[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Nicolas Liochon updated HDFS-347: - Description: One of the major strategies Hadoop uses to get scalable data processing is to move the code to the data. However, putting the DFS client on the same physical node as the data blocks it acts on doesn't improve read performance as much as expected. After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem is due to the HDFS streaming protocol causing many more read I/O operations (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB disk block from the DataNode process (running in a separate JVM) running on the same machine. The DataNode will satisfy the single disk block request by sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, this is done in the sendChunk() method, relying on Java's transferTo() method. Depending on the host O/S and JVM implementation, transferTo() is implemented as either a sendfilev() syscall or a pair of mmap() and write(). In either case, each chunk is read from the disk by issuing a separahitting te I/O operation for each chunk. The result is that the single request for a 64-MB block ends up the disk as over a thousand smaller requests for 64-KB each. Since the DFSClient runs in a different JVM and process than the DataNode, shuttling data from the disk to the DFSClient also results in context switches each time network packets get sent (in this case, the 64-kb chunk turns into a large number of 1500 byte packet send operations). Thus we see a large number of context switches for each block send operation. I'd like to get some feedback on the best way to address this, but I think providing a mechanism for a DFSClient to directly open data blocks that happen to be on the same machine. It could do this by examining the set of LocatedBlocks returned by the NameNode, marking those that should be resident on the local host. Since the DataNode and DFSClient (probably) share the same hadoop configuration, the DFSClient should be able to find the files holding the block data, and it could directly open them and send data back to the client. This would avoid the context switches imposed by the network layer, and would allow for much larger read buffers than 64KB, which should reduce the number of iops imposed by each read block operation. was: One of the major strategies Hadoop uses to get scalable data processing is to move the code to the data. However, putting the DFS client on the same physical node as the data blocks it acts on doesn't improve read performance as much as expected. After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem is due to the HDFS streaming protocol causing many more read I/O operations (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB disk block from the DataNode process (running in a separate JVM) running on the same machine. The DataNode will satisfy the single disk block request by sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, this is done in the sendChunk() method, relying on Java's transferTo() method. Depending on the host O/S and JVM implementation, transferTo() is implemented as either a sendfilev() syscall or a pair of mmap() and write(). In either case, each chunk is read from the disk by issuing a separate I/O operation for each chunk. The result is that the single request for a 64-MB block ends up hitting the disk as over a thousand smaller requests for 64-KB each. Since the DFSClient runs in a different JVM and process than the DataNode, shuttling data from the disk to the DFSClient also results in context switches each time network packets get sent (in this case, the 64-kb chunk turns into a large number of 1500 byte packet send operations). Thus we see a large number of context switches for each block send operation. I'd like to get some feedback on the best way to address this, but I think providing a mechanism for a DFSClient to directly open data blocks that happen to be on the same machine. It could do this by examining the set of LocatedBlocks returned by the NameNode, marking those that should be resident on the local host. Since the DataNode and DFSClient (probably) share the same hadoop configuration, the DFSClient should be able to find the files holding the block data, and it could directly open them and send data back to the client. This would avoid the context switches imposed by the network layer, and would allow for much larger read buffers than 64KB, which should reduce the number of iops imposed by each read block operation. > DFS read performance suboptimal when client co-located on nodes with data > - > > Ke
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Todd Lipcon updated HDFS-347: - Resolution: Fixed Fix Version/s: 2.0.5-beta Hadoop Flags: Reviewed Status: Resolved (was: Patch Available) I've backported this to branch-2. Thanks for the big contribution, Colin and everyone who helped to review and test this. > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Fix For: 3.0.0, 2.0.5-beta > > Attachments: 2013.01.28.design.pdf, 2013.01.31.consolidated2.patch, > 2013.01.31.consolidated.patch, 2013.02.15.consolidated4.patch, > 2013-04-01-jenkins.patch, all.tsv, a.patch, BlockReaderLocal1.txt, > full.patch, HADOOP-4801.1.patch, HADOOP-4801.2.patch, HADOOP-4801.3.patch, > HDFS-347-016_cleaned.patch, HDFS-347.016.patch, HDFS-347.017.clean.patch, > HDFS-347.017.patch, HDFS-347.018.clean.patch, HDFS-347.018.patch2, > HDFS-347.019.patch, HDFS-347.020.patch, HDFS-347.021.patch, > HDFS-347.022.patch, HDFS-347.024.patch, HDFS-347.025.patch, > HDFS-347.026.patch, HDFS-347.027.patch, HDFS-347.029.patch, > HDFS-347.030.patch, HDFS-347.033.patch, HDFS-347.035.patch, > HDFS-347-branch-20-append.txt, hdfs-347-merge.txt, hdfs-347-merge.txt, > hdfs-347-merge.txt, hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Aaron T. Myers updated HDFS-347: Target Version/s: 2.0.5-beta Fix Version/s: 3.0.0 Since the merge vote passed, I have merged the branch to trunk. Leaving the JIRA open for now until we also do the merge to branch-2. Colin, thanks a ton for the monster contribution. This is a long time in coming. > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Fix For: 3.0.0 > > Attachments: 2013.01.28.design.pdf, 2013.01.31.consolidated2.patch, > 2013.01.31.consolidated.patch, 2013.02.15.consolidated4.patch, > 2013-04-01-jenkins.patch, all.tsv, a.patch, BlockReaderLocal1.txt, > full.patch, HADOOP-4801.1.patch, HADOOP-4801.2.patch, HADOOP-4801.3.patch, > HDFS-347-016_cleaned.patch, HDFS-347.016.patch, HDFS-347.017.clean.patch, > HDFS-347.017.patch, HDFS-347.018.clean.patch, HDFS-347.018.patch2, > HDFS-347.019.patch, HDFS-347.020.patch, HDFS-347.021.patch, > HDFS-347.022.patch, HDFS-347.024.patch, HDFS-347.025.patch, > HDFS-347.026.patch, HDFS-347.027.patch, HDFS-347.029.patch, > HDFS-347.030.patch, HDFS-347.033.patch, HDFS-347.035.patch, > HDFS-347-branch-20-append.txt, hdfs-347-merge.txt, hdfs-347-merge.txt, > hdfs-347-merge.txt, hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Tsz Wo (Nicholas), SZE updated HDFS-347: Attachment: a.patch Sure, let's try. > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: 2013.01.28.design.pdf, 2013.01.31.consolidated2.patch, > 2013.01.31.consolidated.patch, 2013.02.15.consolidated4.patch, > 2013-04-01-jenkins.patch, all.tsv, a.patch, BlockReaderLocal1.txt, > full.patch, HADOOP-4801.1.patch, HADOOP-4801.2.patch, HADOOP-4801.3.patch, > HDFS-347-016_cleaned.patch, HDFS-347.016.patch, HDFS-347.017.clean.patch, > HDFS-347.017.patch, HDFS-347.018.clean.patch, HDFS-347.018.patch2, > HDFS-347.019.patch, HDFS-347.020.patch, HDFS-347.021.patch, > HDFS-347.022.patch, HDFS-347.024.patch, HDFS-347.025.patch, > HDFS-347.026.patch, HDFS-347.027.patch, HDFS-347.029.patch, > HDFS-347.030.patch, HDFS-347.033.patch, HDFS-347.035.patch, > HDFS-347-branch-20-append.txt, hdfs-347-merge.txt, hdfs-347-merge.txt, > hdfs-347-merge.txt, hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: 2013-04-01-jenkins.patch patch for jenkins > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: 2013.01.28.design.pdf, 2013.01.31.consolidated2.patch, > 2013.01.31.consolidated.patch, 2013.02.15.consolidated4.patch, > 2013-04-01-jenkins.patch, all.tsv, BlockReaderLocal1.txt, full.patch, > HADOOP-4801.1.patch, HADOOP-4801.2.patch, HADOOP-4801.3.patch, > HDFS-347-016_cleaned.patch, HDFS-347.016.patch, HDFS-347.017.clean.patch, > HDFS-347.017.patch, HDFS-347.018.clean.patch, HDFS-347.018.patch2, > HDFS-347.019.patch, HDFS-347.020.patch, HDFS-347.021.patch, > HDFS-347.022.patch, HDFS-347.024.patch, HDFS-347.025.patch, > HDFS-347.026.patch, HDFS-347.027.patch, HDFS-347.029.patch, > HDFS-347.030.patch, HDFS-347.033.patch, HDFS-347.035.patch, > HDFS-347-branch-20-append.txt, hdfs-347-merge.txt, hdfs-347-merge.txt, > hdfs-347-merge.txt, hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: 2013.02.15.consolidated4.patch > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: 2013.01.28.design.pdf, 2013.01.31.consolidated2.patch, > 2013.01.31.consolidated.patch, 2013.02.15.consolidated4.patch, all.tsv, > BlockReaderLocal1.txt, full.patch, HADOOP-4801.1.patch, HADOOP-4801.2.patch, > HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, HDFS-347.016.patch, > HDFS-347.017.clean.patch, HDFS-347.017.patch, HDFS-347.018.clean.patch, > HDFS-347.018.patch2, HDFS-347.019.patch, HDFS-347.020.patch, > HDFS-347.021.patch, HDFS-347.022.patch, HDFS-347.024.patch, > HDFS-347.025.patch, HDFS-347.026.patch, HDFS-347.027.patch, > HDFS-347.029.patch, HDFS-347.030.patch, HDFS-347.033.patch, > HDFS-347.035.patch, HDFS-347-branch-20-append.txt, hdfs-347-merge.txt, > hdfs-347-merge.txt, hdfs-347-merge.txt, hdfs-347.png, hdfs-347.txt, > local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: (was: 2013.02.15.consolidated4.patch) > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: 2013.01.28.design.pdf, 2013.01.31.consolidated2.patch, > 2013.01.31.consolidated.patch, 2013.02.15.consolidated4.patch, all.tsv, > BlockReaderLocal1.txt, full.patch, HADOOP-4801.1.patch, HADOOP-4801.2.patch, > HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, HDFS-347.016.patch, > HDFS-347.017.clean.patch, HDFS-347.017.patch, HDFS-347.018.clean.patch, > HDFS-347.018.patch2, HDFS-347.019.patch, HDFS-347.020.patch, > HDFS-347.021.patch, HDFS-347.022.patch, HDFS-347.024.patch, > HDFS-347.025.patch, HDFS-347.026.patch, HDFS-347.027.patch, > HDFS-347.029.patch, HDFS-347.030.patch, HDFS-347.033.patch, > HDFS-347.035.patch, HDFS-347-branch-20-append.txt, hdfs-347-merge.txt, > hdfs-347-merge.txt, hdfs-347-merge.txt, hdfs-347.png, hdfs-347.txt, > local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: (was: 2013.02.15.consolidated4.patch) > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: 2013.01.28.design.pdf, 2013.01.31.consolidated2.patch, > 2013.01.31.consolidated.patch, 2013.02.15.consolidated4.patch, all.tsv, > BlockReaderLocal1.txt, full.patch, HADOOP-4801.1.patch, HADOOP-4801.2.patch, > HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, HDFS-347.016.patch, > HDFS-347.017.clean.patch, HDFS-347.017.patch, HDFS-347.018.clean.patch, > HDFS-347.018.patch2, HDFS-347.019.patch, HDFS-347.020.patch, > HDFS-347.021.patch, HDFS-347.022.patch, HDFS-347.024.patch, > HDFS-347.025.patch, HDFS-347.026.patch, HDFS-347.027.patch, > HDFS-347.029.patch, HDFS-347.030.patch, HDFS-347.033.patch, > HDFS-347.035.patch, HDFS-347-branch-20-append.txt, hdfs-347-merge.txt, > hdfs-347-merge.txt, hdfs-347-merge.txt, hdfs-347.png, hdfs-347.txt, > local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: 2013.02.15.consolidated4.patch > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: 2013.01.28.design.pdf, 2013.01.31.consolidated2.patch, > 2013.01.31.consolidated.patch, 2013.02.15.consolidated4.patch, all.tsv, > BlockReaderLocal1.txt, full.patch, HADOOP-4801.1.patch, HADOOP-4801.2.patch, > HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, HDFS-347.016.patch, > HDFS-347.017.clean.patch, HDFS-347.017.patch, HDFS-347.018.clean.patch, > HDFS-347.018.patch2, HDFS-347.019.patch, HDFS-347.020.patch, > HDFS-347.021.patch, HDFS-347.022.patch, HDFS-347.024.patch, > HDFS-347.025.patch, HDFS-347.026.patch, HDFS-347.027.patch, > HDFS-347.029.patch, HDFS-347.030.patch, HDFS-347.033.patch, > HDFS-347.035.patch, HDFS-347-branch-20-append.txt, hdfs-347-merge.txt, > hdfs-347-merge.txt, hdfs-347-merge.txt, hdfs-347.png, hdfs-347.txt, > local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: 2013.02.15.consolidated4.patch > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: 2013.01.28.design.pdf, 2013.01.31.consolidated2.patch, > 2013.01.31.consolidated.patch, 2013.02.15.consolidated4.patch, all.tsv, > BlockReaderLocal1.txt, full.patch, HADOOP-4801.1.patch, HADOOP-4801.2.patch, > HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, HDFS-347.016.patch, > HDFS-347.017.clean.patch, HDFS-347.017.patch, HDFS-347.018.clean.patch, > HDFS-347.018.patch2, HDFS-347.019.patch, HDFS-347.020.patch, > HDFS-347.021.patch, HDFS-347.022.patch, HDFS-347.024.patch, > HDFS-347.025.patch, HDFS-347.026.patch, HDFS-347.027.patch, > HDFS-347.029.patch, HDFS-347.030.patch, HDFS-347.033.patch, > HDFS-347.035.patch, HDFS-347-branch-20-append.txt, hdfs-347-merge.txt, > hdfs-347-merge.txt, hdfs-347-merge.txt, hdfs-347.png, hdfs-347.txt, > local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: (was: 2012.02.15.consolidated3.patch) > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: 2013.01.28.design.pdf, 2013.01.31.consolidated2.patch, > 2013.01.31.consolidated.patch, 2013.02.15.consolidated4.patch, all.tsv, > BlockReaderLocal1.txt, full.patch, HADOOP-4801.1.patch, HADOOP-4801.2.patch, > HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, HDFS-347.016.patch, > HDFS-347.017.clean.patch, HDFS-347.017.patch, HDFS-347.018.clean.patch, > HDFS-347.018.patch2, HDFS-347.019.patch, HDFS-347.020.patch, > HDFS-347.021.patch, HDFS-347.022.patch, HDFS-347.024.patch, > HDFS-347.025.patch, HDFS-347.026.patch, HDFS-347.027.patch, > HDFS-347.029.patch, HDFS-347.030.patch, HDFS-347.033.patch, > HDFS-347.035.patch, HDFS-347-branch-20-append.txt, hdfs-347-merge.txt, > hdfs-347-merge.txt, hdfs-347-merge.txt, hdfs-347.png, hdfs-347.txt, > local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: 2012.02.15.consolidated3.patch > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: 2012.02.15.consolidated3.patch, 2013.01.28.design.pdf, > 2013.01.31.consolidated2.patch, 2013.01.31.consolidated.patch, all.tsv, > BlockReaderLocal1.txt, full.patch, HADOOP-4801.1.patch, HADOOP-4801.2.patch, > HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, HDFS-347.016.patch, > HDFS-347.017.clean.patch, HDFS-347.017.patch, HDFS-347.018.clean.patch, > HDFS-347.018.patch2, HDFS-347.019.patch, HDFS-347.020.patch, > HDFS-347.021.patch, HDFS-347.022.patch, HDFS-347.024.patch, > HDFS-347.025.patch, HDFS-347.026.patch, HDFS-347.027.patch, > HDFS-347.029.patch, HDFS-347.030.patch, HDFS-347.033.patch, > HDFS-347.035.patch, HDFS-347-branch-20-append.txt, hdfs-347-merge.txt, > hdfs-347-merge.txt, hdfs-347-merge.txt, hdfs-347.png, hdfs-347.txt, > local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: 2013.01.31.consolidated2.patch > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: 2013.01.28.design.pdf, 2013.01.31.consolidated2.patch, > 2013.01.31.consolidated.patch, all.tsv, BlockReaderLocal1.txt, full.patch, > HADOOP-4801.1.patch, HADOOP-4801.2.patch, HADOOP-4801.3.patch, > HDFS-347-016_cleaned.patch, HDFS-347.016.patch, HDFS-347.017.clean.patch, > HDFS-347.017.patch, HDFS-347.018.clean.patch, HDFS-347.018.patch2, > HDFS-347.019.patch, HDFS-347.020.patch, HDFS-347.021.patch, > HDFS-347.022.patch, HDFS-347.024.patch, HDFS-347.025.patch, > HDFS-347.026.patch, HDFS-347.027.patch, HDFS-347.029.patch, > HDFS-347.030.patch, HDFS-347.033.patch, HDFS-347.035.patch, > HDFS-347-branch-20-append.txt, hdfs-347-merge.txt, hdfs-347-merge.txt, > hdfs-347-merge.txt, hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: 2013.01.31.consolidated.patch consolidated patch > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: 2013.01.28.design.pdf, 2013.01.31.consolidated.patch, > all.tsv, BlockReaderLocal1.txt, full.patch, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch2, HDFS-347.019.patch, > HDFS-347.020.patch, HDFS-347.021.patch, HDFS-347.022.patch, > HDFS-347.024.patch, HDFS-347.025.patch, HDFS-347.026.patch, > HDFS-347.027.patch, HDFS-347.029.patch, HDFS-347.030.patch, > HDFS-347.033.patch, HDFS-347.035.patch, HDFS-347-branch-20-append.txt, > hdfs-347-merge.txt, hdfs-347-merge.txt, hdfs-347-merge.txt, hdfs-347.png, > hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: 2013.01.28.design.pdf update design document > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: 2013.01.28.design.pdf, all.tsv, BlockReaderLocal1.txt, > full.patch, HADOOP-4801.1.patch, HADOOP-4801.2.patch, HADOOP-4801.3.patch, > HDFS-347-016_cleaned.patch, HDFS-347.016.patch, HDFS-347.017.clean.patch, > HDFS-347.017.patch, HDFS-347.018.clean.patch, HDFS-347.018.patch2, > HDFS-347.019.patch, HDFS-347.020.patch, HDFS-347.021.patch, > HDFS-347.022.patch, HDFS-347.024.patch, HDFS-347.025.patch, > HDFS-347.026.patch, HDFS-347.027.patch, HDFS-347.029.patch, > HDFS-347.030.patch, HDFS-347.033.patch, HDFS-347.035.patch, > HDFS-347-branch-20-append.txt, hdfs-347-merge.txt, hdfs-347-merge.txt, > hdfs-347-merge.txt, hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: full.patch > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, full.patch, > HADOOP-4801.1.patch, HADOOP-4801.2.patch, HADOOP-4801.3.patch, > HDFS-347-016_cleaned.patch, HDFS-347.016.patch, HDFS-347.017.clean.patch, > HDFS-347.017.patch, HDFS-347.018.clean.patch, HDFS-347.018.patch2, > HDFS-347.019.patch, HDFS-347.020.patch, HDFS-347.021.patch, > HDFS-347.022.patch, HDFS-347.024.patch, HDFS-347.025.patch, > HDFS-347.026.patch, HDFS-347.027.patch, HDFS-347.029.patch, > HDFS-347.030.patch, HDFS-347.033.patch, HDFS-347.035.patch, > HDFS-347-branch-20-append.txt, hdfs-347-merge.txt, hdfs-347-merge.txt, > hdfs-347-merge.txt, hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Todd Lipcon updated HDFS-347: - Attachment: hdfs-347-merge.txt Attaching another merge patch for Jenkins to run upstream, since both trunk and the branch have had a few changes since the last QA bot run. > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch2, HDFS-347.019.patch, > HDFS-347.020.patch, HDFS-347.021.patch, HDFS-347.022.patch, > HDFS-347.024.patch, HDFS-347.025.patch, HDFS-347.026.patch, > HDFS-347.027.patch, HDFS-347.029.patch, HDFS-347.030.patch, > HDFS-347.033.patch, HDFS-347.035.patch, HDFS-347-branch-20-append.txt, > hdfs-347-merge.txt, hdfs-347-merge.txt, hdfs-347-merge.txt, hdfs-347.png, > hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Todd Lipcon updated HDFS-347: - Attachment: hdfs-347-merge.txt New merge patch for jenkins, which incorporates Colin's fix for the findbugs as well as a couple other miscellanea. I also fixed the RAT exclude list to ignore CHANGES.HDFS-347.txt. > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch2, HDFS-347.019.patch, > HDFS-347.020.patch, HDFS-347.021.patch, HDFS-347.022.patch, > HDFS-347.024.patch, HDFS-347.025.patch, HDFS-347.026.patch, > HDFS-347.027.patch, HDFS-347.029.patch, HDFS-347.030.patch, > HDFS-347.033.patch, HDFS-347.035.patch, HDFS-347-branch-20-append.txt, > hdfs-347-merge.txt, hdfs-347-merge.txt, hdfs-347.png, hdfs-347.txt, > local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Todd Lipcon updated HDFS-347: - Attachment: hdfs-347-merge.txt Attached a consolidated patch from trunk to the branch (git diff from 8360a7a6a4497c47cf6a389a2663a4a2b4867a19..681737e78ba0ce574b92ff0ef3bd1794492af27e). (The actual merge will be done with an svn merge command, rather than applying this patch -- this is just to get a full Jenkins run). > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch2, HDFS-347.019.patch, > HDFS-347.020.patch, HDFS-347.021.patch, HDFS-347.022.patch, > HDFS-347.024.patch, HDFS-347.025.patch, HDFS-347.026.patch, > HDFS-347.027.patch, HDFS-347.029.patch, HDFS-347.030.patch, > HDFS-347.033.patch, HDFS-347.035.patch, HDFS-347-branch-20-append.txt, > hdfs-347-merge.txt, hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Status: Patch Available (was: Open) > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch2, HDFS-347.019.patch, > HDFS-347.020.patch, HDFS-347.021.patch, HDFS-347.022.patch, > HDFS-347.024.patch, HDFS-347.025.patch, HDFS-347.026.patch, > HDFS-347.027.patch, HDFS-347.029.patch, HDFS-347.030.patch, > HDFS-347.033.patch, HDFS-347.035.patch, HDFS-347-branch-20-append.txt, > hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347.035.patch bugfixes. > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch2, HDFS-347.019.patch, > HDFS-347.020.patch, HDFS-347.021.patch, HDFS-347.022.patch, > HDFS-347.024.patch, HDFS-347.025.patch, HDFS-347.026.patch, > HDFS-347.027.patch, HDFS-347.029.patch, HDFS-347.030.patch, > HDFS-347.033.patch, HDFS-347.035.patch, HDFS-347-branch-20-append.txt, > hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347.033.patch This is an updated patch which gives you some idea of how we can eliminate the Socket dependency as we talked about earlier. The JNI part still needs some work, mostly moving a few things around. > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch2, HDFS-347.019.patch, > HDFS-347.020.patch, HDFS-347.021.patch, HDFS-347.022.patch, > HDFS-347.024.patch, HDFS-347.025.patch, HDFS-347.026.patch, > HDFS-347.027.patch, HDFS-347.029.patch, HDFS-347.030.patch, > HDFS-347.033.patch, HDFS-347-branch-20-append.txt, hdfs-347.png, > hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Status: Open (was: Patch Available) > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch2, HDFS-347.019.patch, > HDFS-347.020.patch, HDFS-347.021.patch, HDFS-347.022.patch, > HDFS-347.024.patch, HDFS-347.025.patch, HDFS-347.026.patch, > HDFS-347.027.patch, HDFS-347.029.patch, HDFS-347.030.patch, > HDFS-347.033.patch, HDFS-347-branch-20-append.txt, hdfs-347.png, > hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347.030.patch this revision fixes an issue with the socket path permission checking > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch2, HDFS-347.019.patch, > HDFS-347.020.patch, HDFS-347.021.patch, HDFS-347.022.patch, > HDFS-347.024.patch, HDFS-347.025.patch, HDFS-347.026.patch, > HDFS-347.027.patch, HDFS-347.029.patch, HDFS-347.030.patch, > HDFS-347-branch-20-append.txt, hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347.029.patch fixes: * put DomainSockets back in the cache right after constructing a {{BlockReaderLocal}}, rather than holding them for the lifetime of the reader. * disable short-circuit local reads for blocks under construction * add back support for old path-based RPC > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch2, HDFS-347.019.patch, > HDFS-347.020.patch, HDFS-347.021.patch, HDFS-347.022.patch, > HDFS-347.024.patch, HDFS-347.025.patch, HDFS-347.026.patch, > HDFS-347.027.patch, HDFS-347.029.patch, HDFS-347-branch-20-append.txt, > hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347.027.patch This doesn't address all the points in the reviewboard (still working on another rev which does.) However it does have the path security validation, the addition of {{dfs.client.domain.socket.data.traffic}}, some refactoring of BlockReaderFactory and the addition of DomainSocketFactory, and renaming of {{getBindPath}} to {{getBoundPath}}. > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch2, HDFS-347.019.patch, > HDFS-347.020.patch, HDFS-347.021.patch, HDFS-347.022.patch, > HDFS-347.024.patch, HDFS-347.025.patch, HDFS-347.026.patch, > HDFS-347.027.patch, HDFS-347-branch-20-append.txt, hdfs-347.png, > hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347.026.patch address todd's comments (see reviewboard) > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: datanode, hdfs-client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch2, HDFS-347.019.patch, > HDFS-347.020.patch, HDFS-347.021.patch, HDFS-347.022.patch, > HDFS-347.024.patch, HDFS-347.025.patch, HDFS-347.026.patch, > HDFS-347-branch-20-append.txt, hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347.025.patch It looks like junit needs some prodding not to try to directly execute the methods in {{TestParallelReadUtil}} > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: data-node, hdfs client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch2, HDFS-347.019.patch, > HDFS-347.020.patch, HDFS-347.021.patch, HDFS-347.022.patch, > HDFS-347.024.patch, HDFS-347.025.patch, HDFS-347-branch-20-append.txt, > hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347.024.patch * Use the previous BlockReaderLocal code, with a few updates. * test some different combinations: UNIX domain sockets UNIX domain sockets + short circuit UNIX domain sockets + short circuit + skip checksum > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: data-node, hdfs client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch2, HDFS-347.019.patch, > HDFS-347.020.patch, HDFS-347.021.patch, HDFS-347.022.patch, > HDFS-347.024.patch, HDFS-347-branch-20-append.txt, hdfs-347.png, > hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347.022.patch fix warnings > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: data-node, hdfs client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch2, HDFS-347.019.patch, > HDFS-347.020.patch, HDFS-347.021.patch, HDFS-347.022.patch, > HDFS-347-branch-20-append.txt, hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347.021.patch * rebase on HADOOP-6311 fixes * nicer logs when short-circuit can't be enabled * use {{TemporarySocketDirectory}} in {{TestParallelLocalRead}}, {{TestShortCircuitLocalRead}} > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: data-node, hdfs client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch2, HDFS-347.019.patch, > HDFS-347.020.patch, HDFS-347.021.patch, HDFS-347-branch-20-append.txt, > hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347.020.patch This is an updated patch based on a discussion we had in HADOOP-6311. Basically, the current design is to pass file descriptors over a new class named {{DomainSocket}}, which represents UNIX domain sockets. This is accomplished by adding a new message to the {{DataTransferProtocol}}, {{RequestShortCircuitFd}}. The {{DataXceiverServer}} can manage these UNIX domain sockets just as easily as it manages existing the IPv4 sockets, because they implement the same interfaces. One thing I refactored in this patch is {{BlockReaderFactory}}. It formerly contained only static methods; this patch changes it to be a "real" class with instance methods and instance data. I felt that the {{BlockReaderFactory}} methods were getting too unwieldy because we were passing a tremendous amount of parameters, many of which could be considered properties of the factory in a sense. Using instance data also allows the factory to keep a blacklist of which {{DataNodes}} do not support file descriptor passing. It uses this information to avoid making unnecesary requests. This patch also introduces the concept of a format version number for blocks. The idea here is that if we later change the block format on-disk, we want to be able to tell clients that they can't short-circuit access these blocks unless they can understand the corresponding version number. (One change we've talked a lot about doing in the past is merging block data and metadata files.) This makes it possible to have a cluster where you have some block files in one format and some in another-- a necessity for doing a real-world transition. The clients are passed the version number, so they can act intelligently-- or simply refuse to read the newer formats if they don't know how. Because this patch depends on the {{DomainSocket}} code, it currently incorporates that code. HADOOP-6311 is the best place to comment about {{DomainSocket}}, since that is what that JIRA is about. > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: data-node, hdfs client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch2, HDFS-347.019.patch, > HDFS-347.020.patch, HDFS-347-branch-20-append.txt, hdfs-347.png, > hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be reside
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347.019.patch * rebase on HDFS-347 * DataNode#shutdown should close the fdServer so that all published file descriptors are closed. > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: data-node, hdfs client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch2, HDFS-347.019.patch, > HDFS-347-branch-20-append.txt, hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347.018.patch2 > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: data-node, hdfs client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch2, HDFS-347-branch-20-append.txt, > hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: (was: HDFS-347.018.patch) > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: data-node, hdfs client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch2, HDFS-347-branch-20-append.txt, > hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347.018.patch re-submit > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: data-node, hdfs client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch, HDFS-347-branch-20-append.txt, > hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: (was: HDFS-347.018.patch) > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: data-node, hdfs client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch, HDFS-347-branch-20-append.txt, > hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347.018.patch New combined HADOOP-6311 + HDFS-347 patch for jenkins. > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: data-node, hdfs client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347.018.patch, HDFS-347-branch-20-append.txt, > hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347.018.clean.patch * SuppressWarnings("deprecation") seems to not work; skip it, and just don't mark BlockLocalPathInfo as {{\@deprecated}}, to avoid creating a lot of warnings. * re-arrange the InputStream close methods in hopes of placating findbugs. This doesn't fix any bugs, but hopefully it quiets it down. * TestBlockReaderLocal: when reading the checksum is disabled, we should not expect to detect checksum errors. > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: data-node, hdfs client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347.018.clean.patch, HDFS-347-branch-20-append.txt, hdfs-347.png, > hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347.017.patch * The combined version for Jenkins to test. > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: data-node, hdfs client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347.017.patch, > HDFS-347-branch-20-append.txt, hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347.017.clean.patch This verison fixes some bugs with the fallback case where the JNI libraries are not installed. It also adds a few more junit tests. > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: data-node, hdfs client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347.017.clean.patch, HDFS-347-branch-20-append.txt, > hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Assignee: Colin Patrick McCabe (was: Todd Lipcon) Status: Patch Available (was: Open) > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: data-node, hdfs client, performance >Reporter: George Porter >Assignee: Colin Patrick McCabe > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347-branch-20-append.txt, hdfs-347.png, > hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347.016.patch > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: data-node, hdfs client, performance >Reporter: George Porter >Assignee: Todd Lipcon > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347-branch-20-append.txt, hdfs-347.png, > hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: (was: HDFS-347.016.patch) > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: data-node, hdfs client, performance >Reporter: George Porter >Assignee: Todd Lipcon > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347-branch-20-append.txt, hdfs-347.png, > hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347-016_cleaned.patch This patch only includes HDFS-347. * DataChecksum#newDataChecksum: correctly handle offset values other than 0. * BlockReader / BlockReaderUtil: add skipFully and available methods. Add JavaDoc for skip method. The available method returns a rough approximation of how much data might be available without doing any more network I/O. This helps us optimize in the case where we are reading from a local file descriptor, since we never do network I/O in that case. * BlockReaderLocal: simpler implementation that uses raw FileChannel objects. We don't need to cache anything, or make RPCs to the DataNode. * DFSClient / DFSInputStream: update getLocalBlockReader to work with fd passing. Rather than overloading AccessControlException to mean "local reads were not enabled," create a new exception called LocalReadsDisabledException and throw it when that is the case. This will prevent confusion in the future. Use skipFully instead of skip, since the latter may give us short skips. * DFSConfigKeys: don't need dfs.block.local-path-access.user any more. Local reads are now on by default rather than disabled by default. * RPC stuff: add BlockLocalFdInfo. Deprecate BlockLocalPathInfo. Implement the DataNode, FsDatasetIMpl, etc. methods. Add GetBlockLocalFdInfoResponseProto. The old RPC is now deprecated and will always throw an AccessControlException, so that older clients will fall back to remote reads. * MiniDFSCluster: add getBlockMetadataFile which is like getBlockFile except that it returns .meta files. * Tests: TestBlockReaderLocal now includes more tests of BlockReaderLocal in isolation. TestParallelRead now explictly disables local reads (that case is testsed by TestParalellLocalRead). TestShortCircuitLocalRead: add testDeprecatedGetBlockLocalPathInfoRpc to test the deprecated RPC. > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: data-node, hdfs client, performance >Reporter: George Porter >Assignee: Todd Lipcon > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-016_cleaned.patch, > HDFS-347.016.patch, HDFS-347-branch-20-append.txt, hdfs-347.png, > hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the c
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Colin Patrick McCabe updated HDFS-347: -- Attachment: HDFS-347.016.patch * testable patch including HDFS-3753, HADOOP-6311 and HDFS-347 > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: data-node, hdfs client, performance >Reporter: George Porter >Assignee: Todd Lipcon > Attachments: all.tsv, BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347.016.patch, > HDFS-347-branch-20-append.txt, hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Todd Lipcon updated HDFS-347: - Component/s: performance hdfs client data-node > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement > Components: data-node, hdfs client, performance >Reporter: George Porter >Assignee: Todd Lipcon > Attachments: BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-branch-20-append.txt, > all.tsv, hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] Updated: (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] ryan rawson updated HDFS-347: - Attachment: HDFS-347-branch-20-append.txt applies to head of branch-20-append > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement >Reporter: George Porter >Assignee: Todd Lipcon > Attachments: BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, HDFS-347-branch-20-append.txt, > all.tsv, hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. - For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] Updated: (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] dhruba borthakur updated HDFS-347: -- Attachment: BlockReaderLocal1.txt This is a patch that I have done to the 0.20 branch to improve hbase performance. HBase caches the read-blocks from HDFS in a block cache. If the block is local, it is also cached by the local filesystem in the OS cache. I am internally referring this patch as the "Killer of the HBase block cache". Our experiments are still continuing, but this patch (and a bunch of other habse patches) will probably mean that HBase does not need to cache local blocks in the block cache anymore! thanks is a huge huge gain! > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement >Reporter: George Porter >Assignee: Todd Lipcon > Attachments: BlockReaderLocal1.txt, HADOOP-4801.1.patch, > HADOOP-4801.2.patch, HADOOP-4801.3.patch, all.tsv, hdfs-347.png, > hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. - For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] Updated: (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Todd Lipcon updated HDFS-347: - Attachment: hdfs-347.png all.tsv Took some time to rebase this work against trunk (with HADOOP-5205 and HDFS-755 patched in as well). Here's a graph (and the data that made it) comparing the following: - checksumfs.tsv - reading a file:/// URL with an associated checksum file on my local disk - raw.tsv - reading the same file, but with no checksum file - without.tsv - pseudo-distributed HDFS with dfs.client.use.unix.sockets=false - with.tsv - same HDFS, but with dfs.client.use.unix.sockets=true For all of these tests, I used a 691MB file, and double checked md5sum output to make sure they were all reading it correctly. Each box plot shows the distribution of 50 trials of fs -cat /path/to/file. io.file.buffer.size was set to 64K for all trials. The big surprise here is that somehow HDFS with this patch came out faster than ChecksumFileSystem. The sys time for the same doesn't show any difference, but HDFS is using less CPU time. Since this doesn't make much sense, I reran both the HDFS and ChecksumFs benchmarks a second time and the results were the same. If anyone cares to wager a guess about how this could be possible, I'd appreciate it :) Otherwise, I will try to dig into this. The inclusion of raw shows the same 200-300% difference referenced in earlier comments in this jira. There's no optimization we can make here aside from speeding up checksumming. The HADOOP-5205/HDFS-755 patches improved this a bit, but it's still the major difference. As noted above, this patch makes reading from the local DN perform at least as well as reading from a local checksummed system (if not inexplicably better). > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement >Reporter: George Porter >Assignee: Todd Lipcon > Attachments: all.tsv, HADOOP-4801.1.patch, HADOOP-4801.2.patch, > HADOOP-4801.3.patch, hdfs-347.png, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online.
[jira] Updated: (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Todd Lipcon updated HDFS-347: - Attachment: hdfs-347.txt Here's a patch that implements the design as detailed above. More benchmarks and discussion of remaining work to come. This depends on the core jars built from HADOOP-6311, plus running ant like: ant -Djava.library.path=/home/todd/git/hadoop-common/build/native/Linux-amd64-64/lib/ run-dfsclient-test > DFS read performance suboptimal when client co-located on nodes with data > - > > Key: HDFS-347 > URL: https://issues.apache.org/jira/browse/HDFS-347 > Project: Hadoop HDFS > Issue Type: Improvement >Reporter: George Porter > Attachments: HADOOP-4801.1.patch, HADOOP-4801.2.patch, > HADOOP-4801.3.patch, hdfs-347.txt, local-reads-doc > > > One of the major strategies Hadoop uses to get scalable data processing is to > move the code to the data. However, putting the DFS client on the same > physical node as the data blocks it acts on doesn't improve read performance > as much as expected. > After looking at Hadoop and O/S traces (via HADOOP-4049), I think the problem > is due to the HDFS streaming protocol causing many more read I/O operations > (iops) than necessary. Consider the case of a DFSClient fetching a 64 MB > disk block from the DataNode process (running in a separate JVM) running on > the same machine. The DataNode will satisfy the single disk block request by > sending data back to the HDFS client in 64-KB chunks. In BlockSender.java, > this is done in the sendChunk() method, relying on Java's transferTo() > method. Depending on the host O/S and JVM implementation, transferTo() is > implemented as either a sendfilev() syscall or a pair of mmap() and write(). > In either case, each chunk is read from the disk by issuing a separate I/O > operation for each chunk. The result is that the single request for a 64-MB > block ends up hitting the disk as over a thousand smaller requests for 64-KB > each. > Since the DFSClient runs in a different JVM and process than the DataNode, > shuttling data from the disk to the DFSClient also results in context > switches each time network packets get sent (in this case, the 64-kb chunk > turns into a large number of 1500 byte packet send operations). Thus we see > a large number of context switches for each block send operation. > I'd like to get some feedback on the best way to address this, but I think > providing a mechanism for a DFSClient to directly open data blocks that > happen to be on the same machine. It could do this by examining the set of > LocatedBlocks returned by the NameNode, marking those that should be resident > on the local host. Since the DataNode and DFSClient (probably) share the > same hadoop configuration, the DFSClient should be able to find the files > holding the block data, and it could directly open them and send data back to > the client. This would avoid the context switches imposed by the network > layer, and would allow for much larger read buffers than 64KB, which should > reduce the number of iops imposed by each read block operation. -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online.
[jira] Updated: (HDFS-347) DFS read performance suboptimal when client co-located on nodes with data
[ https://issues.apache.org/jira/browse/HDFS-347?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Todd Lipcon updated HDFS-347: - Attachment: local-reads-doc Attaching v1 of a design document for this feature. This does not include a test plan - that will follow once implementation has gone a bit further. Pasting the design doc below as well: -- h1. Design Document: Local Read Optimization h2. Problem Definition Currently, when the DFS Client is located on the same physical node as the DataNode serving the data, it does not use this knowledge to its advantage. All blocks are read through the same protocol based on a TCP connection. Early experimentation has shown that this has a 20-30% overhead when compared with reading the block files directly off the local disk. This JIRA seeks to improve the performance of node-local reads by providing a fast path that is enabled in this case. This case is very common, especially in the context of MapReduce jobs where tasks are scheduled local to their data. Although writes are likely to see an improvement here too, this JIRA will focus only on the read path. The write path is significantly more complicated due to write pipeline recovery, append support, etc. Additionally, the write path will still have to go over TCP to the non-local replicas, so the throughput improvements will probably not be as marked. h2. Use Cases # As mentioned above, the majority of data read during a MapReduce job tends to be from local datanodes. This optimization should improve MapReduce performance of read-constrained jobs significantly. # Random reads should see a significant performance benefit with this patch as well. Applications such as the HBase Region Server should see a very large improvement. Users will not have to make any specific changes to use the performance improvement - the optimization should be transparent and retain all existing semantics. h2. Interaction with Current System This behavior needs modifications in two areas: h3. DataNode The datanode needs to be extended to provide access to the local block storage to the reading client. h3. DFSInputStream DFSInputStream needs to be extended in order to enable the fast read path when reading from local datanodes. h2. Requirements h3. Unix Domain Sockets via JNI In order to maintain security, we cannot simply have the reader access blocks through the local filesystem. The reader may be running as an arbitrary user ID, and we should not require world-readable permissions on the block storage. Unix domain sockets offer the ability to transport already-open file descriptors from one peer to another using the "ancillary data" construct and the sendmsg(2) system call. This ability is documented in unix(7) under the SCM_RIGHTS section. Unix domain sockets are unfortunately not available in Java. We will need to employ JNI to access the appropriate system calls. h3. Modify DFSClient/DataNode interaction The DFS Client will need to be able to initiate the fast path read when it detects it is connecting to a local DataNode. The DataNode needs to respond to this request by providing the appropriate file descriptors or by reverting to the normal slow path if the functionality has been administratively disabled, etc. h2. Design h3. Unix Domain Sockets in Java The Android open source project currently includes support for Unix Domain Sockets in the android.net package. It also includes the native JNI code to implement these classes. Android is Apache 2.0 licensed and thus we can freely use the code in Hadoop. The Android project relies on a lot of custom build infrastructure and utility functions. In order to reduce our dependencies, we will copy the appropriate classes into a new org.apache.hadoop.net.unix package. We will include the appropriate JNI code in the existing libhadoop library. If HADOOP-4998 (native runtime library for Hadoop) progresses in the near term, we could include this functionality there. The JNI code needs small modifications to work properly in the Hadoop build system without pulling in a large number of Android dependencies. h3. Fast path initiation When DFSInputStream is connecting to a node, it can determine whether that node is local by simply inspecting the IP address. In the event that it is a local host and the fast path has not been prohibited by the Configuration, the fast path will be initiated. The fast path is simply a different BlockReader implementation. h3. Fast path interface BlockReader will become an interface, with the current implementation being renamed to RemoteBlockReader. The fast-path for local reads will be a LocalBlockReader, which is instantiated after it has been determined that the target datanode is local. h3. Fast path mechanism Currently, when the DFSInputStream connects to the DataNode, it sends OP_READ_BLOCK, including the access token, blo