[
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 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
