[ https://issues.apache.org/jira/browse/HDFS-5776?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13877145#comment-13877145 ]
Liang Xie commented on HDFS-5776: --------------------------------- [~saint....@gmail.com], [~cmccabe], hedged reads doesn't need to be aware of weather local read or not. pread -> fetchBlockByteRange/fetchBlockByteRangeSpeculative -> actualGetFromOneDataNode -> getBlockReader we don't need to handle weather local reader or not, the hedged reads only focus on picking a secondary dn, then wait the winner. In my test, the region server(rs) instance was in the same box with one dn(say dn1). Take my first case(dfs.dfsclient.hedged.read.threshold.millis = 500ms) for example, in that case, in deed, nearly all of the preads go to the local dn1, i could observe io util% always is 100%(really io bound, aha, btw, i filed HDFS-5727, hope with that we can alleviate the io somehow, i am planning to dig that jira after this done, if you guys have any comments, just put there) during the testing, and dn2/dn3's io util% keep 0%. and in following cases, i could see the io util% of dn2/dn3 began to increment to around 5%~20%. > Support 'hedged' reads in DFSClient > ----------------------------------- > > Key: HDFS-5776 > URL: https://issues.apache.org/jira/browse/HDFS-5776 > Project: Hadoop HDFS > Issue Type: Improvement > Components: hdfs-client > Affects Versions: 3.0.0 > Reporter: Liang Xie > Assignee: Liang Xie > Attachments: HDFS-5776-v2.txt, HDFS-5776-v3.txt, HDFS-5776.txt > > > This is a placeholder of hdfs related stuff backport from > https://issues.apache.org/jira/browse/HBASE-7509 > The quorum read ability should be helpful especially to optimize read outliers > we can utilize "dfs.dfsclient.quorum.read.threshold.millis" & > "dfs.dfsclient.quorum.read.threadpool.size" to enable/disable the hedged read > ability from client side(e.g. HBase), and by using DFSQuorumReadMetrics, we > could export the interested metric valus into client system(e.g. HBase's > regionserver metric). > The core logic is in pread code path, we decide to goto the original > fetchBlockByteRange or the new introduced fetchBlockByteRangeSpeculative per > the above config items. -- This message was sent by Atlassian JIRA (v6.1.5#6160)