[ https://issues.apache.org/jira/browse/HDFS-16016?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17345051#comment-17345051 ]
Viraj Jasani edited comment on HDFS-16016 at 5/16/21, 7:03 AM: --------------------------------------------------------------- {quote}Thanks involve me here. It is good idea. I just wonder if meet some performance issue? {quote} Thanks for the comment [~hexiaoqiao]. I understand this might not be a huge gain in perf but at least better than performing all 3 activities in same thread serially. But anyways if there are any logs or more information on any known perf issue that [~jianghuazhu] can provide, I would also be interested to dig further into it. Anyways, as per my test coverage with hints enabled, I can see IBR triggered by BPServiceActor with significantly higher frequency with the patch compared to without patch (e.g ~1.7 M times hints with patch compared to ~500k times hints without patch) meaning frequency of sending IBR is now higher and non-blocking to main thread of BPServiceActor. Beyond this, I would also be interested to know of any issue or slowness observed in production env (because I have not seen so far in prod). However, hopefully this should provide good improvement overall by making IBR a non blocking activity in BPServiceActor. Thanks was (Author: vjasani): {quote}Thanks involve me here. It is good idea. I just wonder if meet some performance issue? {quote} Thanks for the comment [~hexiaoqiao]. I understand this might not be a huge gain in perf but at least better than performing all 3 activities in same thread serially. But anyways if there are any logs or more information on any known perf issue that [~jianghuazhu] can provide, I would also be interested to dig further into it. Anyways, as per my test coverage with hints enabled, I can see IBR triggered by BPServiceActor with significantly higher frequency with the patch compared to without patch (e.g ~1.7 M times hints with patch compared to ~500k times hints without patch) meaning frequency of sending IBR is now higher and non-blocking to main thread of BPServiceActor. Beyond this, I would also be interested to know of any issue or slowness observed in production env (because I have not seen so far in prod). Thanks > BPServiceActor add a new thread to handle IBR > --------------------------------------------- > > Key: HDFS-16016 > URL: https://issues.apache.org/jira/browse/HDFS-16016 > Project: Hadoop HDFS > Issue Type: Improvement > Reporter: JiangHua Zhu > Assignee: Viraj Jasani > Priority: Minor > Labels: pull-request-available > Time Spent: 1h 40m > Remaining Estimate: 0h > > Now BPServiceActor#offerService() is doing many things, FBR, IBR, heartbeat. > We can handle IBR independently to improve the performance of heartbeat and > FBR. -- This message was sent by Atlassian Jira (v8.3.4#803005) --------------------------------------------------------------------- To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org