[
https://issues.apache.org/jira/browse/HBASE-10156?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13861184#comment-13861184
]
stack commented on HBASE-10156:
-------------------------------
[~xieliang007] Yeah. Did more tests today. For single thread, the patch is
~8% faster which would take us about back to where we used to be for the
single-threaded case. For 250 threads, we seem to be 50% faster with the
patch. BUT, for ten threads we are the same (the patch is actually slightly
slower by a few percent). Let me figure what is going on here. Also, doing a
dumb dfsclient write/sync/write/sync/write/sync and comparing, there is a big
gap between what it can do and what our WAL can do throughput-wise; our WAL has
a way more going on but am trying to see if there is any more to be milked here.
I'll clean up the patch and describe what it does better after I finish a few
more experiments.
> Fix up the HBASE-8755 slowdown when low contention
> --------------------------------------------------
>
> Key: HBASE-10156
> URL: https://issues.apache.org/jira/browse/HBASE-10156
> Project: HBase
> Issue Type: Sub-task
> Components: wal
> Reporter: stack
> Assignee: stack
> Attachments: 10156.txt, 10156v2.txt, Disrupting.java
>
>
> HBASE-8755 slows our writes when only a few clients. Fix.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
