hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Nicolas Liochon (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-9869) Optimize HConnectionManager#getCachedLocation
Date Thu, 14 Nov 2013 18:07:23 GMT

    [ https://issues.apache.org/jira/browse/HBASE-9869?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13822692#comment-13822692
] 

Nicolas Liochon commented on HBASE-9869:
----------------------------------------

I've done some tests with TestClientNoCluster with 10000 regions

||#clients|#puts|time without the patch|time with the patch||
||1 client| 100 million| 94 seconds|65 seconds||
||2 clients| 50 million each| 82 seconds|56 seconds||
||5 clients| 20 million each| 105 seconds|66 seconds||

With 5 clients, we have 10 threads trying to insert as much as possible, so more clients means
more context switches on more memory pressure (it's different if they have to wait for an
answer from a server of course).
I need to do more tests with more regions. But so far so good I would say.

> Optimize HConnectionManager#getCachedLocation
> ---------------------------------------------
>
>                 Key: HBASE-9869
>                 URL: https://issues.apache.org/jira/browse/HBASE-9869
>             Project: HBase
>          Issue Type: Bug
>          Components: Client
>    Affects Versions: 0.98.0, 0.96.0
>            Reporter: Nicolas Liochon
>            Assignee: Nicolas Liochon
>             Fix For: 0.98.0, 0.96.1
>
>         Attachments: 9869.v1.patch, 9869.v1.patch, 9869.v2.patch
>
>
> It javadoc says: "TODO: This method during writing consumes 15% of CPU doing lookup".
This is still true, says Yourkit. With 0.96, we also spend more time in these methods. We
retry more, and the AsyncProcess calls it in parallel.
> I don't have the patch for this yet, but I will spend some time on it.



--
This message was sent by Atlassian JIRA
(v6.1#6144)

Mime
View raw message