hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Anoop Sam John (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-15509) Avoid copy of block data in HFileBlock$Writer.finishBlock
Date Wed, 30 Mar 2016 16:30:25 GMT

    [ https://issues.apache.org/jira/browse/HBASE-15509?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15218260#comment-15218260

Anoop Sam John commented on HBASE-15509:

bq.This patch + 4 more in HBASE-15542 will allow us, potentially, reduce memory allocation
during compaction by more than 90% (10x times). That, in turn, will allow us to run compactions
more aggressively, without throughput controller and with 5-10 threads in parallel. This is
the eventual goal.

So any chance for a test which measures the compaction speed up because of it?  And mainly
the GC graphs.  Ya this will surely increase the GC interval and reduce the #GCs..   What
am interested in seeing is whether the overall GC pause how that will be affected.  Because
some of the tests we did shows that the GC pause increased even if #GCs less.. So accumulated
pause is almost same !    Which all patches u mean which make the compaction allocation reduce
by 90%. I will try to apply those also in my write path perf tests.   Some of these should
help in flush time also right?

> Avoid copy of block data in HFileBlock$Writer.finishBlock
> ---------------------------------------------------------
>                 Key: HBASE-15509
>                 URL: https://issues.apache.org/jira/browse/HBASE-15509
>             Project: HBase
>          Issue Type: Sub-task
>          Components: HFile
>            Reporter: Vladimir Rodionov
>            Assignee: Vladimir Rodionov
>         Attachments: HBASE-15509-v1.patch, HBASE-15509-v2.patch
> This calls ByteArrayOutputStream.toByteArray() which creates a copy of data block. 

This message was sent by Atlassian JIRA

View raw message