[
https://issues.apache.org/jira/browse/HBASE-16502?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15436830#comment-15436830
]
binlijin edited comment on HBASE-16502 at 8/25/16 1:16 PM:
-----------------------------------------------------------
I am doing HBASE-16213, and the HBASE-16213_branch1_v3.patch have the same
problem, when i test the E2E performance, i find valueLength=1k, the
performance slow than NONE encoding, and i find the difference is gc, than i
find it.
was (Author: aoxiang):
I am doing HBASE-16213, and the HBASE-16213_branch1_v3.patch have the same
problem, when i test the E2E performance, i find valueLength=1k, the
performance slow than NONE encoding, and i find the different is gc, than i
find it.
> Reduce garbage in BufferedDataBlockEncoder
> ------------------------------------------
>
> Key: HBASE-16502
> URL: https://issues.apache.org/jira/browse/HBASE-16502
> Project: HBase
> Issue Type: Improvement
> Reporter: binlijin
> Assignee: binlijin
> Attachments: HBASE-16502-master.patch
>
>
> In BufferedDataBlockEncoder.SeekerState every read will new a tagsBuffer for
> compressTags. There is no need when no tags or compressTags=false, so we can
> reduce this new byte[] to reduce garbage.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
