You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "Zhihong Yu (Commented) (JIRA)" <ji...@apache.org> on 2012/01/31 01:11:10 UTC
[jira] [Commented] (HBASE-5303) Separate out rows into multiple
files during compaction for better block cache utilization
[ https://issues.apache.org/jira/browse/HBASE-5303?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13196564#comment-13196564 ]
Zhihong Yu commented on HBASE-5303:
-----------------------------------
Interesting idea.
We need to define (pluggable) policies that help us determine the hotness of rows.
> Separate out rows into multiple files during compaction for better block cache utilization
> ------------------------------------------------------------------------------------------
>
> Key: HBASE-5303
> URL: https://issues.apache.org/jira/browse/HBASE-5303
> Project: HBase
> Issue Type: Improvement
> Components: regionserver
> Reporter: dhruba borthakur
> Assignee: dhruba borthakur
>
> This seems odd at first sight, a compaction is supposed to reduce the number of files. But there is great advantage to creating hbase blocks where adjacent keys are accessed together. One approach would be to make compaction write out hot-rows into one file and the cold-rows into another file. This increases the efficiency of the block cache because only the blocks in the hot-rows file will most frequently be populated in the block cache.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira