[
https://issues.apache.org/jira/browse/HBASE-8420?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13649353#comment-13649353
]
Jimmy Xiang commented on HBASE-8420:
------------------------------------
You are right. It is not exactly the same as before. The difference is mostly
in the ClientScanner and related, as I mentioned in the RB. The main reason for
that is because how the ClientScanner moves to the next scanner: asking for a
number of rows, moving to the next scanner/region if getting less. The existing
logic is to cache exactly the "caching" number of rows. This patch could
caching 1-less or many more rows, instead. Not sure if it is ok. But I think it
makes the logic more efficient.
bq. Lastly in 0.94, we could use scan attributes to indicate this per scanner
in a backward compatible way.
Cool, we can use scan attributes to indicate prefetching instead of a global
flag. I will update the 0.94 patch.
As to the caching size, for 0.94, should we leave it as before (1), to make it
the same as the trunk (100)?
> Port HBASE-6874 Implement prefetching for scanners from 0.89-fb
> -----------------------------------------------------------------
>
> Key: HBASE-8420
> URL: https://issues.apache.org/jira/browse/HBASE-8420
> Project: HBase
> Issue Type: Improvement
> Reporter: Jimmy Xiang
> Assignee: Jimmy Xiang
> Attachments: 0.94-8420_v1.patch, trunk-8420_v1.patch
>
>
> This should help scanner performance. We should have it in trunk.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
