[
https://issues.apache.org/jira/browse/HBASE-1505?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12762378#action_12762378
]
stack commented on HBASE-1505:
------------------------------
Turns out, we do seek and then read rather than pread. I had it wrong way
around. We should look into this.
> [performance] hfile should change how it reads from hdfs -- pread/seek+read
> -- dependent on recent history
> ----------------------------------------------------------------------------------------------------------
>
> Key: HBASE-1505
> URL: https://issues.apache.org/jira/browse/HBASE-1505
> Project: Hadoop HBase
> Issue Type: Improvement
> Reporter: stack
> Fix For: 0.21.0
>
>
> hfile preads all the time at moment. its good we do this when we want to up
> our random reads. but apparently pread involves open socket, seek, read,
> close socket whereas seek+read could leave us aligned at next block to read
> and make for some savings. hfile could have a short memory and change how it
> pulls from hdfs dependent on recent history.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
