[jira] [Commented] (HBASE-2247) [performance] Small scans should run like a random-read

[LiuLei (JIRA)]

    [ 
https://issues.apache.org/jira/browse/HBASE-2247?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13461331#comment-13461331
 ] 

LiuLei commented on HBASE-2247:
-------------------------------

Hi stack， I  think We can new smallScan() method in HTable, the method use 
pread to read data and  the method return all data in one hfile block.  I think 
we can let users to decide use smallScan() method or getScanner method.


                
> [performance] Small scans should run like a random-read
> -------------------------------------------------------
>
>                 Key: HBASE-2247
>                 URL: https://issues.apache.org/jira/browse/HBASE-2247
>             Project: HBase
>          Issue Type: Improvement
>          Components: Performance
>            Reporter: stack
>
> hbase-2180 added pread whenever we do a get -- random-read -- and kept the 
> old sync+position+read for when scanning.   In between is the case of small 
> scans.  Small scans of 0-100 or so rows where the cells are small will likely 
> fit a single hfile blocksize, especially if its the default 64k.  We should 
> recognize small scans and flip to random-read to satisfy (somehow).  It'll up 
> the performance a bit.

--
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