[ 
https://issues.apache.org/jira/browse/HBASE-5898?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13267005#comment-13267005
 ] 

Jean-Daniel Cryans commented on HBASE-5898:
-------------------------------------------

A not too scientific comparison.

Before:

total=6.26 GB, free=1.04 GB, max=7.3 GB, blocks=101753, accesses=5078311, 
hits=4333154, hitRatio=85.32%, cachingAccesses=5078311, cachingHits=4333154, 
cachingHitsRatio=85.32%, evictions=54, evicted=643404, 
evictedPerRun=11914.888671875

After:

total=6.45 GB, free=878.07 MB, max=7.3 GB, blocks=104747, accesses=5810869, 
hits=4345850, hitRatio=74.78%, cachingAccesses=5810869, cachingHits=4345850, 
cachingHitsRatio=74.78%, evictions=52, evicted=627755, 
evictedPerRun=12072.2119140625

This is using the benchmark tool I'm working on right now, I had to port Todd's 
patch since I'm not going through the HFile code but I'm doing exactly what he 
does. The test is started completely cold. It uses 5 threads.

The big takeout is that it did 14% more cache accesses which had the side 
effect of lowering the hit ratio. The number of actual cache hits was a big 
higher.
                
> Consider double-checked locking for block cache lock
> ----------------------------------------------------
>
>                 Key: HBASE-5898
>                 URL: https://issues.apache.org/jira/browse/HBASE-5898
>             Project: HBase
>          Issue Type: Improvement
>          Components: performance
>    Affects Versions: 0.94.1
>            Reporter: Todd Lipcon
>            Assignee: Todd Lipcon
>            Priority: Critical
>         Attachments: 5898-TestBlocksRead.txt, hbase-5898.txt
>
>
> Running a workload with a high query rate against a dataset that fits in 
> cache, I saw a lot of CPU being used in IdLock.getLockEntry, being called by 
> HFileReaderV2.readBlock. Even though it was all cache hits, it was wasting a 
> lot of CPU doing lock management here. I wrote a quick patch to switch to a 
> double-checked locking and it improved throughput substantially for this 
> workload.

--
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

        

Reply via email to