[
https://issues.apache.org/jira/browse/HBASE-11331?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14109769#comment-14109769
]
Nick Dimiduk commented on HBASE-11331:
--------------------------------------
This data is from increasing the query range to size=100; more data than fits
in cache with either config. I had to run the warmup and test for longer
periods because it took longer for the oscache and BlockCache to reach
steady-state.
|| ||=false, 100g||=true, 100g||delta|
|hbase.regionserver.server.Get_num_ops|313.83|466.69|{color:green}49%{color}|
|hbase.regionserver.server.Get_mean|28.91 ms|20.00 ms|{color:green}-31%{color}|
|hbase.regionserver.server.Get_99th_percentile|221.06 ms|197.30
ms|{color:green}-11%{color}|
|hbase.regionserver.jvmmetrics.GcTimeMillis|26.99 ms|48.84
ms|{color:red}81%{color}|
|proc.loadavg.1min|11.71|12.00|{color:red}2%{color}|
|proc.stat.cpu.percpu{type=iowait}|343.11|404.10|{color:red}18%{color}|
|hbase.regionserver.server.blockCacheCount|181.85 k|716.79
k|{color:green}294%{color}|
Overall, I'd say that you want this feature enabled unless:
# your data decompressed fits in BlockCache
# your machines are not your own and CPU time is at a premium (ie, AWS).
2nd point above is just a guess. It's likely other factors are at play in these
environments.
> [blockcache] lazy block decompression
> -------------------------------------
>
> Key: HBASE-11331
> URL: https://issues.apache.org/jira/browse/HBASE-11331
> Project: HBase
> Issue Type: Improvement
> Components: regionserver
> Reporter: Nick Dimiduk
> Assignee: Nick Dimiduk
> Attachments: HBASE-11331.00.patch, HBASE-11331.01.patch,
> HBASE-11331.02.patch, HBASE-11331.03.patch, HBASE-11331.04.patch,
> HBASE-11331.05.patch, HBASE-11331LazyBlockDecompressperfcompare.pdf,
> hbase-hbase-master-hor17n36.gq1.ygridcore.net.log, lazy-decompress.02.0.pdf,
> lazy-decompress.02.1.json, lazy-decompress.02.1.pdf, v03-20g-045g-false.pdf,
> v03-20g-045g-true-16h.pdf, v03-20g-045g-true.pdf
>
>
> Maintaining data in its compressed form in the block cache will greatly
> increase our effective blockcache size and should show a meaning improvement
> in cache hit rates in well designed applications. The idea here is to lazily
> decompress/decrypt blocks when they're consumed, rather than as soon as
> they're pulled off of disk.
> This is related to but less invasive than HBASE-8894.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
