[
https://issues.apache.org/jira/browse/HBASE-16417?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15692572#comment-15692572
]
Anoop Sam John commented on HBASE-16417:
----------------------------------------
Make sense.. Agree to points raised by Eshcar and Ram.
40% default size for both memstore and block cache seems not a good choice. We
should change this specially after G1GC. This makes the working memory size to
be so high. The IHOP is very imp for G1GC to get a more predictable GC pause.
This defaults to 45% only.
May be if you tune the memstore and block cache size and IHOP accordingly, you
might see better results.
And best solution what we propose would be as Ram said. Use L2 off heap BC. The
latest results shared by Alibaba , after backporting HBASE-11425 work , reveal
that we will do better with new off heap L2 cache compared to L1 cache (on heap)
> In-Memory MemStore Policy for Flattening and Compactions
> --------------------------------------------------------
>
> Key: HBASE-16417
> URL: https://issues.apache.org/jira/browse/HBASE-16417
> Project: HBase
> Issue Type: Sub-task
> Reporter: Anastasia Braginsky
> Assignee: Eshcar Hillel
> Fix For: 2.0.0
>
> Attachments: HBASE-16417-benchmarkresults-20161101.pdf,
> HBASE-16417-benchmarkresults-20161110.pdf,
> HBASE-16417-benchmarkresults-20161123.pdf
>
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
