[
https://issues.apache.org/jira/browse/HBASE-16417?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15692359#comment-15692359
]
Eshcar Hillel commented on HBASE-16417:
---------------------------------------
My explanation is that with 100% writes the block cache is empty and does not
take any memory from the heap.
Whereas when there is even just 5% reads (will be the same for even less) the
block cache is full taking 40% (!!!) of the heap space (or 38% to be precise in
our settings). Only 62% of the heap space are left to be used by memstore,
chunk pool, compactions (both from disk and memory) and also the GC which is
know to take a lot of space (g1gc).
So at some point there is just not enough space for all these to work well, or
even to work at all.
> 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)
