[
https://issues.apache.org/jira/browse/HBASE-18442?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16182314#comment-16182314
]
ramkrishna.s.vasudevan commented on HBASE-18442:
------------------------------------------------
bq.MSLAB chunk pool, the initialCount by def is 0 right?
I see. I did not check this value. Will see. If so the patch does not handle
this case.
bq.I mean am not fully sure whether we need this multi threaded fix fo rthe
MSLAB chunk pool
Simply having 'max count 12288, initial count 12288' itself takes more time. If
we go with MSLAB and pool I think in cases where the init and max is almost
same it will make a difference. Ya in cases where init is very small this won't
help.
> Speed up Memstore chunk pool ByteBuffer allocations
> ---------------------------------------------------
>
> Key: HBASE-18442
> URL: https://issues.apache.org/jira/browse/HBASE-18442
> Project: HBase
> Issue Type: Improvement
> Reporter: ramkrishna.s.vasudevan
> Assignee: ramkrishna.s.vasudevan
> Fix For: 2.0.0, 2.0.0-alpha-4
>
> Attachments: HBASE-18442_1.patch, HBASE-18442_2.patch,
> HBASE-18442_3.patch, HBASE-18442_4.patch
>
>
> Like in HBASE-17738 we can speed up the allocation of memstore chunk pool's
> ByteBuffers. {code}
> 2017-07-24 17:51:09,726 INFO [regionserver/stobdtserver6/10.66.254.41:16020]
> regionserver.ChunkCreator: Allocating MemStoreChunkPool with chunk size 2 MB,
> max count 12288, initial count 12288
> 2017-07-24 17:51:19,642 INFO [regionserver/stobdtserver6/10.66.254.41:16020]
> regionserver.HRegionServer: Serving as stobdtserver6,16020,1500898858958,
> RpcServer on stobdtserver6/10.66.254.41:16020, sessionid=0x15d748a9ccc0002
> {code}
> Allocating 12288 buffers of 2MB size takes around 10 secs.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
