[
https://issues.apache.org/jira/browse/HBASE-6619?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13466567#comment-13466567
]
Michal Gregorczyk commented on HBASE-6619:
------------------------------------------
I am not sure if I got the question right.
Previously thread pool was used to read from connections. New task was
submitted every time Listener thread noticed that there is some data to read.
With this patch, round robin is used to distribute new (just accepted)
connections among Readers (threads introduced in this patch). Each of them uses
its own selector and reads all requests from connections that were registered
with its selector (without delegating reads to pool of threads). This change
removes need for registering and deregistering interest in read events
(previously Listener had to unset read interest before submitting read task,
and task had to set it once again when finished reading).
So purposes of ExecutorService and round robin are different.
Does it answer your question ?
> Do no unregister and re-register interest ops in RPC
> ----------------------------------------------------
>
> Key: HBASE-6619
> URL: https://issues.apache.org/jira/browse/HBASE-6619
> Project: HBase
> Issue Type: Bug
> Components: IPC/RPC, Performance
> Reporter: Karthik Ranganathan
> Assignee: Michal Gregorczyk
> Priority: Critical
> Fix For: 0.94.3, 0.96.0
>
> Attachments:
> 0001-jira-HBASE-6619-89-fb-Do-no-unregister-and-re-regist.patch
>
>
> While investigating perf of HBase, Michal noticed that we could cut about
> 5-40% (depending on number of threads) from the total get time in the RPC on
> the server side if we eliminated re-registering for interest ops.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
