You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "Biju Nair (JIRA)" <ji...@apache.org> on 2019/05/15 01:13:00 UTC
[jira] [Commented] (HBASE-22418) Separate queue/executor for Multi
get requests
[ https://issues.apache.org/jira/browse/HBASE-22418?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16839897#comment-16839897 ]
Biju Nair commented on HBASE-22418:
-----------------------------------
Attached changes for feedback which based on the new property {{hbase.ipc.server.callqueue.multiread.ratio}} value will use a fraction of the read queue/handler to service {{multi gets}}.
> Separate queue/executor for Multi get requests
> ----------------------------------------------
>
> Key: HBASE-22418
> URL: https://issues.apache.org/jira/browse/HBASE-22418
> Project: HBase
> Issue Type: Improvement
> Components: regionserver
> Reporter: Biju Nair
> Priority: Major
> Attachments: HBASE-22418.DRAFT
>
>
> Using {{RWQueueRpcExecutor}}, {{SimpleRpcScheduler}} lets us define separate queue/handlers to process write, read and scan requests. If this can be extended to have separate queue/handlers for multi get requests, it will help processing single get requests quickly where there is a mix of multi and single get workload. This is the case for multi put/single put as well. Any drawbacks in doing this segregation?
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)