[jira] [Commented] (HBASE-8836) Separate reader and writer thread pool in RegionServer, so that write throughput will not be impacted when the read load is very high

["Tianying Chang (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HBASE-8836?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13699253#comment-13699253 ] 

Tianying Chang commented on HBASE-8836:
---------------------------------------


  >>  Is there a better way to detect write actions?
Lars, do you have some suggestion? I also feel this is a little too clumsy. This was the most straightforward way I can think of when I start coding for my POC. Anyone can suggest some other way? I would love to try that.  

  >>  Do you really want to tease apart a "multi" action and enqueue the parts individually?
Good point. When I run the test, I found that even single "Put"(and some other method) was actually wrapped into the "multi", instead of the Put method directly, so I have to parse through the argument of the "multi" method so that I can make it functionally correct. But this is not a very clean way. I really appreciate it if someone can give me some hint of a better way to detect this. 

  >  We can do this with less code duplication I think.
Sure. I was working on the non-secure version first, and test on my dev VM cluster which is not secure version. Later on, when I move to test on our testing cluster (50+ nodes) which is a secure enabled, I found my change is not being executed, and realized that it calls the secure path of RPC. To quickly unblock myself and get a perf result on the testing cluster, I just copy/paste my change into the SecureServer.java. I will remove the duplicate later.  
                
> Separate reader and writer thread pool in RegionServer, so that write throughput will not be impacted when the read load is very high
> -------------------------------------------------------------------------------------------------------------------------------------
>
>                 Key: HBASE-8836
>                 URL: https://issues.apache.org/jira/browse/HBASE-8836
>             Project: HBase
>          Issue Type: New Feature
>          Components: Performance, regionserver
>    Affects Versions: 0.94.8
>            Reporter: Tianying Chang
>            Assignee: Tianying Chang
>             Fix For: 0.98.0, 0.95.2, 0.94.10
>
>         Attachments: hbase-8836.patch
>
>
> We found that when the read load on a specific RS is high, the write throughput also get impacted dramatically, and even cause write data loss sometimes. We want to prioritize the write by putting them in a separate queue from the read request, so that slower read will not make fast write wait nu-necessarily long.

--
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