hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Liang Xie (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-10010) eliminate the put latency spike on the new log file beginning
Date Mon, 25 Nov 2013 06:01:39 GMT

    [ https://issues.apache.org/jira/browse/HBASE-10010?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13831195#comment-13831195
] 

Liang Xie commented on HBASE-10010:
-----------------------------------

Oh, [~himanshu@cloudera.com], very very thanks for pointing out this.
After a looking at deeper code, i think the key is SequenceFile Class under hadoop-common
project, see (i extracted below code from cdh4.1.1 modified version)
{code}
    /** Write and flush the file header. */
    private void writeFileHeader() 
      throws IOException {
      out.write(VERSION);
      Text.writeString(out, keyClass.getName());
      Text.writeString(out, valClass.getName());
      
      out.writeBoolean(this.isCompressed());
      out.writeBoolean(this.isBlockCompressed());
      
      if (this.isCompressed()) {
        Text.writeString(out, (codec.getClass()).getName());
      }
      this.metadata.write(out);
      out.write(sync);                       // write the sync bytes
      out.flush();                           // flush header
    }
{code}

yes, i agree with you now that we can just issue "sync" api, it should be enough, since i
did a "svn blame" on SequenceFile.java, it shows this writeFileHeader() method was added long
long time ago.

I had uploaded the simplified patch against 0.94 branch and trunk, could you guys help to
look again? thanks! 
[~himanshu@cloudera.com], [~saint.ack@gmail.com],[~lhofhansl],[~tedyu@apache.org]

> eliminate the put latency spike on the new log file beginning
> -------------------------------------------------------------
>
>                 Key: HBASE-10010
>                 URL: https://issues.apache.org/jira/browse/HBASE-10010
>             Project: HBase
>          Issue Type: Improvement
>          Components: regionserver
>    Affects Versions: 0.94.13
>            Reporter: Liang Xie
>            Assignee: Liang Xie
>         Attachments: HBase-10010-0.94-v2.txt, HBase-10010-0.94-v3.txt, HBase-10010-0.94.txt,
HBase-10010-trunk-v2.txt, HBase-10010-trunk.txt
>
>
> In deed, the original finding came from fb, see HBASE-6813 for detailed discussion.
> Through this improvement doesn't expect obvious gain on 95th or 99th latency, it still
could make the response time more stable to me.



--
This message was sent by Atlassian JIRA
(v6.1#6144)

Mime
View raw message