[jira] [Issue Comment Edited] (HBASE-4812) GC benchmarking and analysis tools

["Jonathan Gray (Issue Comment Edited) (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HBASE-4812?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13152300#comment-13152300 ] 

Jonathan Gray edited comment on HBASE-4812 at 11/17/11 8:37 PM:
----------------------------------------------------------------

{{analyzeGC.py}} is a python script that parses raw GC logs (lots of regex) and has some mechanisms for aggregating and summarizing.  It's not done but most of the nastiest bits are completed or near complete.

{{BlockCacheBench.java}} is the isolated block cache benchmarker.  In addition to threading and all that, it uses RandomGenerator to generate various distributions of operations (needs improvements but baseline is there).
                
      was (Author: streamy):
    <b>analyzeGC.py</b> is a python script that parses raw GC logs (lots of regex) and has some mechanisms for aggregating and summarizing.  It's not done but most of the nastiest bits are completed or near complete.

<b>BlockCacheBench.java</b> is the isolated block cache benchmarker.  In addition to threading and all that, it uses RandomGenerator to generate various distributions of operations (needs improvements but baseline is there).
                  
> GC benchmarking and analysis tools
> ----------------------------------
>
>                 Key: HBASE-4812
>                 URL: https://issues.apache.org/jira/browse/HBASE-4812
>             Project: HBase
>          Issue Type: Task
>          Components: io
>            Reporter: Jonathan Gray
>            Assignee: Jonathan Gray
>            Priority: Minor
>         Attachments: GCBench.patch
>
>
> One of the challenges of debugging GC issues is that it can take a long time to reproduce GC pauses.  Isolating the LRU block cache and benchmarking it independently can more easily cause measurable and reproducible GC issues.
> In addition, making sense of the noisy GC logs is difficult.
> This may not be for commit, but I wanted to share some code and scripts I've written to tackle these problems.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira