[jira] [Commented] (OAK-4631) Simplify the format of segments and serialized records

["Michael Dürig (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/OAK-4631?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15406106#comment-15406106 ] 

Michael Dürig commented on OAK-4631:
------------------------------------

As mentioned in the [discussion|http://markmail.org/thread/3oxp6ydboyefr4bg] on the list I would like to get a better understanding on mainly the performance impact of such a change. Having lifted the 255 segment references limit (aka OAK-2896) is a good thing and I would expect it to have an positive impact on performance. However, I suspect most of the positive impact only to become visible on big stores with many small records and a lot of concurrent traffic. IMO we should

* run the micro benchmarks from {{oak-run}} and compare numbers with and without the patch. I wouldn't expect big changes but still we might be missing something. 

* with the help of [~volteanu], run longevity tests with concurrent traffic comparing to the same test without the patch. Apart from overall system performance I would also be interested in the difference of repository sizes with and without the patch, the hit/miss statistic for the segment cache with and without the patch and the segment size histogram with and without the patch. 

> Simplify the format of segments and serialized records
> ------------------------------------------------------
>
>                 Key: OAK-4631
>                 URL: https://issues.apache.org/jira/browse/OAK-4631
>             Project: Jackrabbit Oak
>          Issue Type: Improvement
>          Components: segment-tar
>            Reporter: Francesco Mari
>            Assignee: Francesco Mari
>             Fix For: Segment Tar 0.0.10
>
>         Attachments: OAK-4631-01.patch
>
>
> As discussed in [this thread|http://markmail.org/thread/3oxp6ydboyefr4bg], it might be beneficial to simplify both the format of the segments and the way record IDs are serialised. A new strategy needs to be investigated to reach the right compromise between performance, disk space utilization and simplicity.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)