[
https://issues.apache.org/jira/browse/OAK-3715?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15035878#comment-15035878
]
Thomas Mueller commented on OAK-3715:
-------------------------------------
It looks good to me.
Where the difference will probably show up is on Java GC logs, and in heap
allocations statistics, when using
{noformat}
-Xrunhprof:heap=sites,depth=4
{noformat}
> SegmentWriter reduce buffer size for reading binaries
> -----------------------------------------------------
>
> Key: OAK-3715
> URL: https://issues.apache.org/jira/browse/OAK-3715
> Project: Jackrabbit Oak
> Issue Type: Improvement
> Components: segmentmk
> Reporter: Alex Parvulescu
> Assignee: Alex Parvulescu
> Priority: Minor
> Fix For: 1.3.12
>
> Attachments: OAK-3715.patch
>
>
> The SegmentWriter uses an initial buffer size of 256k for reading input
> streams binaries that need to be persisted, then it checks if the input is
> smaller than 16k to verify if it can be inlined or not. [0]
> In the case the input binary is small and can be inlined (<16k), the initial
> buffer size is too wasteful and could be reduced to 16k, and if needed
> increased to 256k after the threshold check is passed.
> [0]
> https://github.com/apache/jackrabbit-oak/blob/trunk/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/segment/SegmentWriter.java#L495
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
