You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@qpid.apache.org by "Keith Wall (JIRA)" <ji...@apache.org> on 2016/10/03 14:03:21 UTC
[jira] [Commented] (QPID-6803) ByteBuffer slices/duplicates
contribute significantly to garbage overhead in busy Brokers
[ https://issues.apache.org/jira/browse/QPID-6803?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15542484#comment-15542484 ]
Keith Wall commented on QPID-6803:
----------------------------------
Examination shows that this change has reduced the memory requirements for the Broker, especially for the 0-8..0-91 use-case. We are know able to accommodate more than 500,000 queue entries (*) in heap (default memory settings), vs. 320,000 before this change.
> ByteBuffer slices/duplicates contribute significantly to garbage overhead in busy Brokers
> -----------------------------------------------------------------------------------------
>
> Key: QPID-6803
> URL: https://issues.apache.org/jira/browse/QPID-6803
> Project: Qpid
> Issue Type: Bug
> Components: Java Broker
> Reporter: Keith Wall
> Assignee: Keith Wall
> Fix For: qpid-java-6.1
>
>
> Profile whilst running auto-ack/transacted test at higher messages rates shows that the garbage collection is dominated being dominated by DIrectByteBuffers resulting from slicing/duplicating.
> We could refactor QpidByteBuffer (maintaining our own remaining/capacity/limit and keeping a reference to the original buffer) ourselves to avoid this overhead.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@qpid.apache.org
For additional commands, e-mail: dev-help@qpid.apache.org