[jboss-jira] [JBoss JIRA] (JGRP-155) TP.loopback should always be used

[Bela Ban (JIRA)]

    [ https://issues.jboss.org/browse/JGRP-155?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12751965#comment-12751965 ] 

Bela Ban commented on JGRP-155:
-------------------------------

The reason for the perf degradation is probably that we loop back messages *individually*, e.g. if A multicasts 1000 messages, TP.loopback=true will use 1 thread / message.
SOLUTION: we should move the loopback handling after the bundling code, so that looped-backed messages are bundled (batched) as well. This way, we'll use fewer threads to handled looped backed messages.
                
> TP.loopback should always be used
> ---------------------------------
>
>                 Key: JGRP-155
>                 URL: https://issues.jboss.org/browse/JGRP-155
>             Project: JGroups
>          Issue Type: Task
>    Affects Versions: 2.2.8, 2.2.9
>            Reporter: Bela Ban
>            Assignee: Bela Ban
>             Fix For: 3.x
>
>
> Investigate the 3 potential problem areas below:
> - Memory leak caused by FC when credits are received too quickly
> - Deadlocks caused by same threads calling themselves (e.g. VIEW_OK ack msgs)
> - Preemption doesn't work: Scheduler is meaningless here

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira