[jboss-jira] [JBoss JIRA] Commented: (JGRP-181) Threadless stack

[Bela Ban (JIRA)]

    [ http://jira.jboss.com/jira/browse/JGRP-181?page=comments#action_12341748 ] 
            
Bela Ban commented on JGRP-181:
-------------------------------

One idea is to have a thread pool handle incoming messages, so that for messages A1, B10, B11, B12, A2, {A1,A2} and {B10-12} can be processed concurrently with respect to each other.
However, in the following case: A1, A2, A3, A4, B1. Let's assume we have a threadpool of 2 configured. In this example, thread-1 (T1) will handle A1 and T2 will handle A2. T2 will have to block until A1 has been processed. This is bad, as we could deliver B1 on a separate thread in the meantime.
So, using a threadpool for each message, regardless of the message's sender, isn't good. We should rather categorize messages by sender, e.g. have 1 queue per sender: {A1-4} and {B1}, and have T1 process the queue {A1-4} and T2 process the queue {B1}.
Need to investigate whther this is possible with QueuedExecutor

> Threadless stack
> ----------------
>
>                 Key: JGRP-181
>                 URL: http://jira.jboss.com/jira/browse/JGRP-181
>             Project: JGroups
>          Issue Type: Task
>    Affects Versions: 2.2.9.1, 2.2.9, 2.2.8
>            Reporter: Bela Ban
>         Assigned To: Bela Ban
>            Priority: Critical
>             Fix For: 2.5
>
>
> Investigate setting all down_thread and up_thread properties to false by default
> Performance tests have shown that in almost all cases this is always faster, so why not make it the default. This is e.g. already the case in fc-fast-minimalthreads.xml

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: http://jira.jboss.com/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira