[jbosscache-issues] [JBoss JIRA] Updated: (JBCACHE-1564) Using asynchronous API of JGroups for asynchronous replication

[Krzysztof Sobolewski (JIRA)]

     [ https://jira.jboss.org/jira/browse/JBCACHE-1564?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Krzysztof Sobolewski updated JBCACHE-1564:
------------------------------------------

    Attachment: GroupRequest.diff


The changes to the "hacked" GroupRequest class.

> Using asynchronous API of JGroups for asynchronous replication
> --------------------------------------------------------------
>
>                 Key: JBCACHE-1564
>                 URL: https://jira.jboss.org/jira/browse/JBCACHE-1564
>             Project: JBoss Cache
>          Issue Type: Patch
>      Security Level: Public(Everyone can see) 
>          Components: Replication
>    Affects Versions: 3.2.1.GA
>            Reporter: Krzysztof Sobolewski
>            Assignee: Manik Surtani
>         Attachments: async-replication.patch, GroupRequest.diff
>
>
> As I mentioned on Infinispan list, I run into some problems with NBST and asynchronous replication. Quoting my message:
> ***
> The thing is that the "asynchronous" replication uses the usual method of 
> submitting a task to an Executor. This is no problem without NBST, because the 
> tasks are not interested in responses and finish immediately, but with NBSP 
> they need responses to facilitate replay. And since they need to wait for 
> responses, they clog the Executor's threads and under sufficiently heavy load, 
> it all degrades to effectively synchronous replication. And it happens quite 
> fast, really.
> ***
> The fix in the attached patch creates a new class, AsynchronousRpcDispatcher, and makes CommandAwareRpcDispatcher extend it. The new class adds a new variant of castMessage() method - castMessageAsync(). I takes a ScheduledExecutorService where it schedules a timeout task, and a "closure" callback object, an implementation of AsyncCallback interface, which is invoked when the replication task is complete. "Completion" also means a timeout, in which case the callback is invoked with whatever responses were received.
> One thing I had to do was to "hack" the GroupRequest class from JGroups to expose responsesComplete() method. I just copied it and did the necessary modification, diff of which is also attached. The method I needed is used to check whether the received responses are enough to complete the request.
> There are also some minor stylistic tweaks in the patch unrelated to the real thing.

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