[jboss-jira] [JBoss JIRA] (JGRP-1989) Bundlers: reuse send buffer when transport == sync.

[Bela Ban (JIRA)]

    [ https://issues.jboss.org/browse/JGRP-1989?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13140134#comment-13140134 ] 

Bela Ban commented on JGRP-1989:
--------------------------------

You don't give up, do you? :-)
Even the most light-weight impl adds ~16 bytes to every message [1], and I'm not willing to add that cost.

Looking at my previous comment, that may not even be necessary, because it seems we can reuse buffers *for sure* for 75% of our use cases:
* Sending of message batches: OK
* Sending of single messages: OK
* Reception of message batches: OK
* Reception of single messages: MAYBE

The last 25% (reception of single messages) is (1) not a big use case as most messages are received as batches anyway and (2) we could come up with requirements for how the {{receive()}} callback needs to use the buffer in the message argument. For instance, we could say that a buffer needs to be read and data in it needs to be de-serialized before {{receive()}} returns, or a copy has to be made. I also need to look at how retransmission is affected in {{NAKACK2}} and {{UNICAST3}}.

Actually, most callbacks do perform de-serialization right away, with some exceptions (e.g. Async Invocation API), so they should be conformant.

[1] https://github.com/belaban/workshop/blob/master/slides/status2015.adoc#size-of-new-message-with-buffer

> Bundlers: reuse send buffer when transport == sync.
> ---------------------------------------------------
>
>                 Key: JGRP-1989
>                 URL: https://issues.jboss.org/browse/JGRP-1989
>             Project: JGroups
>          Issue Type: Enhancement
>            Reporter: Bela Ban
>            Assignee: Bela Ban
>             Fix For: 3.6.7
>
>
> With the addition of {{TCP_NIO2}}, all bundlers now create new send buffers for every message (or message list). This generates a lot of memory allocations, perhaps it is better to revert this change for *synchronous transports* such as {{UDP}} and {{TCP}}, and still create new buffers for *asynchronous transports* such as {{TCP_NIO2}}.
> Synchronous transports guarantee a message has been put on the wire when {{TP.send()}} returns, whereas asynchronous transports may only have completed a partial write (so we cannot reuse the buffer).
> The code in the bundler should check for this, and copy if async or not copy if sync.
> Whether or not a transport is sync is determined by a new abstract method that needs to be overridden by every transport.



--
This message was sent by Atlassian JIRA
(v6.4.11#64026)