[ https://issues.jboss.org/browse/JGRP-1717?page=com.atlassian.jira.plugin.... ] Bela Ban commented on JGRP-1717: -------------------------------- I think the underlying problem is that bundled *OOB messages* will get delivered as message *batches* at the receiver(s). This means that if we have a batch of 5 OOB RPCs [1,2,3,4,5], then RPC #1 will get invoked, then RPC #2 etc, *even if these RPCs are all tagged as OOB*. RPC #5 for example will be delayed by the cummulative delays of all RPCs ahead of it. So while message batching increases throughput, it also increases latency, which decreases performance of synchronous RPCs. Note that the RPC responses are also going to be delivered as a batch. Therefore, currently message batching and sync OOB RPCs only makes sense if the receivers use the Async Invocation API and and internal thread pool. If this is not the case, sync OOB RPCs should be marked with a {{DONT_BUNDLE}} flag. -- 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