Thanks for your feedback Andy. I take your point and I think that the benefits of having larger queue buffers is something that might be evaluated in more depth at some point to give some guidance on where these provide benefit. The story is more nuanced, as you say.

Paul,

You state "Note that large queue buffers may not be needed in your configuration. Writes to the queue files are asynchronous, taking advantage of OS buffering. Large buffers were set in the runs here, as a precaution only".

This seems like a very broad statement, I might have been expecting a bit more qualification. Although the writes to the queue files are unforced, the queue manager will periodically force any buffers that contain persistent messages that are not near immediately consumed.  The OS has no idea which buffers relate to which messages and so by and large cannot optimize this process nearly as well as the queue managers own buffer management is able to do so. Buffer management was something I used to chip away at before I retired but lacked significant investment. I'm unaware if greater investment has been made in this area in recent years but I suspect not ?

One of the big conceptual differences between NativeHA and RDQM used to be that NativeHA only ships log writes whereas RDQM ships both log writes and queue files writes (I'm guessing that this is still true). Again I might expect this to have a knock on effect where giving MQ's own buffer  manager more resources (bigger buffers) would have a more significant effect in an RDQM environment.

Best wishes

Andy.