MQSeries.net :: View topic

mustang · Posted: Wed Feb 08, 2006 4:37 am Post subject:

A J2EE developer created a service using the WAS Service Integration
Bus features. I was asked to create a synchronous request/response
subflow client that uses the service by writing a request to its JMS
queue. This subflow would then be used in any message flow that wants
to be a client of that service.

I wrote a subflow using the aggregation fan-in fan-out flow design.
The fan-in and fan-out flow are in the same message flow. The flow
works fine in the lab. I wrote a message flow that calls the subflow,
and when a message is placed on the MQInput queue, it goes through and uses the subflow and JMS service just fine. When I walk it through in
debug mode, it behaves just as you would hope it would. Our concern is
although it works fine in a lab, how will it work in a production
environment? We are having a difficult time understanding how the
fan-in flow knows which original request to match the response to. The
documentation mentions the "ReplyIdentifier", but that gets modified as
it goes through various stages in the request/reply cycle (not the
least of which is using the JMS service). Yet, despite this, the
fan-in flow matches the response to the aggregation control message.
It works, and we don't know why, and that is what concerns us. What
happens if many messages are coming through? Will it be a bottleneck?
Will the fan-in really not care which request the response belongs to,
and just send it FIFO? We will try to test this out, but we really
would like to understand better how it works.

We are on, and this was written in, version 5. Eventually, we will be
on version 6.

Thank you very much.