MQSeries.net :: View topic - Workload Exits

Deeko

Hi,

We're considering using a workload exit to route traffic to destinations where state is stored.

The algorithim is based on maintaining a "cache" of destinations against state identifiers. First time requests (with no state at a destination) would take the default destiuantion and create an entry in this cache (with a unique id & destination). Subsequent requests would look for an entry and route accordingly. There can be more than other executable doing PUTs so the cache has to visible to each process.

This is obviously a quite different type of usage for a workload exit and I have some questions I can't find answers to.

My questions are:

Is this a suitable use for an exit? Are we being a bit over ambitious?

What's the best implementation choice for the cache (DB, shared memory, heap (if possible) etc)? We were thinking of a combination of DB and heap for performance reasons).

Where is the exit physically executed? Is it in the agent process (amqzlaa0) on the queue manager machine? We need to know this in order to deal with cache placement.

Are the destinations passed into the exit guaranteed not to change? If we have a number of valid destinations normally passed into the exit (say 1-4) what happens if 3 becomes unavailable? Does the exit get passed in 1-3 or 1, 2 & 4? This obviously would cause bother if it does and we'd have to navigate the MQWDR structures to validate the choices (plus store the destination queue managers in the cache).

Thanks,

Derek.