The peak load is going to be important (and what an acceptable time from putting in the queue to being processed is), as 1 million messages a day is on average slightly over 10 requests per second, which is definitely possible (and scales to higher numbers if you spread out processing on a number of machines), but it will depend on the architecture you set up, how many different processes you have, how many persistent points, how your JMS queues will perform, etc.
I would recommend creating a simple prototype that represents your typical use case and then running performance tests on that.
Thanks Kris for reply. My apologies to reply you back on it. For this one currently there are some processes are triggered through event from another processes where we are facing problem is passing data from parent process to process triggered through an event.
We are going with CDI implementation and currently will trying this load test for around 3 simple process for each message triggered using events. There are around 6 persistent points in all together.
We are in process of doing it and will surely let you know results.