I am attempting to create a computing grid. I setup the grid servers to pull requests from the mutliple brokers using the instructions from "HowDoIConfigureAnMDBToTalkToARemoteQueue". Once the MDB running on the grid is finished processing the message, a message needs to be sent back to the broker.
The only design option I see at this point is to configure queues on each of the compute grid servers and have the MDB running on the broker listen for messages on each of the grid servers. However, we could potentially have hundreds of compute grid servers and I believe this would be a lot of overhead if the MDB running on the broker needed to maintain connections to every grid server. In addition, I would need to add jms proxy configuration in the broker for every grid server.....it seems there should be a better solution.
Is it possible to create a design where the grid servers can talk back to a queue residing on the broker? In addition, I would like the grid servers to be able to detect a failed broker to ensure that the message is sent to a running broker.
Any advice would be very much welcomed.