I have a N-1 fan in model. I am using ROUTER-DEALER-REP. (I use ZMQ.Poll to pull the requests from the ROUTER socket and push to DEALER socket.) The ROUTER and DEALER sockets bind at the server and client simply connects to the ROUTER to push requests. In my case, around 200 clients connect to the ROUTER and send multiple requests. I am trying to make best use of the client side message queue feature provided by zeromq.
In case of a REQ-REP its very clear. The message stays in the message queue (on the client end) until the REP socket is ready to process. How does this concept translate to a more complicated model like REQ-ROUTER-DEALER-REP. In the doc, it also says that the queueing happens closer to the receiver. I am trying to avoid the situation where all the messages are queued up on the server. (For example, 200 clients * 100 messages queued up). Any help is much appreciated. Regards Rama
_______________________________________________ zeromq-dev mailing list [email protected] http://lists.zeromq.org/mailman/listinfo/zeromq-dev
