Hi Omer, > I'm testing a scenario in which I have one publisher sending messages > to multiple subscribers. > In the message the publisher sends I'm sending the value of > QueryPerformanceCounter and when the subscriber gets a message it > calculates the difference between the current QueryPerformanceCounter > to the one stored in the message. > > It seems that for 1000 messages sent from the publisher to the > subscribers, the last one takes around 20 times than the average time > it takes for the previous ones to arrive. > Seems as the last one is "stuck" and in some queue and is released > probably after a timeout.
There's no flush timeout in 0MQ code, so the delay is either a bug in 0MQ or a side effect of how latency is measured. Is the problem reproducible? If so, can you provide the test program? Martin _______________________________________________ zeromq-dev mailing list [email protected] http://lists.zeromq.org/mailman/listinfo/zeromq-dev
