Reputation: 43499
This is probably a very naïve question, but I'm really a newbie in that stuff.
I'd like to test 0MQ performances (latency, throughput) according to different communication patterns: REQ/REP, PUB/SUB, PUSH/PULL, ROUTER/DEALER and so on, ... and estimate how well, performance-wise, 0MQ would handle the various communication scenarios we encounter in our software.
When everything runs on the same machine, it is relatively easy to measure things and do basic statistics according to message size, etc. I know for sure when my messages are sent, and when they are received.
But how can I do measurements across the network without a common time reference (which is accurate enough, I mean)? Do I measure round-trips (from machine A to machine B and back)? Is that a meaningful test?
Upvotes: 3
Views: 772
Reputation: 6669
ZeroMQ comes with performance testing tools; look in the perf/ directory. E.g. to test throughput, run local_thr on one machine, and remote_thr on the other. You can set message sizes and counts. Do test with sufficient messages to get accurate figure (test should run for at least 5-10 seconds).
Upvotes: 3