MPS is above all a simple system, and does not go for performance as one of its primary goals. However, it does perform quite well, given its naivety. It may also be worth noting that I haven't even profiled the code yet, let alone optimised it. Also, I didn't even use the -O flags to GCC yet...
I've only done some very rough measurements so far; but the numbers I have may be of some use:
- On a Celeron 400 MHz with 64 MB RAM (not that the amount of RAM actually makes a difference) running Redhat Linux 6.2, 100,000 method calls run in ~10.8 seconds giving ~9000 method calls per second (this is two processes on the same machine communicating via sockets).
- Performance is roughly 3000 method calls per second if the server and client are run on different machines over an unloaded 10Mbps Ethernet.
- Performance over a 56k modem is lousy :-) Latency is more important than bandwidth in the tests I was performing.
(last modified 04 June 2000 by tonyg)