<br>When I've done similar tests, I've usually benchmarked one <br>erlang node talking to another over such a TCP socket.<br>Depending on machine, I've usually landed around 100-150us<br>round-trip. With blocking select on the external program,
<br>you should be able to get better results.<br><br>I think that 50 us round-trip overhead when calling an erlang<br>function from a C program would feel comfortable. I've not <br>been able to reach that through normal means, but I believe
<br>that it was done once using a blocking linked-in driver (by<br>Lars Björnfot at CSLab, while evaluating VIA)<br><br>BR,<br>Ulf W<br><br><div><span class="gmail_quote">2007/4/14, James Hague <<a href="mailto:james.hague@gmail.com">
james.hague@gmail.com</a>>:</span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">I wrote an Erlang program that uses a local TCP socket (with nodelay)
<br>to communicate with a tiny external program (which I wrote). I was<br>curious what the turnaround time was on sending data to the external<br>program and getting back a result, so I wrote a little test function.<br><br>
Under Windows XP, I'm getting ~3500 round trips per second. I was<br>surprised at how hight that number was.<br><br>Under OS X, I'm getting ~90 round trips per second. 90? Wow, that's<br>low! I'm sure something is wrong, but I haven't found it yet.
<br><br>Anyone else ever run tests like this? I'd be curious what kind of<br>numbers you got.<br><br>Thanks!<br>_______________________________________________<br>erlang-questions mailing list<br><a href="mailto:erlang-questions@erlang.org">
erlang-questions@erlang.org</a><br><a href="http://www.erlang.org/mailman/listinfo/erlang-questions">http://www.erlang.org/mailman/listinfo/erlang-questions</a><br></blockquote></div><br>