Yeah.. Yaws v. lighttpd, nginx, thttpd or tornado would be interesting. When's the last time apache won a benchmark of anything?

Apache 2.0.39? When was this test made?

Also, which apache model, prefork or worker was used?

Apache hasn't changed that such since 2.0 when it comes to basic connection handling. When it comes to handling tons and tons of connection, its I/O architecture is the problem, irregardless of prefork or worker.

That said, the comparison is pretty yawning. It's been known for years that Apache with its non-evented architecture is not good at handling massive amounts of connections; there's really no point at comparing with Apache anymore. I'm more interested in how this stacks up against Lighttpd or Nginx.

This is really old.

They're essentially comparing apples to oranges:

Apache's worker MPM uses system threads, whereas Erlang uses its own internal scheduler, so that at the system level, there is only one process.

I would like to see memory and CPU-time allocation during the tests, as well as some more information on the test workload.

Have a little dig about on the web. There are benchmarks galore.

I found nothing that would give me this kind of info. Most limit themselves to how much memory the box had and load numbers.

No ref to apache hece, but might be of interest:

http://www.planeterlang.org/en/planet/article/Nginx_vs_Yaws_...

anyone have yaws versus nginx data?

nginx would do very well too. But what's interesting about yaws or mochiweb or other Erlang systems is that they could also handle long-running calculations just about as well as they handle a static file. Well, perhaps there would be some loss of performance, but not nearly so much as if you have to spawn external threads or processes to handle said long running calculations.

This is a very old test, I recall seeing this back in 2006 or so. It would interesting to see a comparison with Mochiweb also.

Erlang really rocks!