tux3-2.6.8.1-A5 performance

Sapan Bhatia sapan.bhatia at inria.fr
Wed Oct 6 19:18:57 UTC 2004



> > On Wednesday 06 October 2004 18:31, Yaoping Ruan wrote:
> > > 1K single file, none persistent connection: TUX - 221 Mb/s, Flash - 299
> > > Mb/s 10K single file, none persistent connection: TUX - 1232 Mb/s, Flash -
> > > 1461 Mb/s
> > > 1K single file, persistent connection: TUX - 391 Mb/s, Flash - 892 Mb/s
> > > 10K single file, persistent connection: TUX - 1627 Mb/s, Flash - 2033 Mb/s
> > With what kind of program are you testing this?
> 
> A tool similar to httperf. I don't think measurement tools matter. BTW, can
> ApacheBench run on multiple clients?

No it can't. Unlike httperf, it calculates requests/second and the total
bandwidth based on the total number of requests served/data transferred
through the full benchmark period, so it'd be quite difficult to
synchronize across multiple clients.

> > I'm using apachebench at the moment to test an opteron 248 (and 242) against
> > an nocona 3GHz and 3.6GHz. If I use a 1kb file i dont get anywhere near 400
> > (or 900) mbit/s. (I hit that with 10k files, using 100k files i even get more
> > than 13 gbit/s)
> >
> > What concurrency levels are you using, and is the file compressable (ie; just
> > a bunch of zero's or some file which gets bigger if you gzip it)
> 
> I tried 1200 - 2400 concurrent connections but didn't notice any difference of
> them. The files do gets bigger if gzipped, i.e. random chars. But "compression"
> is disabled and tests with a file of zeros have the same results.
> 
> Are there something I am missing?
> 
> -Yaoping
> 
> 
> _______________________________________________
> tux-list mailing list
> tux-list at redhat.com
> https://www.redhat.com/mailman/listinfo/tux-list




More information about the tux-list mailing list