httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dean Gaudet <dgau...@arctic.org>
Subject Re: Benchmarking - For future reference
Date Mon, 18 Aug 1997 22:43:02 GMT
I think that with 1.3 if the server is on a unix host (well, a reasonable
one like freebsd, solaris, linux, and probably others that have all the
tunings done) that we don't really have to worry about this much.  I worry
about NT though (everything about NT performance says "completion ports,
threads, fibers, miss any of those and you suck").

I'm now giving thought to what data we'd have to log from a live server to
generate a test sequence that was both representative of real-life and
repeatable.  It's really hard to do this with dynamic stuff so I'm
thinking I'll ask for Wired's permission to instrument one of the
static.wired.com servers for a week, and permission to make a big tarball
of all the necessary files.  Then we might have something repeatable and
significant. 

Wired's apache is already instrumented in a manner that makes it possible
to discover exactly which hits were part of which keep-alive session; and
possible to determine exactly what each child did (to look at locality
within a child not just across a server).  The only thing it's missing at
the moment is a log of the beginning timestamp of the requests, which is
far more useful for benchmarking than the ending timestamp; because the
beginning is what you'd need to sequence events later.

I won't be able to do any of this for a month though. 

Dean

On Mon, 18 Aug 1997, Brian Behlendorf wrote:

> At 11:06 PM 8/15/97 -0500, Randy Terbush wrote:
> >Anyone here ever read the MS IIS license agreement?
> >
> >It specifically states:
> >
> >No Performance or Benchmark Testing. You may not disclose the 
> >results of any benchmark test of either the Server Software or 
> >Client Software for Internet Information Server to any third party 
> >without Microsoft's prior written approval.
> 
> As stupid as it sounds, if we had something like this it might cut down on
> the number of clueless test reports out there.
> 
> 	Brian
> 
> 
> --=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=--
> "Why not?" - TL           brian@organic.com - hyperreal.org - apache.org
> 


Mime
View raw message