[Zope] ZOPE, Medusa and the ZD WebBench 3.0 test suite...

[Chris Petrilli]

Lies, damn lies and benchmarks.
     -Samuel Clements via me.

> My question, and the one I expect is on many of your minds, 
> is... How does a
> ZOPE/Medusa server config stack up? My hope and expectation that the
> underlying Medusa architecture might really shine when compared to the
> pre-allocating multi-process approach of Apache, etc.

Well, this is for lack of a better statement, irrelevent to Zope.  They
are asking (last I looked) for a static-page system, with some tiny
CGIs, but this isn't really what Zope is about.  EVERY page is Zope is
dynamic, so if you compared a 100% CGI site with Zope, then it *might*
be a valid comparison, if you also pulled all the data out of a database
on the fly as well...

IMPORTANT:
The only benchmark that matters is the one that is your application, and
on that, Zope excells.

As for asyncronous v. threading/forking servers.  Zeus, which is the
fastest 'static' web server int he world, and what everyone uses for
benchmarking on UNIX is also an asyncronous server.  Different servers
have different behaviours.

One of the common problems with IIS v. Apache/etc is that Apache isn't
tuned to saturate a small number of clients (which is all they are
simulating) with very big pipes, but instead, a huge number of clients
with smaller bandwidth (what the Internet really looks like).  Also,
IIS4 runs a "cache" inside the server, which you could simulate iwth
Squid, or some such... One must be careful comparing the flavor of two
things sold as oranges when one is distinctly shaped like a pumpkin.

Chris
--
| Christopher Petrilli                        Digital Creations
| petrilli@digicool.com                 http://www.digicool.com