"Mark Burns" <mark@iolofpa.com> wrote:
> "Bob O`Bob" <bob@cluestick.org> wrote in message


> >
> > I guess I've "fallen for it" then, because NT doesn't scale.


> > Clustering does serve the same purpose, and often does so at
> > reduced total cost, but it is not really the same thing.


I should clarify what I meant here. When people argue that it's impossible
to build really large scale systems (particularly the server-side web stuff
that I do, and like doing) on Windows, I usually think the person making the
case is either uninformed or a *nix bigot.

Granted, there are some apps that really take a really huge server with a
lot of CPUs, and with the possible exception of the 2000 series of MS
enterprise apps, there isn't much out there that scales beyond 4-8
CPUs/server right now, but how often are there real-world apps that require
a ton of horsepower and don't scale out?

> Bob, uh... ever hear of Windows 200 DataCenter Server?
> I think THAT puppy probably "scales" just fine. Never put mits
> to it myself, but from what specs I've read...it should f'n
> ro...uh,..scale well<g>.


I haven't seen any independent numbers on DataCenter Server yet.
What I consider a fair test:

Give Microsoft, Sun, IBM, and Red Hat a budget of $500K (or something around
that) to build a highly scalable app [detailed spec provided] -- they can
spend the cash however they like, though all programmer time will be figured
at a fixed, hourly rate. Tell all the participants at the start what numbers
you're trying to maximize.

Then run performance and reliability numbers on the resulting app. No using
beta software.

I'd bet that you'd get 4 really different solutions in terms of how they
were architected, and there wouldn't be too much of a performance
difference.

--
David A. Rothgery
Consultant
Spherion, Inc.
davidrothgery@spherion.com
drothgery@myrealbox.com