Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 14 Apr 1999 21:52:07 +0000
From:      Neil Blakey-Milner <nbm@mithrandr.moria.org>
To:        advocacy@FreeBSD.ORG
Subject:   Re: Benchmarks
Message-ID:  <19990414215207.A64591@rucus.ru.ac.za>
In-Reply-To: <Pine.BSF.4.10.9904141517130.29587-100000@freebie.dp.ny.frb.org>; from Seth on Wed, Apr 14, 1999 at 03:21:36PM -0400
References:  <XFMail.990414204955.asmodai@wxs.nl> <Pine.BSF.4.10.9904141517130.29587-100000@freebie.dp.ny.frb.org>

next in thread | previous in thread | raw e-mail | index | archive | help
Hi

Most of this culled from an email I sent to a local list:

> Just something that was posted to #freebsd yesterday.  I think it's pretty
> interesting for a couple reasons I'll mention below:
> 
> What's hidden in here is that Microsoft sponsored this test (paid
> Mindcraft to do it, I guess).

The numbers may seem slightly high, but there are some reasons why NT
on that machine might outperform Linux [and possibly FreeBSD] (running
those sorts of services).

NT's integrated SMB support in it's multiprocessor kernel versus Linux's
[and again, FreeBSD's] userland daemon - there might not be much in that
fight really.  Backed up with nice RAID drives, NT is not (that?) slow
for file serving.

A previous ZDnet article 
(http://www.zdnet.com/pcweek/stories/news/0,4153,387766,00.html/0,4153,387766,00.html), 
however, claims that with a VA Research-provided quad-xeon machine gave
"performance [is] in line with Windows NT 4.0 on comparable hardware" in 
a PC Week lab test.

I'm quite sure that Apache's performance knobs probably weren't turned,
and probably not the Linux high-power knobs either. (as documented in
the Apache manual)

Also, I've seen cited before in similar tests that NT provides a really
fast function (getfile?) that doesn't have an equivalent in our world.
Don't take my word on that, though.

It would be interesting to know what section the bottlenecks occurred
in, in each test.

> Something I found interesting.  There was a post a while back that
> suggested that Microsoft would not allow benchmarks of its own products to
> be posted by 3rd parties.  Did I misread this?  If not, it seems mighty
> unfair of them to post benchmarks only when it's in their interest to do
> so.

Of course it's unfair.  And, of course, Microsoft must have known
beforehand that they were likely to win.  It's also good business
sense - you don't want nitwits benchmarking machines that are poorly
configured with your OS (which would probably be as poorly configured).

It would not surprise me if Microsoft provided the hardware specs,
aiming at the best difference between NT and Linux hardware, memory
system, and multiprocessor support.  The key may be in the hardware, as
the PC Week test seems to indicate.

I'll leave FreeBSD's likely performance to the experts.

Neil
-- 
Neil Blakey-Milner
nbm@rucus.ru.ac.za


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-advocacy" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19990414215207.A64591>