[rescue] Re: Re: FreeBSd/SCSI, PPro Overdrives

[Greg A. Woods]

[ On Wednesday, January 30, 2002 at 07:50:54 (-0800), Robert Novak wrote: ]
> Subject: [rescue] Re: Re: FreeBSd/SCSI, PPro Overdrives
>
> So no comparison on Pentium Pro Overdrive vs Pentium 2/3?

I dunno about PPro-Overdrive, but I do notice a steady, but nowhere near
stunning, improvement between PPro, P-II, and P-III.  Of course the
steady and far more stunning improvement in clock frequencies is
definitely visible!  (250MHz vs. 300MHz vs 700MHz for the machines I've run)

As for FreeBSD performance, well so far it is the only *BSD with which
I've been able to get full performance from Intel's STL-2 motherboards
and Ultra-160 LVD drives like the new Seagate Cheetah.

In particular that's because of the highly advanced FreeBSD SCSI
subsystem.  It fully supports the AIC-7899 chip (Ultra-160 + Ultra-80 on
one chip).  (NetBSD's port of that driver is now nearly two years old
and it can only drive the Ultra-160 bus at half speed, and it fails
miserably with more than two disks attached).  I've got a couple of
those machines at a customer site running squid and they fly.  We peak
at 4-5 thousand requests per minute on each one with aggregate
throughputs of well over a megabyte/s to the clients.  When I did simple
disk benchmarks with Seagate Cheetahs a disk-to-disk copy was an average
of 35 megabytes/s.  Another very similar machine runs in Karachi,
Pakistan with about 8000 dialup ports and not nearly enough backbone
bandwidth.  It's got two processors spinning under FreeBSD-4.4 now and
IIRC it almost always exceeds 3000 requests per minute, peaking at well
over 6000 requests/s.

There are lots of things I don't really like about FreeBSD, but
performance and reliability, even on high-end hardware, are not among
them.

-- 
								Greg A. Woods

+1 416 218-0098;  <gwoods at acm.org>;  <g.a.woods at ieee.org>;  <woods at robohack.ca>
Planix, Inc. <woods at planix.com>; VE3TCP; Secrets of the Weird <woods at weird.com>