Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.carno.net.au!harbinger.cc.monash.edu.au!munnari.OZ.AU!news.ecn.uoknor.edu!feed1.news.erols.com!cpk-news-hub1.bbnplanet.com!news.bbnplanet.com!su-news-hub1.bbnplanet.com!csn!nntp-xfer-1.csn.net!csn!nntp-xfer-2.csn.net!yuma!pyeatt
From: pyeatt@.cs.colostate.edu (Larry Pyeatt)
Newsgroups: comp.os.linux.misc,comp.os.linux.setup,comp.unix.bsd.bsdi.misc,comp.unix.bsd.misc
Subject: Re: IDE vs SCSI (was Re: Linux vs BSD)
Date: 24 Jan 1997 12:09:30 -0700
Organization: Colorado State University, Computer Science Department
Lines: 47
Sender: pyeatt@ravel.cs.colostate.edu
Distribution: world
Message-ID: <5cb1da$1kk@ravel.cs.colostate.edu>
References: <32DFFEAB.7704@usa.net> <87k9p4rckd.fsf_-_@murkwood.gaffaneys.com> <5c8a39$7tn@usenet.srv.cis.pitt.edu> <5c8odp$14q@ravel.cs.colostate.edu> <5c97cd$f6h@cynic.portal.ca> <87g1zrb7gv.fsf@murkwood.gaffaneys.com>
NNTP-Posting-Host: ravel.cs.colostate.edu
Xref: euryale.cc.adfa.oz.au comp.os.linux.misc:154110 comp.os.linux.setup:93818 comp.unix.bsd.bsdi.misc:5728 comp.unix.bsd.misc:2022
In article <87g1zrb7gv.fsf@murkwood.gaffaneys.com>, Zach Heilig <zach@blizzard.gaffaneys.com> writes:
|> cjs@cynic.portal.ca (Curt Sampson) writes:
|>
|> > In article <5c8odp$14q@ravel.cs.colostate.edu>,
|> > Larry Pyeatt <pyeatt@.cs.colostate.edu> wrote:
|>
|> > >While a modern SCSI system can sustain > 16MB/s and my 5 year old
^^^^^^
|> > >SCSI controller and drives can sustain > 8MB/s
^^^^^^
|> > Which SCSI drive do you have that sustains 16 MB/sec, and how did you
|> > benchmark this?
|>
|> More than likely, he had 4 or 5 disks on the same controller, with
|> some sort of concatenated filesystem (FreeBSD does this). These sorts
|> of figures are sustainable when you spread the work over among many
|> drives.
That is correct, but it doesn't take THAT many drives.
That throughput can be attained with a Buslogic
PCI ultra SCSI card and 3 quantum drives (8ms access). The SCSI
controller can issue up to 32 requests for each drive and the
drives can service those requests in any order. This arrangement
allows the SCSI bus to be almost totally saturated with data.
I used a slightly hacked version of byte unix benchmarks.
The hack made it perform simultaneous tests on all hard drives.
I don't have the byte unix results for the above, and don't have
access to that machine right now, but here are the results from
my 5 year old controller. ( I have replaced the 3 old drives
with 2 newer ones and did not run the benchmark until today)
Unfortunately, there were 9 users at the time, so the benchmark
numbers are a bit low:
Buslogic 445S VLB SCSI II adapter (5 years old)
Cyrix 5x86 120Mhz (pretty new)
2 quantum drives (8ms access) (2? years old)
The benchmark gave 3069 Kbytes/sec copy (read and write)
which is 6138 Kbytes/sec thoughput. Adding a 3rd drive and
kicking off the 9 users would significantly impact performance.
--
Larry D. Pyeatt All standard disclaimers apply.
pyeatt@cs.colostate.edu Void where prohibited.
http://www.cs.colostate.edu/~pyeatt - If it won't go with a sledgehammer,
- don't force it.