*BSD News Article 75450


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!news.eng.convex.com!newshost.convex.com!newsgate.duke.edu!news.mathworks.com!newsfeed.internetmci.com!in3.uu.net!newsxfer2.itd.umich.edu!news.itd.umich.edu!naughty.monkey.org!pha
From: pha@monkey.org (Paul H. Anderson)
Newsgroups: comp.unix.bsd.freebsd.misc
Subject: Re: what is the fastest scsi controller card?
Date: 5 Aug 1996 13:07:51 GMT
Organization: University of Michigan
Lines: 48
Message-ID: <4u4rn7$dhc@lastactionhero.rs.itd.umich.edu>
References: <4tjllu$he@brians-pc.brians-domain.com> <31FDB961.5AAB@cccd.edu> <4u4gda$k8v@itchy.serv.net>
NNTP-Posting-Host: 141.211.26.102
X-Newsreader: TIN [version 1.2 PL2]

Sean T. Lamont (zeno@serv.net) wrote:
: In article <31FDB961.5AAB@cccd.edu>, Michael Peer  <mpeer@cccd.edu> wrote:
: >Brian Bayorgeon wrote:
: >> 
: >> What is the balls to the wall fastest scsi controller card
: >> I can lay my hands on for a news box??
: >> 
: >> Thanks
: >I am using 2 AHA-2940UW with FAST and wide SCSI drives for my spool and
: >over areas.  It seems to be working fine.
: >
: >

I've been trying to find the fastest combination for sequential access, which is
different than your case, but still hopefully interesting.

I've found the combination of the Adaptec driver and 2940UW to give higher throughput
for multiple drives than the NCR wide controller.  When doing large sequential reads,
I want to be able to keep a bunch of drives busy, and the 2940UW with 4 wide drives
gave an aggregate of about 18.5MB/sec through the controller through large
sequentially laid out UFS files.  This was on a P6/200 with 256MB RAM (the files were
each in the gigabyte range, so caching wasn't an issue). 

The NCR controller (both that I've tested) seems to peak at about 8MB/sec for the
controller, no matter how many drives are being read or written.

For a single drive, max throughput for seqential IO seems to be directly proportional
to rotational speed and bit density on the disk - when going through the UFS, there
is some overhead, maybe 20% or so (not carefully measured).  The fastest drive I saw
specs for is one that wasn't shipping last I checked a few weeks ago - the Barracuda
9.  I think the potential range is 9-12MB/sec, but check the specs to be sure.  BSD
will probably see about 7-10MB/sec with the overhead factored in. 

I don't have specs directly comparing the Barracuda-2 with the Barracuda-2LP, but one
person who wrote to me indicated the Barracuda-2LP or maybe 4LP was doing somewhat
better performance than I was getting on my Barracuda-2 drives (8mb/sec instead of
about 6.5mb/sec).  I don't know how he was testing - raw or through the filesystem.

So, barring me being able to get Barracuda-9 drives anytime soon (Christmas was one
quote I heard), I bought a bunch of Quantum 4.3GB drives at $800 apiece, and am
pretty pleased with their price performance.

None of the testing I've done has anything much to do with random I/O - all I care
about is sequential... (until freebsd is fully multithreaded, anyhow...).

Paul Anderson
pha@pdq.com