*BSD News Article 34151


Return to BSD News archive

Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!yeshua.marcam.com!MathWorks.Com!news2.near.net!news.delphi.com!usenet
From: John Dyson <dysonj@delphi.com>
Newsgroups: comp.os.386bsd.misc
Subject: Re: SCASI or IDE disk?
Date: Thu, 11 Aug 94 13:05:17 -0500
Organization: Delphi (info@delphi.com email, 800-695-4005 voice)
Lines: 25
Message-ID: <JI5wEFV.dysonj@delphi.com>
References: <salem.136.2E48D0EF@hauk.hsr.no> <michaelv.776576787@ponderous.cc.iastate.edu> <salem.146.776611896@hauk.hsr.no>
NNTP-Posting-Host: bos1g.delphi.com
X-To: Salem, Lazaro <salem@hauk.hsr.no>

 
This is great... You are really getting a discussion going!!!
 
One of the comments made is that the command overhead is greater under
SCSI.  That is true, but the new *BSDs do clustering which lessens the
effect of that overhead significantly.  You mentioned that you have
IDE interfaces -- are they indeed faster than the standard AT 2MB/s
type?  I have a WD540, and the best that I can get is 2MB/s (really) under
FreeBSD V2.0 with a standard IDE controller.  I can easily get over 3MB/s
with a SCSI 2217.  Of course that is comparing apples and oranges.  I think
that the standard AT bus does limit you to around 2MB/s for PIO without
some tricks.  If you are connected to the local bus, the situation is
better, but I think that you are still limited to the performance of the
drive, causing the CPU to block on each byte transferred.  Using SCSI and
busmastering, the overhead can be less because of less CPU intervention.
Of course, the controller still grabs the bus and slows down or stops the
CPU.
 
Some of the EIDE methods will be better than IDE, but I tend towards the
conservative -- and my recent experience with busmastering SCSI says that
it is currently better than standard IDE.  BTW, 2-3yrs ago, many times
SCSI was not very good, but now it is.
 
John
dyson@implode.root.com