*BSD News Article 7512


Return to BSD News archive

Xref: sserve comp.arch:27939 comp.unix.bsd:7561 comp.os.linux:15022
Newsgroups: comp.arch,comp.unix.bsd,comp.os.linux
Path: sserve!manuel.anu.edu.au!munnari.oz.au!uunet!utcsri!torn!nott!cunews!revcan!quantum!danh
From: danh@quantum.on.ca (Dan Hildebrand)
Subject: Re: IDE faster than Mips SCSI disk
Message-ID: <7j+q=2d@quantum.on.ca>
Date: Fri, 06 Nov 92 15:08:46 GMT
Organization: Quantum Software Systems
References: <1992Nov6.033942.21194@ntuix.ntu.ac.sg>
Lines: 51

In article <1992Nov6.033942.21194@ntuix.ntu.ac.sg> eoahmad@ntuix.ntu.ac.sg (Othman Ahmad) writes:
>These figures are not meant to be a thorough study. More of a provocation
>for more soul searching. The widespread belief that SCSI-2 is "defininely
>faster than IDE" is questionable, in fact under some circumstances, entirely
>false.

One significant point you're missing is that IDE drives use the processor 
to move the data in and out of the controller card, while SCSI on a PC with 
an Adaptec 1542 or 1742 controller card is a bus master. This fact lets the 
host processor continue computing while the SCSI card does the data 
transfer. If the processor is consuming or producing data at near the rate 
the SCSI card moves it, the resulting pipelining provides in significantly 
better performance. For a DOS system, which just waits while the disk I/O 
is in progress, it doesn't make a lot of difference, but with a 
multitasking OS, or a filesystem which is doing pre-reads and post-writes, 
the overlap of I/O with computation is a worthwhile goal.

>The mips machine under test runs on ultrix. Although it has up to 10Gbyte 
>of hard-disk, it is not so heavily loaded. We only use it for email and
>news feed. Fragmentation can be severe because I cannot even have 32 
>megabyte free space in /usr/tmp , only 30 Mbytes.
>
>	IOZONE: Performance Test of Sequential File I/O  --  V1.14b (1/31/92)
>	Operating System: ULTRIX
>
>IOZONE performance measurements:
>	345684 bytes/second for writing the file
>	641985 bytes/second for reading the file
>
>486/33 Maxtor 7120 200Mbyte
>
>IOZONE performance measurements:
>	367586 bytes/second for writing the file
>	499942 bytes/second for reading the file

Running IOZONE with QNX on an ISA bus, 33 MHz 486 with only the 8K 
processor cache ( no motherboard processor cache ), I sustain about 1.6 
Mbytes / second read and about 800 Kbytes / second write ( 32 Mbyte files 
). This is essentially the platter speed of the drive. Of all the IDE 
drives I've tried so far, the best performance I've seen was about half 
that on both read and write, respectively. Local bus will help to speed up 
IDE drives, but local bus will also speed up SCSI. Bus mastering is still 
the important difference. I suspect a bus mastering IDE controller card 
could compete with an Adaptec SCSI controller and disk drive.


-- 
Dan Hildebrand                     email: danh@qnx.com
Quantum Software Systems, Ltd.     QUICS: danh  (613) 591-0934 (data)
(613) 591-0931 x204 (voice)        mail:  175 Terrence Matthews          
(613) 591-3579      (fax)                 Kanata, Ontario, Canada K2M 1W8