Subject: Re: lmbench disk figures on SCSI vs. IDE
To: Laine Stump <laine@MorningStar.Com>
From: Curt Sampson <cjs@portal.ca>
List: port-i386
Date: 01/05/1997 21:55:32
On Sun, 5 Jan 1997, Laine Stump wrote:

> [IDE:]
>    <QUANTUM FIREBALL_TM2100A>
>    Disk /dev/rwd0a bandwidth          4474 KB/sec
>    File /usr/tmp/XXX write bandwidth: 3595 KB/sec
> [SCSI:]
>    ahc0: aic7880 Wide Channel, SCSI Id=7, 16 SCBs
>    sd0 at scsibus0 targ 0 lun 0: <SEAGATE, ST15150W, 0023> SCSI2...
>    Disk /dev/rsd0a bandwidth          6677 KB/sec
>    File /usr/tmp/XXX write bandwidth: 1466 KB/sec

This seagate drive is a 7200 RPM Barracuda, right? You should be
able to get much better performance out of it than that. On a P90
running 1.1, I see about 5-6 MB/sec of filesystem performance on
7200 RPM Quantum Atlas drives, with both an NCR 53c810 and Buslogic
controller. What controller are you using? Can you try out a
different controller?

For better filesystem performance figures, you may want to grab a
copy of bonnie (in <http://www2.portal.ca/~cjs/computer/benchmark/>)
and look at those results, too.

I have an IBM 5400 RPM drive on my Multia (266 MHz DEC Alpha; NCR
SCSI controller). This is what lmbench and bonnie say about it:

    Disk /dev/rsd1e bandwidth4229 KB/sec
    File /usr1/tmp/XXX write bandwidth: 3219 KB/sec

Local Disks  -------Sequential Output-------- ---Sequential Input-- --Random--
             -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine   MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
M2B IBM  100  3330 72.5  3279 25.8  1241 15.7  3518 81.1  3605 27.6  54.1  5.3
M2B IBM  100  3295 69.5  3262 25.3  1240 15.3  3486 77.7  3615 26.8  54.1  5.5

So at first blush, lmbench results seem to fairly accurate.

cjs