OSEC

Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com
 
Subject: Re: hmmmm....
From: Liviu Daia (Liviu.Daiaimar.ro)
Date: Tue Sep 19 2000 - 18:01:18 CDT

On 19 September 2000, Harry <mail-nanofizbin.com> wrote:
> on 9/19/00 9:44 AM, Wietse Venema at wietseporcupine.org wrote:
>
> > so that even a steenkeeng IDE disk might do the job.
>
> 'steenkeeng' IDE disks have come a long ways in the past 5 years - the
> stereotype of the slow IDE drive is becoming more and more of a myth
> (unless you use some way old hardware, like a P90 with an original RLL
> drive).
>
> I have no performance issues with properly tuned IDE drives (hdparm is
> my friend)

    Well, yes and no. Thanks to various improvements that come standard
with the IDE controllers these days (ATA-100, ATA-66, UDMA etc.), newer
IDE disks are indeed capable of much better throughputs than, say, 5
years ago. But the real problem with these disks is elsewhere: the
controller would still hog the CPU for the full duration of a command.
That's fine for a single-user system (unless you do some really weird
things, you won't see any spectacular difference if you switch to a
SCSI disk on your machine at home), but it's a severe limitation for a
server. Run 10 processes at the same time, each of them trying to do
many "small" disk operations (say, Postfix with 10 smtpds, receiving
thousands of 1k messages at full speed :-)), and watch your system slow
down to a crawl. Do that with a SCSI disk, and you won't even notice
the load. This is an architectural problem, which won't go away if you
buy a faster IDE disk. If you're stuck with a PeeSee and you expect to
have several processes continuously thrashing the disk, or more than 4-5
people logged in at the same time, do yourself a favor and get a real
controller and a SCSI disk. Trust me, it worth the money.

    Regards,

    Liviu Daia 

-- 
Dr. Liviu Daia               e-mail:   Liviu.Daiaimar.ro
Institute of Mathematics     web page: http://www.imar.ro/~daia
of the Romanian Academy      PGP key:  http://www.imar.ro/~daia/daia.asc