[lug] hard drive speeds

[Sean Reifschneider]

On Tue, Sep 28, 1999 at 01:49:49PM -0700, Jeffrey B. Siegal wrote:
>Ultra2 (which is also wide) is 80MB/sec.  Ultra-wide is 40MB/sec.

A lot of people are talking *INTERFACE* speeds, when the more important
thing in most low-end systems is the actual mechanism performance.  If
the interface is Ultra2, but the mech can only provide 10MBps, the only
advantage is if you are putting 8 drives on one channel.

Evi in her presentation several months ago to BLUG said they had tested
various hard drives and found that IBMs were actually significantly
faster than the others (I presume this was using a SCSI interface).
I'll second the recommendation -- while I haven't done side-by-side
tests, IBMs are the only drives I use these days.

For performance, RAM at 10ns is a huge win compared to a hard drive at
10ms.

Remember that while spindle speed is an easy measure to get, density is just
as important.  I would expect a 5400RPM 18GB drive with two platters to give
better performance than a 10kRPM 18GB drive with 4 or 5 platters.  If the
density is twice as high, it effectively doubles the throughput.

This is one place where IDE drives are really winning.  For some reason, the
really high density drives tend to be coming out IDE first.  Modern IDE
with a single drive per channel is quite a respectable performer.

As Rob mentioned the biggest win is if you can increase the number of
mechanisms providing data.

Remember that with UltraSCSI, you have SERIOUS cabling limitations --
1 foot between devices, 1 to 1.5 meter max cable lengths limit you to
around 3 devices.  Ultra2 is Low Voltage Differential, which increases
the tolerance in the cabling dramaticly, but you're still talking
40Mhz signalling speed -- make sure you're using high-quality "woven"
cable.

Another consideration is that 7200RPM discs operate quite a lot warmer than
5400RPMs, and 10KRPM are even worse.  If you're using 10KRPM drives, you'd
better make sure that your case can handle it.  Many of the 10KRPM drives
will not void their warranties unless you have them directly fan-fed with
air, and some even recommend mounting heat-sinks to the drive mechanisms.
Make sure the case is extra rigid so it doesn't set up resonant frequencies
that end up literally shaking the drive apart...

>Usually the lower sector numbers (which usually correspond to the ouside edge of
>the platter) are faster than the higher numbers (inside).  Partition the drive

Unfortunately, there's no real way to tell how a particular drive allocates
it's sectors.  Wether lower is inside or outside, wether they allocate
from one platter to the next in a single cylindar, or allocate all cylindars
on one platter and then start over on the next.

There was a fascinating presentation at Usenix a couple of years ago where
somone was trying to apply these sorts of optimizations to the file-system,
and he found that the only way one could tell was to actually benchmark
every drive mechanism.  He was seeing relatively good performance improvements
by allocating physicly at the outer edge of the platter though.

Performance optimization is hard.

Sean
-- 
 Got Source?
Sean Reifschneider, Inimitably Superfluous <jafo at tummy.com>
URL: <http://www.tummy.com/xvscan> HP-UX/Linux/FreeBSD/BSDOS scanning software.