Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!cuae2!ihnp4!ihlpe!daryl
From: daryl@ihlpe.UUCP
Newsgroups: comp.sys.atari.st
Subject: Disk speeds and benchmarks
Message-ID: <1510@ihlpe.UUCP>
Date: Sat, 31-Jan-87 14:16:05 EST
Article-I.D.: ihlpe.1510
Posted: Sat Jan 31 14:16:05 1987
Date-Received: Sun, 1-Feb-87 00:03:08 EST
References: <8701290450.AA16048@cory.Berkeley.EDU>
Organization: Engineering Design Process Department, AT&T Bell Labs
Lines: 25

> The test is a simple one... simply start with a blank disk,
> write out a very large file, then see how long it takes to read it all in.

Unless your OS writes contiguous files almost always, this is not a very
useful benchmark.  Disk performance depends on so many things, it is almost
impossible to get a meaningful benchmark.  The DMA rate is freqently such
a small part of disk bandwidth that it doesn't matter.  Some considerations
are:
	Seek time
	Rotational latency
	Operating system disk block buffering
	Seek and latency OS optimizations (only useful in a multiprocessing
		environment)

It is very obvious on my ATARI 8 bit w/810 that even short sessions on the
machine fragment the disk files across the disk, significantly increasing
the seek time in typical file access.  (The sound the 810 makes when it
is seeking makes it painfully obvious when it is doing so.)

I have found few OS's that allow contiguous files, and none of them wrote
them by default.

Daryl Monge				UUCP:	...!ihnp4!iheds!dlm
AT&T					CIS:	72717,65
Bell Labs, Naperville, Ill		AT&T	312-979-3603