Benchmark results

[David Collier-Brown]

Denis Sbragion wrote:
> Are you sure about this? I benchmarked about 12 Mb/s of transfer rate on a
> recent IDE drive (Fijitsu 10 Gb, Linux 2.0.36, PII 266, Asus MB).

	The maximum throughput of a drive, if reading purely sequentially,
	is a simple factor of bit-density/track and rotational speed...
	Adrian Cockroft provides data for a 5400 rpm disk, using a
	seek-rotate-read scenario, that gives 432 KB/S and a response time
	of 19ms [In Sun Performance & Tuning, 2nd ed p207].  This is 
	reasonable: by counting start-ios I got 60/sec, and 480 KB/S with 
	8 KB reads.

	I like the old approach of counting startios on a real system:
	(iostat's tps column). This is an old DEC/IBM measure, and
	is easy enough to do.

	You may be measuring buffering on your fast disk: I've seen
	some amazing transfer rates for slow devices which can do
	full-track reads.  This is A Good Thing, but you need to lay
	our the data to avoid much seeking.

	By the way: I'm used to getting very different results in seemingly
	similar tests... vendors make it hard to compare their stuff (:-))

--dave (I hate all vendors: I am a vendor) c-b
-- 
David Collier-Brown,  | Always do right. This will gratify some people
185 Ellerslie Ave.,   | and astonish the rest.        -- Mark Twain
Willowdale, Ontario   | http://java.science.yorku.ca/~davecb
Work: (905) 477-0437 Home: (416) 223-8968 Email: davecb at canada.sun.com