Re: Disk throughput Questions

From: Anthony D'Atri <aad_at_statsci.com>
Date: Fri, 3 Mar 1995 09:43:58 -0800 (PST)

>a) Why is the Filesystem tests better that the raw devices?

You've got a lot of memory -- you're probably writing mostly into
buffers, which the OS then asynchronously writes to disk. These sorts
of benchmarks are best done with something like iozone, but it might be
interesting in this case to add a sync after the dd to the fileystem and
see how the number changes.

>d) Are the tests that I have performed valid at all?

That's a can-o-worms I won't touch.

>Is there another more valid method to use?

There's a utility out there called "iozone" for doing this sort of
thing. Archie around for it.
Received on Fri Mar 03 1995 - 12:45:07 NZDT

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:45 NZDT