[rescue] Sun 711

[David Passmore]

On Fri, May 03, 2002 at 01:08:44AM -0400, George Adkins wrote:

> this was a 1 gig file, and it took just longer than 10 seconds.  To get 1
> full minute, I's probably have to cat together 8 or 9 gigs of file.  The
> system has 1 gig of RAM and the controller has 32 Megs of cache, but I don't
> see how that would falsely affect this test since it's straight read
> performance (no write caching).

For large sequential reads the storage subsystem, sometimes the filesystem,
and the drives themselves (drives have no battery-backing thus their caches
are purely read cache) will read-ahead several sectors, tracks, or even
entire cylinders-- since the platters rotate under the heads once for either
one sector, track, or a cylinder (only very high-end drives can activate
more than one head at a time), might as well read it all. This is why read
performance is typically so much faster than writes, but it is illusory--
actual reads and writes on the raw platters operate at the same speed.

> I looked at Bonnie, it looks interesting.  Tests several performance aspects, 
> it's small, comes in source code, is non-proprietary, etc.  I would not be 
> using IOZone (Don't do Winders...).

I use a combination of Solaris TNF tracing to profile an application's
storage access, and a Veritas utility called vxbench to generate simulated
loads. TNF tracing is very useful for many other things as well.

David