Re: stochastic fair queueing in the elevator [Re: [BENCHMARK]

[Rod Van Meter]

On Mon, 2003-02-10 at 05:30, ext Rik van Riel wrote:
> On Mon, 10 Feb 2003, Andrew Morton wrote:
> 
> > Could be that sending out a request which is larger than a track is
> > saving a rev of the disk for some reason.
> 
> I guess disks are optimised for the benchmarks that are
> run by popular PC magazines ...

Yes, absolutely they are (I used to work for one such company that no
longer exists).  However, there's a catch -- the disk drive companies
are optimizing for the test as currently constituted, but the test keeps
changing.  So they're optimizing for version N-1 while the magazine is
publishing N.  Can't help it.  Ever heard of the Nyquist frequency?

> 
> After all, those benchmarks and the sales price are the
> main factors determining sales ;)
> 

Yup.  In general, it's capacity first, then reliability, then
performance.  No, wait, not quite; the FIRST criterion is the ability to
ship when you say you will.  If you make Dell late on shipping a bunch
of machines, you WILL feel the pain.

		--Rod