[nSLUG] 2.6 kernel performance

George White aa056 at chebucto.ns.ca
Wed Jun 22 17:40:28 ADT 2005


On Tue, 21 Jun 2005, Donald Teed wrote:

> I read years ago about a study where a group of non-savvy
> PC users had their rigs upgraded while they were out at
> lunch, and no one told them.  The subject of the
> experiment was to determine when an average person 
> notices, all on their own, that the system is faster.
> The study concluded that the average user does not report
> the speed up until there is a 2x performance factor
> increase.  I've often used that as my formula on
> planning upgrades.

Our lab runs a model developed locally that was recently determined in a
NASA study to give the best results of over 30 models in the evaluation. 
Unfortunately, NASA wants to get results this year that will take 
years to compute using our model.  When a single calculation takes 
days to run, a 2x speedup would be real nice to see, but even nicer 
might be to throw a few hundred PIII's at the problem and finish in a
month.

I've seen many new systems that the user thought was fine but 
seemed sluggish to me -- running a few benchmarks often reveals
underlying configuration issues, or, these days, some infestation.
I've also noticed that many inexperienced programmers don't
have a feeling for what run-times to expect, so don't feel there is a
problem when their C code has run times 10x longer than the equivalent
awk script.


--
George White <aa056 at chebucto.ns.ca> <gnw3 at acm.org>
189 Parklea Dr., Head of St. Margarets Bay, Nova Scotia  B3Z 2G6


!DSPAM:42b9ccd167767688110247!




More information about the nSLUG mailing list