[Subject Prev][Subject Next][Thread Prev][Thread Next][Subject Index][Thread Index]

RE: [LI] Linux Myths



> I tried the Apache Benchmarker today on one Apache on Linux
> and one IIS on
> NT on _exactly_ similar machines, and similar loads on them
> (I guess, the
> Linux one was more loaded) and Apache definitely outperformed IIS.
>
> But this was a definitely crude way to do. Can anyone give
> any pointers to
> documents that give guidelines for benchmarking?
>
> regards,
> jaju
>
First, Benchmarking is not a trivial task.
Second, No benchmark can be generic and you can do two benchmarks in   
different ways
to show exactly the opposite results...
So Benchmarks should be done within a project, on the specific hardware   
that the
project wants to use, with the datas and the programs that the project   
wants to
execute and serve. Then the benchmark should be used to mesure precise   
things (user point
of vue (speed of answers to requests), system point of vue (run queue   
length, cpu load,
disk load). A benchmark should have a goal (usability, choice of   
hardware, choice of
algorithm, choice of software, choice of OS, lessen the cost of hardware)
and not combine all the goals at the same time. The results depends on   
things as scientific
as the mesures you take to things as human as the capacity of the guy who   
will have to
use the system afterwards.
Otherwise, the benchmark is useless.
All these benchmarks are just journalist games devoted to make some trees   
cut and
pseudo scientific papers sell.

have fun
pascal  
--------------------------------------------------------------------
The Linux India Mailing List Archives are now available.  Please search
the archive at http://lists.linux-india.org/ before posting your question
to avoid repetition and save bandwidth.