[Benchmarking] Use of the machines, location of the data sources

[Andrea Aime]

Hi,
as you may know, we do have two server machines, which for
the moment I just named "Server A" and "Server B" (given
the hardware differences we could rename them Goliath
and David respectively ;-) )

The main question is, how do we use them?
Possibilities:
1) run the server software on A, use B to run JMeter
2) run the server on both, and JMeter on both
    (run JMeter on A when testing B and vice versa)

1) is less work and covers the most interesting case,
2) is more work and allows to check the influence
of the hardware size on the results

Note: on A we have enough memory to keep GeoServer and Mapserver
fcgi processes running all time, on B, I guess we'll start
fighting with swap space unless we start them only
when we need to benchmark them.

As for data sources interest has been shown my our
host to have Oracle and SDE benchmarked as well.
First off, did we get permission to do so?

Secondly, it seems Postgis will run on A, meaning
it will be local, so the comparison may be unfair
to one side or the other of the equation, depending
on whether CPU or the network prove to be the bottleneck.
Opinions?

Cheers
Andrea