[Beowulf] Three notes from ISC 2006

[Joachim Worringen]

Patrick Geoffray wrote:
> Recently, I have been thinking about something that you may like. With 
> motherboards with 4 good PCIE slots coming on the marketing (driven by 
> SLI and such), it could be doable to have a reasonably sized machine, 
> let's say 64 nodes, with 4 different interconnects in it. If Intel or 
> AMD (or any good will) would donate the nodes, and the interconnect 
> vendors would donate NICs + switch + cables, and a academic or 
> governmental entity would volunteer to host it, you could have a testbed 
> accessible by people to do benchmarking. The deal would be: you can use 
> the test bed but you have to allow your benchmark code to be available 
> to everyone and the code will be run on all interconnects and the 
> results public.
> 
> What do you think of that ?

It's every MPI/compiler/... implementor's fantasy (including me), but 
unfortunately, it won't happen.

SPEC benchmarking is an example for this. The advantage of SPEC is that you only 
need one machine to reproduce the results, not a whole cluster. This does of 
course mean such a reference cluster is even more important from the purely 
technical or scientific perspective.

It won't happen partly for cost reasons: you need not only the hardware, but 
also manpower and other running costs. You need to update the cluster once or 
even twice a year to be always on the edge of technology - other numbers are not 
interesting for marketing. No marketing interest, no funding.

But mostly because the vendors want to make sure they get "the right" numbers 
for their evil whitepapers. Not that I assume that any vendor will actually 
publish fake numbers, but every vendor wants to make sure that the numbers are 
the best numbers possible, and needs his own people to ensure this.

-- 
Joachim - reply to joachim at domain ccrl-nece dot de

Opinion expressed is personal and does not constitute
an opinion or statement of NEC Laboratories.