[Beowulf] Re: Re: Home beowulf - NIC latencies

[Rossen Dimitrov]

>>>
>>> So if you run an MPI application and it sucks, this is because the 
>>> application is poorly written ?
>>
>> Patrick, here the argument is about whether and how you "measure" the 
>> "performance of MPI". I guess you may have missed some of the 
>> preceding postings.
> 
> No, I was pulling your leg :-) The bigger picture is that MPI has no 
> performance in itself, it's a middleware. You can only measure the way 
> an MPI implementation enable a specific application to perform. Only 
> benchmarking of applications is meaningful, you can argue that 
> everything else is futile and bogus.

Actually, we have been arguing this exact argument for quite some time, 
which might sound odd as we are a commercial MPI vendor :) The whole 
idea is that focusing too much on microbenchmarks and then extending the 
results from these to characterize a whole parallel system does not seem 
to be right thing to do (or at least not the only thing to do) but on 
the other hand, I often see it being done.