[Beowulf] New HPCC results and the Myri viewpoint

Keith D. Underwood kdunder at sandia.gov
Thu Jul 21 12:07:46 PDT 2005

> ... did I miss you signing up to run on our customer benchmark
> cluster? ;-) In any case, I can point you to quite a few sets of
> online performance numbers where scalability is being hurt by
> short-message performance, which is the thing that everyone means when
> they say "latency sensitive". For example, quantum chemistry is very
> well known for being hard to scale. Climate involves running small
> datasets with large numbers of timesteps, and that inevitably ends up
> being latency sensitive. And so on.

In many of those cases, the latency impact on throughput (i.e. the
streaming numbers) are what is important.  I don't know of anything that
does a tight loop of "small amount of work, ping-pong with one other
processor, repeat". In most networks, message throughput == 1/latency 
(approximately, anyway).  It doesn't have to be that way.


Keith D. Underwood <kdunder at sandia.gov>

More information about the Beowulf mailing list