Newsgroups: comp.parallel,comp.parallel.pvm
From: tom@elsa.gsfc.nasa.gov (Tom Head)
Subject: Re: Distributed Proccessing - How to measure speedups?
Organization: Oceans & Ice Branch
Date: Mon, 1 Aug 1994 15:45:47 GMT
Message-ID: <Ctv54C.8tw@dcs.ed.ac.uk>

In article <Ct8JBz.KDJ@dcs.ed.ac.uk> nfotis@theseas.ntua.gr (Nick C. Fotis) writes:
>
>- How can / should I measure the efficient execution of programs in a
>  heterogeneous network?
>
>
>We don't know what to measure anymore - the CPU seconds spent in each CPU
>are rather irrelevant, as we may have CPUs from 30 SPECfp to 300 SPECfp each
>- and the network delays aren't the same on each machine.
>
>We cannot isolate the network, since it's not our own, and the wall-clock
>time is not adequate metric, since he does research on efficient parallel
>algorithms (till now on homogeneous, shared memory machines)
>

  The problem can be worse than that if cache-based CPU's are part of your 
heterogeneous network. As your data set per processor decreases with the
increase in cache-based CPU's allocated to your problem, your cache miss rate 
can be substantially lowered, giving possibly super-linear timing results that 
are not a result of your algorithm, but the hardware employed for a given 
problem size.

Thomas Head



