Performance Variations using MPI/Myrico

[Greg Lindahl]

On Wed, Apr 25, 2001 at 10:26:27AM -0700, Thomas Davis wrote:

> We are looking for sites that run Intel Linux/SMP(dual)/MPI/Myricom 2k,
> and have experienced performance variations.  IE, you've ran the NAS/FT
> parallel benchmark, densely packed (using all CPU's on the nodes), and
> noted that the runs come up different each time - and the difference is
> not minor (as much as 80%).

There are several sites with exactly that setup, such as the new Maui
machine, and the new NCSA machine.

I can say that with dual Alpha/MPI/Lanai 7 cards, performance
variations on a reasonable variety of benchmark codes (bonus points
for naming the bid this was for) were small, on order of 2%. I am
currently running dual Alphas with Myrinet2000 and haven't formally
measured variability, but it isn't large.

Come to think of it, I think HPTi's bid for the last NERSC cluster RFP
(which you have access to, yes?) has numbers for the NAS benchmarks on
dual Intel/MPI/Lanai 7 cards. These times were not highly variable and
we got the "right answer" compared to similar systems.

-- g