[Beowulf] $2500 cluster. What it's good for?

Dean Johnson dtj at uberh4x0r.org
Mon Dec 20 12:35:36 PST 2004

On Sun, 2004-12-19 at 18:31, Andrew M.A. Cater wrote:
> If you have a long running problem - DON'T start it now. If it needs to
> run for two years - buy next year's equipment (which is twice as fast as
> today's) and run it for just one year. One year wait then one years intensive
> compute - and you're still ahead. Next year's computer is
> _automatically_ faster and potentially much better value for your $$ :)

If you have a long running problem - don't start it right now. Spend
some time adding checkpointing and then start it. That way you can start
it in a month, get 11 months in, and then restart it on new faster,
cheaper hardware. If you are running a code that pops out '42' after two
years time, you got "some issues". One must not forget that the cluster
part may be the fun part, but ultimately the science is the important
part. It would be fun to report to a funding organization at a review
that your sole progress is that you clicked the refresh button on a
pricewatch.com webpage 63,714 times for the preceding quarter.



More information about the Beowulf mailing list