[Beowulf] Differenz between a Grid and a Cluster???

Mark Hahn hahn at physics.mcmaster.ca
Tue Sep 20 05:30:46 PDT 2005


> Setting up a Grid can be more political than technical.

I'd even go so far as to say that setting up a grid is basically
a PR measure, whereas setting up a cluster is usually done for a 
specific set of practical reasons ;)

grids are based on the absolutely mistaken premise that computing 
is commodified and generic.  you just arrange the plumbing and the 
flops will flow to whereever they're needed.  I've never quite been
clear on whether gridophiles believe this, and understand that there
are different instruction sets, different cache sizes, different clocks,
different sizes and speeds of memory, etc.  maybe they're simply 
comfortable working with the least common denominator - a portable 
language like java/perl/etc and nothing but embarassingly parallel codes.

for example, my organization is currently acquiring >6K cores in the
form of four large clusters.  single admin domain, but we're not trying
to push a grid because each cluster is specialized for a particular kind
of job, and none of the cycles are generic.  if someone does an MPI code
that runs well on our 3072-core "throughput" cluster (small memory,
dual-core, non-full-bisection gigabit), it will by definition NOT make
effective use of, say, the big-memory, single-core, full-bisection cluster.

that said, we will probably link or unify the queueing systems, since if
there are idle cpus on a cluster, some kinds of jobs can profitably run
anywhere.  I'd tend to call that a single cluster, though, not a grid.

regards, mark hahn.




More information about the Beowulf mailing list