We have three clusters, the first two purchased a year ago.
The small one is 20 nodes:
- MSI K8D Maaster MB
- Dual Opteron 242 (1.6GHz)
- 4 GB RAM
- Two WDC 2500
- 1U rackmount chassis (Chenbro, I think)
The large one is 152 nodes, same MB, CPUs, and disks, only 2GB RAM.
- 1U Appro chassis
The new one is 160 nodes, Opteron 250s, 4 GB RAM, and same disks,
- Appro blade chassis.
We also have a sprinkling of development nodes using the 244 CPUs, some
of which are using Adaptec 39320 HBAs to connect to RAIDs for NFS service.
(I would use Coda, but our Solaris clients do not have that.)
These systems are used for processing tropospheric emissions data in
an attempt to retrieve chemical consituents in the air. They are not
used in the normal sense of "clusters"; we just run a lot of independent
parallel jobs.
eyc