OSC is updating its hardware with an Intel Pentium 4 (P4) cluster to be installed later this month. Replacing the AMD Athlon cluster, the P4 doubles the current system’s power with a sizable increase in speed.
With a theoretical peak of 2,457 gigaflops, the P4 cluster contains 256 dual-processor Pentium IV Xeon systems with four gigabytes of memory per node and 20 terabytes of aggregate disk space. It will be connected via a gigabit Ethernet and use Voltair InfiniBand 4x HCA, and a Voltair ISR 9600 InfiniBand switch router for high-speed interconnect.
The P4 cluster is a natural progression from the IA32 cluster, a distributed/shared memory hybrid system constructed from commodity PC components running the Linux operating system. The IA32 cluster of 256 AMD Athlon processors will be cascaded to Ohio faculty research teams as part of the Cluster Ohio project in January.
"We try to make the changes as seamless as possible so that they are more evolutionary than a radical change to users," said Doug Johnson, OSC Systems Developer and Engineer.
The cluster is available for general research use including everything from cracking plant genomic codes to doing computational flow analysis. OSC will transition users off of the AMD cluster and on to the new system during the last two weeks of December. This coincides with the Cluster Ohio Rev3 grant process during which time OSC will cascade five clusters to Ohio faculty. For Rev3, OSC will provide five AMD Athlon clusters, each with 24 compute nodes, one login node, and one storage node.
OSC will bring all 256 systems online within the first week of delivery. The P4 cluster will use Infiniband software, which connects conventional supercomputing software with innovative networking technology allowing it to speed up the data flow.
"We are more aggressive in our schedule this time,” said Johnson. “This goes against the misconception that clusters take a lot of time to install."
More information about OSC’s high performance computing resources is available at www.osc.edu/supercomputing.
