The Jupiter CFD Cluster

Computationally demanding CFD calculations are run on the research group's CFD computing cluster.  This enables CFD simulations to be run with high spatial and temporal resolution.

The CFD computing cluster currently consists of 192 Pentium 4 processors running Linux, with the addition of 2 racks of dual-processor, dual-core AMD Opteron processors in progress, this ongoing work is being performed to obtain a total of 544 processors. This cluster currently has 6 terabytes of storage.

Parallelisation between nodes is achieved by the Message-Passing Interface (MPI) protocol through ethernet links, and parallelisation between cores in the Opteron processors is achieved through the OpenMP API. Management of this system is through a high-speed fibre-optic link to the CFD lab, and monitoring is performed through the Ganglia open-source distributed monitoring system for high-performance computing.