Tor – High Performance Computing Cluster

The Tor cluster is a computational server configured as a resource for computational jobs of a scale too large for desktop computing, but not requiring the immense resources and administrative processes required to run on the major national HPC resources for which time can be requested (SCIAMA, Archer and similar).

It consists of a total of 36 computational `nodes’, each effectively a computer in its own right and used to run distributed tasks across larger numbers of computational cores than can be achieved with desktop computing. 32 of the 36 nodes consist of twin, quad-core processors and 64 GB of RAM; the remaining 4 nodes are twin hex-core processors with 24 GB of RAM.

This gives a total capacity of 296 computational cores, and just over 2.1 TB of RAM across the cluster. It stores data on a RAID-6 storage array, with 12 TB of local space for users. An approximation of its CPU computational capacity is between 1.4 and 3.0 TFLOPS (Trillion Floating Point Operations Per Second).

At average usage loads, Tor could be expected to perform around 1.8 million CPU hours of calculations during a year.