The HPC facility, which can be termed a “Super Computer”, but rather than being a single machine, is a high-performance computing cluster comprised of many standard components (e.g. high specification computer servers) networked together by high performance interconnect, and with the necessary software to make it capable of performing extensive calculations, involving hundreds of computing cores, and several terabytes (TB) of computer memory.
The hardware making up the main cluster includes:
- 8 compute nodes, each with 2x 14 core processors (Intel Xeon E5-2680 2.4GHz), 128GB RAM, and 600GB local disk.
- 2 compute nodes, each with 2x 8 core processors (Intel Xeon E5-2667 3.2GHz ), 256GB RAM, 600GB local disk, and 2x Tesla M60 GPGPU accelerators.
- 4 shelves of Panasas PAS12 object-based, parallel storage, offering 240TB raw, 180TB formatted disk space, available from login nodes and compute nodes.
- 1Gb Ethernet network used for general end-user communication.
- 56Gb bandwidth InfiniBand Quad data-rate (QDR) network used for MPI (“message-passing-interface”) data, and storage traffic.
This cluster, therefore, provides a total of 276 CPU compute cores and is capable of running at over 11 teraFLOPS - that is 11 trillion or 11 million million calculations per second).