The purpose of a centralised High Performance Computing (HPC) resource is to provide a reliable, scalable and economic computing facility to UCT’s researchers. Given the cost of electricity both for powering and cooling CPUs it is imperative that UCT provides an affordable and environmentally-friendly means of granting researchers access to computing resources. Currently HPC operations is catered for in the budget and maintenance for the standard procurement and life cycle of ICTS equipment, however the long term aim is for the cluster hardware to be sustained through research funding.
The HPC clusters are housed in ICTS’s data centres in order to make use of existing support infrastructures such as UPS, generators, cooling, fire suppression and secure access. This ensures that UCT’s investment in HPC is well spent and that the cluster can continue to grow according to researchers’ needs. Permanent staff members have also been given extensive training in supporting the HPC infrastructure and researchers’ computing requirements. This training and time is seen as an investment by the university in providing long term HPC support.
Researchers that make use of the HPC resources are bound by the Acceptable Use Policy and are required to reference the HPC resource in their citations. There is currently no service level agreement (SLA) for HPC resources and support is on a best effort basis. Should a department or a research group need to use specific infrastructure on the HPC cluster that does not fall within the generic offering then they need to include it in their budget and engage with ICTS regarding the provisioning thereof.