Scientific computing

High Performance Computing

The Institute Computing cluster has 1280 CPU cores for High Throughput jobs. There are currently three InfiniBand based, Lustre scratch file systems. The latest supporting 200Gb/s transfer speeds. It also provides 13 dedicated GPU nodes, comprising 3 CUDA nodes with a total of 12 GPUs and 10 ROCm nodes with a total of 80 GPUs. These are used for machine learning, classification and inference jobs. All are managed by a SLURM job scheduler. 

Research area specific databases, application servers, and additional network storage can also be provided. 

If these resources are required for your research, you will be given introductory training and access to the systems. We can provide training or advice on any aspect of computing, including the use of specific applications, computer programming, and the use of high-performance computing facilities.  

Data Storage

We provide tiered multi petabyte storage as well as the HPC scratch storage. Fast, offsite replicated, self-healing bulk storage is provided for general use, with offsite tape backups. A Write Once Read Many (WORM) archive system is also provided. Files copied to the archive are made immutable after a week, check summed and a copy is also written to tape. The archive is vendor agnostic with archived data being migrated and a new tape set being created during a hardware refresh. There are regular filesystem scans to compare checksums and restore from tape any files that may have become corrupted. 

Software & Database Development

The Software Team creates and supports web applications and databases for the Research Groups, Core Facilities and Operations departments at the Institute.