Overview

CSF4 / Advanced Linux Algorithms and Numerics (ALAN)

The CSF4 (also known as ALAN – Advanced Linux Algorithms and Numerics) is a brand-new computational platform at the University to complement the existing CSF3.

It provides new computing capacity, primarily replacing the now-retired large contribution to the original CSF by the School of MACE (in 2011), when the CSF was originally setup.

The system was initially funded by the Faculty of Science and Engineering (FSE) and by the departments of MACE and CEAS in that faculty. Many individual researchers have contributed funds subsequently to expand the system.

Resources

Updated Dec 2022: 245 x 40-core Intel Cascade Lake compute nodes provide 9800 cores in total.

Importantly, the CSF4 has been built on new brand-new infrastructure using different technology to the CSF3. In particular a new 1-Petabyte scratch filesystem built on GPFS, a new batch system running SLURM and completely new management software for better monitoring and scalability are in use. This will help to ensure reliability and scalability for future CSF contributions.

The compute resources in CSF4 are mostly dedicated to multi-node parallel jobs, although there is some resource for single-node multicore parallel jobs. A very small amount of resource is available for serial (1-core) jobs, mainly for pre/post processing. The preference for larger multi-node jobs reflects the workload of the contributors. This also helps to free up capacity on the CSF3 where the workload tends to be single-node multicore or serial high-throughput (job-array) jobs.

The CSF4 delivers the same benefits as the CSF3 and uses the same contribution model.

Please note that the existing CSF3 remains in active use and provides access to CPUs (mostly serial and multicore jobs), GPUs (Nvidia v100 and A100) and the HPC Pool.