Summary

Resources per team: HPC Cluster: 100k core hours and 1k GPU hours

Resource Access:  SSH access to a login node or a web user portal 

Resource management: A Slurm batch queue manager. Users will have access to a shared network file system.

Software management: Software module environment, singularity containers or user compiled code. 

Documentation: https://docs.hpc.cam.ac.uk/hpc/

Support: support@hpc.cam.ac.uk please put [SDC3] in the subject of the email

Resource location: University of Cambridge, UK

Additional Information: On request we can provide limited resources running on a private OpenStack cloud, in the Azimuth Cloud Portal environment. This environment can support user-deployed and user-managed applications clusters such as Slurm or Kubernetes, but for user-managed clusters support is more limited in scope

Technical specifications

Overview

The UKSRC resource at the University of Cambridge comprises a multi node HPC/GPU cluster running a Slurm batch scheduler. An OpenStack-hosted Platform-as-a-Service Azimuth applications portal is also available on request. 

Technical specifications

https://docs.hpc.cam.ac.uk/hpc/user-guide/cclake.html

https://docs.hpc.cam.ac.uk/hpc/user-guide/icelake.html

https://docs.hpc.cam.ac.uk/hpc/user-guide/a100.html

Per user resource

The cluster operates a fair-share algorithm between all the users within a project.

Software installed

A wide range of software packages described in the documentation (https://docs.hpc.cam.ac.uk/hpc/index.html) is available via modules.

Singularity/Apptainer containers are also supported.

Users can compile their own code in the project spaces.

Volume of resource

Each SDC3 team will be allocated 100k core hours and 1k GPU hours and 20TB of storage.

GPUs if any

320 x Nvidia A100 GPUs

User access

Account Setup

To setup an account please complete the form https://www.hpc.cam.ac.uk/external-application (unless you are a member of the University of Cambridge, in which case use the usual internal form).

Logging in

SSH access via login.hpc.cam.ac.uk or web access via login-web.hpc.cam.ac.uk.

How to run a workflow

Slurm batch script submission, using your own code, local software modules or containers.

Software management

Users can compile their own codes, make use of module packages or their own singularity containers.

Containerisation

Singularity or docker container images can be run using Singularity/Apptainer.

Documentation

https://docs.hpc.cam.ac.uk/hpc/user-guide/quickstart.html

Cloud Access 

On request we can provide limited resources on our private OpenStack cloud through an Azimuth Cloud Portal environment.


Resource management

Resources on the HPC clusters will be allocated via Slurm projects. 

On request we can provide limited resources on our Openstack cloud through an Azimuth Cloud Portal environment.

Support

Support will be provided via the HPC service desk - email support@hpc.cam.ac.uk.

Credits and acknowledgements 

The UK SKA Regional Centre (UKSRC) is funded by:

https://www.iris.ac.uk/portfolio/stfc-cloud/