Cardiff Cluster Computing Forum (CCCF)
The CCCF is focused on the provision of best working practice around clusters at Cardiff University, and their optimum utilisation. The Forum will provide an expanded role over that currently filled by the System Administration group, and will be coordinated and supported by ARCCA staff. The group will focus on a number key operational activities:
- The optimum deployment of technology in support of research computing within the University, with a focus on the coherent provision of resources by ARCCA and those activities supported by the Schools.
- Information dissemination and discussion around all aspects of cluster technology, best-practice cluster management and optimum utilisation.
- The provision of a common software interface for clusters and their users (possible resonance with proposed Campus Grid activities):
- A common software stack across all clusters (currently Condor, PBS, and SGE)
- A standard accounting system across all clusters (currently Ganglia and Moab)
- A common licensing regime (needed for Matlab, Abaqus, and other applications)
- Access to a full suite of compilers and optimisers from all major vendors
- Enhanced environments for Code Development and Performance
- Enhancing the current development environments though the provision of parallel debuggers and code optimisers.
- Improving the performance of users codes on various clusters
- Understanding the key parallel programming paradigms (MPI, OpenMP etc)
- Algorithms and impact on parallel performance.
- The provision of appropriate Training and Outreach
- Provision of training in the use of all the above to SysAdmins and Users
- Increasing the usage of our visualization assets
- Technology Updates
- Regular technology updates from Cluster Integrators and Tier-1 suppliers
- ARCCA staff work with current sys. admin. staff and line managers in identifying and progressing opportunities for career development.
Some of the aims of the CCCF include:
- Fostering collaboration and communication between ARC SysAdmins
- Attending and presenting to a series of local ARC workshops
- Agreeing methods and delivering standard reporting and monitoring information on Cardiff's ARC facilities
- Developing and documenting best practices for the management and administration of ARC facilities.
The group will report every 6 months to the ARC Working Group.
If you are an administrator of any ARC or mid-range compute systems at the University, or a researcher utilising these systems, you are invited to join the group and related email list. Please contact us for more information.