Cluster computing

Cluster Computing at CSE

Several clusters of hosts are available for special purposes.

The computer support group (CSG) manages several of these clusters and each cluster computer works much like a login server.

Access to cluster computers is given by CSG at the request of the academic supervisor. If you want to use a cluster computer please ask the supervising academic to email CSG for you.

They originate from different funding purposes related to the research groups that use them.

  • Vina Cluster — Supervisor Manuel Chakravarty — 20 computers
  • Rattle Cluster, and Sigmod Cluster — Supervisor Xuemin Lin — 41 computers

General Notes

Research cluster computers managed by CSG are generally used by one person at a time, though CSG don't enforce this. Making sure two people aren't using the same cluster machine is something we leave up to the research group to schedule.

Similarly, we set up cluster computers to not restrict the memory use on the processes running there. The person using the computer will need to manage the memory use of their processes to avoid their computing task crashing.

If you want to start jobs, then disconnect from the remote session and reconnect later there are programs to help you do this. See our notes on screen and VNC.

Lab Computer Cluster

CSE runs a "Condor" system to schedule jobs on the CSE lab computers. More information at: CSE Condor FAQ page

Faculty Clusters

Follow this link for information on the Faculty of Engineering's Clusters. (Wiki documentation is here.)

Excerpt:

The Faculty of Engineering administers a medium-sized computational cluster for use by researchers in Engineering.

The cluster is based on blade servers (HP BL685c G7 blade) and currently consists of:

  • 2,944 compute core (across 56 compute servers),
  • 5.8 TB Memory,
  • 100 TB (usable) shared disk storage. …

The cluster is fully interconnected with dual 10 Gb/sec data paths.

The cluster nodes are Linux-based, running the standard Rocks clustering platform. There are UNSW site licences available for a number of software packages which may be used on the cluster. These currently include Matlab and Ansys. The cluster is also a viable platform for researchers writing their own software.

The cluster is hosted at the UNSW Data Centre at R1 (Randwick Campus). The location has fully redundant networking, power and air-conditioning; and is constantly monitored by UNSW systems and staff.

Last edited by robertd 23/08/2016

Tags for this page:

cluster, computing