Main /

University of Johannesburg Research Cluster

Welcome to the site dedicated to the Research Cluster for High Performance Computing at the University of Johannesburg, South Africa. On this site you will find news, information on the UJ Research Cluster, which services it provides, how it has been installed and more.

News

  • September 2009 4 new working nodes are being installed, and we are migrating the site to Scientific Linux 5
  • 15 April 2009 A draft MoU between the User Groups is available
  • 9-14 April 2009 UJ Research Cluster Training Workshop
  • 25 March 2009 12 new Worker Nodes have been ordered. When installed these will bring us to a respectable total of 192 CPU cores.
  • 20 March 2009 The UJ-OSG Compute Element passes validation and is registered on VORS. UJ RC is now officially a registered OSG site. UJ Research Cluster on the OSG World map
  • (permanent) SAGrid upcoming events in Training and Collaboration Meetings

Introduction

  • The UJRC, which started as an initiative of the new High Energy Physics group at UJ, supported by the OSG, has received the support of the University at large and of various Departments, and is a tool available to researchers in every field.
  • A cluster is a group of computers, all running the same operating system (Scientific Linux 4), which researchers can use for large computational tasks, controlling it from a single access.
  • This cluster is also part of the world-wide computing Grid; it participates both to the OpenScience Grid and to the South African National Compute Grid.
  • Currently the cluster offers primarily a platform for execution of batch jobs (Torque-OpenPBS), which is directly accessible to users with an account on one of the User Interface servers. Researchers belonging to a Virtual Organisation of the Grid can also use the User Interface to submit computing jobs to the Grid.
  • Computing programs which process independently separate subsets of data can very easily benefit from the parallel processing offered by the batch system and by the Grid; typical examples are the event-by-event data analysis and MonteCarlo simulations for High Energy Physics. Programs which operate on single, large calculations will instead need to be adapted to the parallel computing platform. Support for Message Passing APIs for parallelisation in the cluster is possible, and its implementation will be considered based on users' necessities.

Relevant Links: