Parallelisation of a molecular replacement program

The aim of this project is to usefully parallelise (using MPI[1]) the molecular replacement program Qs[2].

Initial tests with a simple-minded (fine-grained) parallelisation showed a useful scale-up of the program for a case of a 1Gbps cluster (the tests were performed on the University of Leeds Faculty of Biological Sciences Linux cluster). Here are the details from testing Qs on the FBS Linux cluster.

The Qs tests on MBG cluster (with the same fine-grained paralellisation) failed to show a useful scale-up of the program for all types of problems mainly due to the slow (100Mbps) network link between the nodes.

Future work will include :

• Implementation of other parallel algorithms
• Benchmarking of these implementations using
  • Different size problems
  • Different number of nodes
• Application to real-life problems