Nowadays, clusters of personal computers are more popular than ever. Compared to shared-memory multiprocessor machines, clusters achieve nearly the same performance rate but at a much lower cost.
DYANA is a program that computes three-dimensional structures of molecules on shared-memory multiprocessor systems. A first version of DYANA, which runs in parallel by means of fork-calls, was developed at the Institute of Molecular Biology and Biophysics at ETH Zurich. It is our opinion that enhancing DYANA to run efficiently on a cluster of PCs will contribute to its distribution: research groups with difficult access to powerful shared-memory systems will be able to run DYANA with low cost/high performance resources.
In this project, we have implemented a new version of DYANA which runs on a cluster of PCs using MPI. We have started investigating the algorithms behind DYANA as well as the critical aspects of the fork version. By means of the data and information gathered during this investigation phase, we have developed the new version of Dyana introducing new paralleization tools such as MPI. Finally, we have compared the performance results obtained with both the original and the new version of DYANA.