Simulation of Resource Usage in Parallel Evolutionary Peptide Optimization using JavaSpaces Technology

Peptide Optimization is a highly complex problem and it takes very long time of computation. This optimization process uses many software applications in a cluster running GNU/Linux Operating System that perform special tasks. The application to organize the whole optimization process had been already developed, namely SEPP (System for Evolutionary Pareto Optimization of Peptides/Polymers). A single peptide optimization takes a lot of computation time to produce a certain number of individuals. However, it can be accelerated by increasing the degree of parallelism as well as the number of nodes (processors) in the cluster. In this master thesis, I build a model simulating the interplay of the programs so that the usage of each resource (processor) can be determined and also the approximated time needed for the overall optimization process. There are two Evolutionary Algorithms that could be used in the optimization, namely Generation-based and Steady-state Evolutionary Algorithm. The results of each Evolutionary Algorithm are shown based on the simulations. Moreover, the results are also compared by using different parameters (the degree of parallelism and the number of processors) in the simulation to give an overview of the advantages and the disadvantages of the algorithms in terms of computation time and resource usage. The model is built up using JavaSpaces Technology.