Author(s): Peter David Mckellar Brady; James E Ball
Linked Author(s): James Ball
Keywords: SWMM; Genetic Algorithm; OpenMP; BOINC; Parallelization
Abstract: Genetic Algorithms have already been successfully applied to the optimization of catchment parameters within hydrologic systems. However, as with all genetic algorithms significant computational effort is required to compute the fitness of the individuals within the population. We present a two stage computational scale out methodology to take the singly threaded Storm Water Management Model (version 4. 4) .Firstly, we develop an Open MP based programme to run SWMM in parallel on a single, multi-core computer. With this wrapper we achieved linear speed up to three cores, which peaked at a 5. 5 times speed up on 12 core machine. Secondly we implement a distributed methodology to deploy the multi-core programme across a cluster of shared workstations using the BOINC middleware. With this middleware in place were able scavenge 20-30%of the computational resources of a 100+node cluster with over 1000 cores. These dual scale out methods allowed us to significantly reduce our computational runtimes and while simultaneously increasing both the parameter space to search and the size of the population of individuals.