Grid-Enabled Simulation-Optimization Framework for Environmental Characterization

A parallel evolutionary strategy based simulation–optimization approach for solving groundwater source identification problems

“…The simulation-optimization framework is based on the LAarge Scale Simulation-Optimization framework (LASSO); a description of the framework is available in Mirghani et al [22]. LASSO consists of a centralized optimization application that utilizes a masterworker task distribution strategy.…”

“…The fine grained parallelism is within the simulation model. The two coarse grained levels are due to independent model evaluations required by the optimization application on parallel or distributed grid resources [21,22]. The optimization application acts as the master process in the master-worker task distribution strategy.…”

“…Aggregating individuals in this manner reduces communications overhead. Worker processes running on distributed computing resources ''grid", having established a TCP-IP socket connection with the master, signal their readiness and draw tasks from the task pool [22]. The worker process transfers the remote task to the MPI's zeroth processor.…”

“…Several approaches, such as surrogate modeling [21,26] and code parallelization [22], are available to improve the simulation model efficiency by reducing the forward model solution time. In the research presented in this paper, the parallelization approach is investigated to improve the execution time of the simulation model.…”