Job-Scheduling of Distributed Simulation-Based Optimization with Support for Multi-Level Parallelism
2015 (English)In: Proceedings of the 56th Conference on Simulation and Modelling (SIMS 56): October, 7-9, 2015, Linköping University, Sweden, Linköping: Linköping University Electronic Press, 2015, 187-197 p.Conference paper (Refereed)
In many organizations, the utilization of available computer power is very low. If it could be harnessed for parallel simulation and optimization, valuable time could be saved. A framework monitoring available computer resources and running distributed simulations is proposed. Users build their models locally, and then let a job scheduler determine how the simulation work should be divided among remote computers providing simulation services. Typical applications include sensitivity analysis, co-simulation and design optimization. The latter is used to demonstrate the framework. Optimizations can be parallelized either across the algorithm or across the model. An algorithm for finding the optimal distribution of the different levels of parallelism is proposed. An initial implementation of the framework, using the Hopsan simulation tool, is presented. Three parallel optimization algorithms have been used to verify the method and a thorough examination of their parallel speed-up is included.
Place, publisher, year, edition, pages
Linköping: Linköping University Electronic Press, 2015. 187-197 p.
Linköping Electronic Conference Proceedings, ISSN 1650-3686 (print), 1650-3740 (online) ; 119
Job-scheduling, parallelism, distributed simulation, optimization
Electrical Engineering, Electronic Engineering, Information Engineering Fluid Mechanics and Acoustics
IdentifiersURN: urn:nbn:se:liu:diva-122752DOI: 10.3384/ecp15119187ISBN: 9789176859001OAI: oai:DiVA.org:liu-122752DiVA: diva2:872644
The 56th Conference on Simulation and Modelling (SIMS 56), “Modelling, Simulation and Optimization”, Linköping, Sweden, 7-9 October 2015