Transaction-Oriented Simulation In Ad Hoc Grids

IAbstractComputer Simulation is an important area of Computer Science that is used in manyother research areas like for instance engineering, military, biology and climateresearch. But the growing demand for more and more complex simulations can lead tolong runtimes even on modern computer systems. Performing complex ComputerSimulations in parallel, distributed across several processors or computing nodes withina network has proven to reduce the runtime of such complex simulations.Large-scale parallel computer systems are usually very expensive. Grid Computing is acost-effective way to perform resource intensive computing tasks because it allowsseveral organisations to share their computing resources. Besides more traditionalComputing Grids the concept of Ad Hoc Grids has emerged that offers a dynamic andtransient resource-sharing infrastructure, suitable for short-term collaborations and witha very small administrative overhead to allow even small organisations or individualusers to form Computing Grids. A Grid framework that fulfils the requirements of AdHoc Grids is ProActive.This paper analyses the possibilities of performing parallel transaction-orientedsimulations with a special focus on the space-parallel approach and discrete eventsimulation synchronisation algorithms that are suitable for transaction-orientedsimulation and the target environment of Ad Hoc Grids. To demonstrate the findings aJava-based parallel transaction-oriented simulator is implemented on the basis of thepromising Shock Resistant Time Warp synchronisation algorithm and using the Gridframework ProActive. The validation of this parallel simulator shows that the ShockResistant Time Warp algorithm can successfully reduce the number of rolled backTransaction moves but it also reveals circumstances in which the Shock Resistant TimeWarp algorithm can be outperformed by the normal Time Warp algorithm. Theconclusion of this paper suggests possible improvements to the Shock Resistant TimeWarp algorithm to avoid such problems.