Dynamic scheduling of computationally intensive applications on unreliable infrastructures
Maria Chtepen, Bart Dhoedt, Filip De Turck, Piet Demeester, Filip H. A. Claeys, Peter A. Vanrolleghem · 2006
simulation, software development Grid is a distributed computing paradigm that has considerably gained importance during the last several years. Typical for grids is that their resources are highly distributed and often belong to different organizations. Therefore, resource availability and stability form an important issue that should be taken into account when designing scheduling mechanisms for grid systems. Currently, most existing grids make use of static schedulers, which is not sufficient to deal with the dynamic nature of grid resources. This paper is touching on the dynamic scheduling concept, justifying its usefulness for computationally intensive applications. All results claimed in this paper are based upon simulations done in the DSiDE simulation environment. The DSiDE simulator is a discrete-event simulator for grids that differs from similar grid simulators, such as NSGrid, GridSim and SimGrid, on its ability to easily simulate the dynamics of grid jobs and resources, as well as on its short simulation times and portability. As a use-case for this study, a computationally intensive generic modelling and simulation tool, named Tornado, was chosen. The effect of dynamic resource availability on the execution of Tornado jobs will be shown.