Approximate Performance Analysis of Job Scheduling on Computational Grids

Our former study had investigated the modeling and performance evaluation of QoS-aware job scheduling on computational grids using the stochastic high-level Petri net (SHLPN). This paper proposes an approximate performance analysis technique, which is based on the decomposition and refinement of the SHLPN model as well as iteration among submodels, to reduce the complexity of the model and cope with the state-space explosion problem. Numerical results of performance analysis show that this approximate analysis technique is effective for accuracy of the numerical results as well as significantly reduces the state complexity of the model.