QoS Enhancement for PDES Grid Based on Time Series Prediction

The combination of parallel/distributed discrete event simulation (PDES) and Grid technology is a new trend in simulation. QoS is important yet difficult in this process. With the special features of Gird architecture and PDES input, such as periodical and predictable inputs, we could enhance QoS with a PDES-specific prediction. A Grid-based framework is presented that is designed to help predict the performance of PDES. Based on this framework, a prediction algorithm using time series theory is presented in the context of large scale Grid simulations. Experiments are executed in the context of GridSim, which shows methods discussed before to help to improve QoS level.