Logical Process Based Sequential Simulation Cloning

Discrete-event simulation is a widely used technique for the performance evaluation of systems. Recently, discrete-event simulation studies which explore alternative scenarios within a single simulation run have gained popularity. The simulation is branched at decision points, and the different branches are simulated one after another. In this paper, we show that concepts from parallel and distributed simulation can considerably speed up this type of discrete-event simulation study. We present a sequential branching mechanism, LPseq, which is based on the logical process paradigm and uses logical process cloning for efficient computation of branches. Our performance evaluation demonstrates that LPseq can achieve considerable speedup compared to the traditional branching approach.