A comparative study of state saving mechanisms for time warp synchronized parallel discrete event simulation

The state saving mechanism constitutes an essential part of any system in which erroneous or undesired executions can be undone by rolling back the system state. One such application is optimistically synchronized parallel discrete event simulation (PDES) systems based on the time warp synchronization mechanism. In this type of system the state saving and restoration mechanism is essential to the performance. Consequently, several mechanisms which could reduce the state saving overhead in time warp based PDES have been proposed. In this study we investigate the performance of several such mechanisms, including both transparent sparse and incremental state saving methods, in the simulation of large, realistic cellular communication simulation models. We also investigate the applicability of two analytical methods proposed to guide the choice of state saving mechanisms. Our empirical results indicate that the best choice of method depends on the characteristics of the simulation model. Furthermore, the cost of transparent state saving methods is found to be low. We also show the need to further investigate analytical methods to better understand how to make the selection of state saving mechanisms.