Efficient large-scale process-oriented parallel simulations

Process oriented views are widely recognized as very useful for modeling, but difficult to implement efficiently in a simulation system, as compared to event oriented views. In particular, the complexity and run-time overheads of the implementation have prevented the widespread use of process oriented views in optimistic parallel simulations. Here, we review the conventional approaches to implementing process-oriented views, and outline some of the sources of problems in those approaches. We also identify an approach that we call stack reconstruction, which is most suited for portably and efficiently supporting optimistic process-oriented views. Benchmark simulations using our preliminary implementation, which is incorporated in the TED modeling and simulation system, confirms the low overheads of this approach, and demonstrates its capability to simulate over one million processes in a process-oriented model.