Christopher D. Carothers, Ph.D.

Associate Professor
Computer Science
Rensselaer Polytechnic Institute


Time Warp Discrete-Event Simulations on Terascale Supercomputers

Discrete-event simulation techniques are used to model a host of complex systems ranging from large-scale networks (i.e., a significant fraction of the Internet) to next generation petascale supercomputer systems. However, because of the computational demand of these models, they must be run on state-of-the-art supercomputer systems. Time Warp is a synchronization protocol that allows a parallel discrete-event simulation to speculatively process event computations, but handles event ordering errors when they occur. The aim of our research is to investigate the performance and overall scalability of a Time Warp discrete-event simulator on both Blue Gene/L and Blue Gene/P supercomputers. On Blue Gene/L, we find that strong scaling out to 16,384 processors is possible. In terms of event-rate, we observed 2.47 billion events per second for a PCS telephone network model when executed on 32,768 processors. At the time, this is the first multi-billion event rate achieved for any Time Warp model. On Blue Gene/P, we are able to push Time Warp performance even further where we obtained 12.5 billion events per second using 65,536 processors for the PHOLD benchmark model.


Dr. Carothers is an Associate Professor in the Computer Science Department at Rensselaer Polytechnic Institute. He received the Ph.D., M.S., and B.S. from Georgia Institute of Technology in 1997, 1996, and 1991, respectively. Prior to joining RPI, he was a research scientist at the Georgia Institute of Technology. Currently, Dr. Carothers is an Associate Editor for the ACM Transactions on Computer Modeling and Simulations as well as SIMULATION: Transactions of the Society for Modeling and Simulation International. His research interests include high-performance parallel and distributed computing, simulation, networking and computer architecture.

updated: 2008-10-15