Parallel implementation of the TRANSIMS micro-simulation

This paper describes the parallel implementation of the TRansportation ANal ysis and SIMulation System (TRANSIMS) traffic micro-simulation. The paralle lization method is domain decomposition, which means that each CPU of the p arallel computer is responsible for a different geographical area of the si mulated region. We describe how information between domains is exchanged, a nd how the transportation network graph is partitioned. An adaptive scheme is used to optimize load balancing. We then demonstrate how computing speeds of our parallel micro-simulations can be systematically predicted once the scenario and the computer architec ture are known. This makes it possible, e.g., to decide if a certain study is feasible with a certain computing budget, and how to invest that budget. The main ingredients of the prediction are knowledge about the parallel im plementation of the micro-simulation, knowledge about the characteristics o f the partitioning of the transportation network graph, and knowledge about the interaction of these quantities with the computer system. In particula r, we investigate the differences between switched and non-switched topolog ies, and the effects of 10 Mbit, 100 Mbit, and Gbit Ethernet. As an example, we show that with a common technology - 100 Mbit switched Et hernet one can run the 20 000-link EMME/2-network for Portland (Oregon) mor e than 20 times faster than real time on 16 coupled Pentium CPUs. (C) 2001 Elsevier Science B.V. All rights reserved.