Reducing the Idle Time of Parallel Transportation Applications Technical Report Cse-96-004

Transportation engineers use traac equilibrium analysis to study traac and road systems; it is also used for planning purposes. This class of applications are generally computationally-intense and requires signiicant storage because of the large networks. Parallel processing provides the memory and computational power needed to solve these problems in a reasonable amount of time. The most computationally-intense component of the equilibrium analysis is the solution of a shortest path problem during each iteration. This paper focuses on reducing the parallel execution time of shortest path problems. In particular, we concentrate on reducing the processor idle time, which was found to be a signiicant component of the parallel execution time for shortest path algorithms used with traac networks. We provide four diierent methods for reducing this time, highlighting and quantifying the advantages and disadvantages of each. The results indicate that for a 9,701 node network solving for 32 sources the adaptive scheme reduces the execution time of the parallel shortest path by as much as 71% as compared to the other non-adaptive implementations. When using these schemes with the traac equilibrium application, a 63% reduction in execution time occurs by using the adaptive implementation.