page 1  (20 pages)
2to next section

A Parallel Distributed Simulation of a
Large-Scale PCS Network: Keeping Secrets

Albert T. Montroy and Brian A. Malloy
Department of Computer Science
Clemson University
Clemson, SC 29634

Recently, research and development of complex PCS (personal communication service) networks has increased due to the rise in demand for mobile cellular communications. The efficiency of a PCS network is crucial in minimizing cost while maintaining quality service to mobile subscribers. Simulation is used extensively to facilitate the development of an efficient network. However, PCS simulation models are time consuming so that small-scale PCS networks are often simulated. Furthermore, small-scale PCS network simulations can produce incorrect results due to the effect of a call leaving the network at the boundary of the simulation; large-scale simulations can offset this boundary effect so that a network containing 1024 or more cells is required. Large-scale PCS networks have been previously simulated on a distributed network using the optimistic approach and on a MasPar using the conservative approach. However the optimistic approach can place large demands on the memory hierarchy and may not be an option on some systems. Also, in our experiments using the approach of [4], we were not able to induce appreciable speedup into the parallel simulation. In this paper we examine a conservative distributed simulation of large-scale PCS networks. The conservative approach that we propose permits large simulations using the PVM software to configure the network into a parallel machine. Using a unique approach to exploit lookahead, we are able to induce speedups comparable to those produced in [1]. Also, using our distributed approach, we have been able to simulate PCS networks containing 100,000 cells.

keywords: modeling, time-driven, PCS, methodology, conservative parallel distributed simulation.