A parallel simulation methodology for speedup and obtaining performance estimates with specific accuracy: Experiences of its application in studies of metropolitan area networks

Performance measures obtained by stochastic simulation are estimates, and o ne must consider the precision of the estimates before making any construct ive conclusions about the investigated systems. This paper applies an autom ated distributed simulation method, called Spectral Analysis in Parallel Ti me Streams, to speed up production of performance estimates, and for run-le ngth determination in the simulation of High-speed Metropolitan Area Networ ks (MANs). This method makes sequential simulators suitable for parallel ex ecution on multiprocessors and/or networked computers. At runtime it estima tes the information content of observations generated during simulation, ge nerates a point estimate and confidence interval, and directs the run to co ntinue until an estimate is obtained that achieves or exceeds our required level of precision. The application of this methodology for studying high-s peed MANs, the speedup, intermachine communication and "warm-up" overhead, and the run lengths needed to produce estimates with a specific level of pr ecision, are reported for each of the parameters investigated. Practical im plications are discussed.