Fast Simulations of Large-Scale Highly Congested Systems

Focusing on resource cycles, the authors developed a semiconductor wafer factory (fab) simulation that executed more than an order of magnitude faster than a job-tracing simulation previously in use. The authors summarize the methodologies used and conclude that the differences in execution speeds are due to the fundamental differences in using an event graph paradigm to model the discrete event system dynamics instead of the more popular process flow paradigm that is used by almost all commercial simulation packages. However, the execution speed of a resource-driven model is insensitive to system congestion, whereas a job-driven model slows dramatically (or halts) as the system becomes heavily loaded. The authors conclude that a resource-driven approach using event scheduling logic offers the best approach to modeling very large-scale highly congested systems such as those found in communication, transportation, and unit-manufacturing operations.