Performance Comparison of Models for Multiple Rule Firing

Abstract

The performance of production programs can be improved by firing multiple rules in a production cycle. Although considerable amount of research has been done on parallel processing of production programs, the problem of multiple rule firing has not been thoroughly investigated yet. In this paper, we begin by identifying the problems associated with multiple rule firing systems: the compatibility problem and the convergence problem and present three multiple rule firing models which address them. The rule dependence model (RDM) addresses the compatibility problem using inter-rule data dependence analysis. The single-context-multiple-rules (SCMR) model and the multiple-contexts-multiple-rules (MCMR) model address both the compatibility and the convergence problems. A production program executed under the SCMR and the MCMR models is guaranteed to reach a solution which is equivalent to the sequential execution. These three multiple rule firing models have been simulated on the RUBIC simulator, and the MCMR model, which has the highest performance, has been implemented on the Intel iPSC/2 hypercube. The simulation and implementation results are reported.