Serializability Improves Parallel Execution of Production System

This paper presents a new production system architecture that uses serializability as a correctness criterion to select a set of productions to be executed in parallel. The use of serializability eliminates global synchronization. This architecture takes advantage of modern associative memory devices to allow parallel production firing, concurrent matching, and overlap among matching, selection, and firing of productions. A comprehensive eventdriven simulator is used to evaluate the scaling properties of the new architecture and to compare it with a parallel architecture using global synchronization before every production firing. Our results indicate that the combination of serializability and associative memories can achieve substantial improvements in speed with a very modest increase in hardware cost. 1 Introduction Attempts to speed up Production Systems (PS) date back to 1979 when Forgy created the Rete network, a state saving algorithm to speed up the matching phase of PS [4]. ..