Parallel execution on production systems

The paper proposes a scheme, dependency propagation algorithm (DPA), to analyze the parallelism of production systems, which have certain conflict resolution strategies. The scheme, based on a marked high-level Petri net, is to find out which rules can be executed in parallel without changing the semantics of OPS5-like production systems. This DPA uses explicit dependency, implicit dependency with degree 1 and dependency propagation to construct dependency matrix and parallel executable matrix. The high probability partitioning algorithm considering both load balance and maximal parallel execution uses a parallel executable matrix to map production rules into a multiprocessing system. The parallel execution model exploits not only parallel matching but also parallel firing. The initial results reveal that the performance of production systems is greatly improved.<<ETX>>

[1]  J. McDermott,et al.  Production system conflict resolution strategies , 1977, SGAR.

[2]  Salvatore J. Stolfo,et al.  Initial Performance of the DADO2 Prototype , 1987, Computer.

[3]  Salvatore J. Stolfo,et al.  DADO: A Parallel Processor for Expert Systems , 1984 .

[4]  Kemal Oflazer,et al.  Partitioning in parallel processing of production systems , 1987 .

[5]  Elizabeth Allen.  YAPS: A Production Rule System Meets Objects , 1983, AAAI.

[6]  Salvatore J. Stolfo,et al.  Towards the Parallel Execution of Rules in Production System Programs , 1985, ICPP.

[7]  Allen Newell,et al.  Parallel OPS5 on the Encore Multimax , 1988, ICPP.

[8]  Salvatore J. Stolfo,et al.  The DADO Production System Machine , 1986, J. Parallel Distributed Comput..

[9]  Anoop Gupta.  Implementing OPS5 production systems on DADO , 1984 .

[10]  David Elliot Shaw,et al.  Execution of OPS5 Production Systems on a Massively Parallel Machine , 1986, J. Parallel Distributed Comput..

[11]  Charles L. Forgy,et al.  OPS5 user's manual , 1981 .

[12]  Allen Newell,et al.  Initial Assessment of Architectures for Production Systems , 1984, AAAI.

[13]  James Lyle Peterson,et al.  Petri net theory and the modeling of systems , 1981 .

[14]  Charles L. Forgy,et al.  Rete: A Fast Algorithm for the Many Patterns/Many Objects Match Problem , 1982, Artif. Intell..

[15]  Charles Lanny Forgy,et al.  On the efficient implementation of production systems. , 1979 .

[16]  Elaine Kant,et al.  Programming expert systems in OPS5 , 1985 .

[17]  Daniel P. Miranker.  TREAT: a better match algorithm for AI production systems , 1987, AAAI 1987.

[18]  C. Forgy,et al.  PRODUCTION SYSTEM CONFLICT RESOLUTION STRATEGIES1 , 1978 .

[19]  John P. McDermott,et al.  R1: A Rule-Based Configurer of Computer Systems , 1982, Artif. Intell..

[20]  Jaideep Srivastava,et al.  A parallel architecture for large scale production systems , 1989, [Proceedings 1989] IEEE International Workshop on Tools for Artificial Intelligence.

[21]  Rudolph E. Seviora,et al.  A Multiprocessor Architecture for Production System Matching , 1987, AAAI.

[22]  Anoop Gupta,et al.  Static and Run-Time Characteristics of OPS5 Production Systems , 1989, J. Parallel Distributed Comput..

[23]  Toru Ishida,et al.  Optimizing Rules in Production System Programs , 1988, AAAI.

[24]  Salvatore J. Stolfo,et al.  DADO: A Tree-Structured Machine Architecture for Production Systems , 1982, AAAI.

[25]  Anoop Gupta,et al.  Comparison of the RETE and TREAT production matchers for soar (A summary) , 1988, AAAI 1988.

[26]  Salvatore J. Stolfo.  Five Parallel Algorithms for Production System Execution on the DADO Machine , 1984, AAAI.