Improving Production System Performance on Parallel Architectures by Creating Constrained Copies of Rules

Pasik, Alexander J.; Stolfo, Salvatore

Production systems have pessimistically been hypothesized to contain only minimal amounts of parallelism [Gupta 1984]. However, techniques are being investigated to extract more parallelism from existing systems. Among these methods, it is desirable to find those which balance the work being performed in parallel evenly among the rules, while at the same time decrease the amount of work which must be performed sequentially in each cycle. The technique of creating constrained copies of culprit rules accomplishes both of the above goals. Production systems are plagued by occasional rules which slow down the entire execution. These rules require much more processing than others and thus cause other processors to idle while the culprit rules continue to match. By creating the constrained copies and distributing them to their own processors, each performs less work while others are busy, yielding increased parallelism, improved load balancing, and less work overall per cycle.



More About This Work

Academic Units
Computer Science
Department of Computer Science, Columbia University
Columbia University Computer Science Technical Reports, CUCS-313-87
Published Here
December 7, 2011