Evaluation of rule processing strategies in expert databases
A. Segev, Jun Zhao · 2002
Rule processing strategies in expert database systems which involve rules conditional on join results of base relations are studied. In particular, those rules that require very fast response time in their evaluation are considered. It is proposed to materialize the results of firing a rule in a relation, the rule relation. Performance evaluation of several strategies shows that under the clustered B-trees, strategies using pattern relations perform better than those without pattern relations. The strategy with skinny pattern relations performs poorly in comparison to that with bulky pattern relations. The selective bulky pattern strategy performs better than the bulky pattern strategy. The selective pattern strategy outperforms other strategies in terms of expected total cost. However, it always uses more storage space than the direct materialization.>