Abstract: Management of primitive subscriptions and matching events with them are fundamental issues in a publish/subscribe system. This paper organized primitive subscriptions into a covering forest and then executes primitive subscription matching on the basis of multi-level predicate indexes. This approach was implemented in our content based publish/subscribe system OncePubSub. This paper also gave the descriptions of the experiments which were conducted to evaluate performance and overhead. Experimental results prove that our approach has efficient matching performance and good scalability.

Key words: Publish/subscribe, Primitive subscription management, Primitive subscription matching