Comparison of the performance of Drools and Jena rule-based systems for event processing on the semantic web

Abstract

Event processing systems (EPS) analyze events in real-time to extract useful information that enhances value for the stakeholders. In order to extract relevant information from the stream of events, EPS's typically embed a rule-based system (RBS) that react to events as soon as they occur. The performance of RBS becomes even more important when the events processed are generated on the semantic web, where events and their metadata may originate from a variety of sources but at a very high rate, often in excess of 107 events per second. For such a system the most appropriate RBS's seem to be Drools and Jena both of which are open-source and are tailored for the Java programming language. In this paper we present the requirements and design of an EPS for the semantic web as required by the Air Force Research Lab, and analyze the performance of Drools and Jena for this EPS. We developed prototype systems using Drools and Jena, and stressed each system with streams of high speed events. Based on our study we conclude that while Drools is about 40% faster, Jena consumes about 15% less memory. We summarize the findings from our study and discuss scenarios when each of these RBS's might be more suitable.