Abstract

Distributed event-based systems are becoming increasingly popular as enabling technology for novel information-driven applications such as transport information monitoring, event-based supply chain management, ubiquitous sensor-rich applications and the like. The latter, however, pose some serious performance and scalability challenges. In particular, the inherent complexity and dynamics of such systems makes it difficult to predict their behavior under load and ensure that enough resources are available to meet the quality of service requirements. While many methods and techniques are available in the literature for workload characterization and performance modeling of conventional distributed systems, no methods have been proposed for distributed event-based systems. The latter have several inherent properties that make their modeling rather challenging. In this talk, I will discuss an approach to performance modeling and evaluation of distributed event-based systems. A practical method for workload characterization and performance analysis using basic analytical techniques will be presented. It will be shown how more detailed performance models can be built to provide more accurate performance prediction.