A Scalable and Elastic Business Event Processing System

Abstract

The event processing system discussed in this paper updates target objects' state according to corresponding events. It is widely used in areas such as activity monitoring, business process automation, and business object state management. Existing event driven systems usually employ centralized single-server architecture, which is not scalable, elastic or robust enough to handle highly dynamic workload. In this paper, we propose a cloud-based event driven system that is highly scalable and elastic, and can tolerate even severe system failures. Our event driven system takes advantage of cloud technologies and adopts a completely distributed architecture. It therefore can scale to large number of servers in the cloud quickly with minimum administration overhead. Results from extensive evaluations show that our system's performance grows almost linearly with the number of servers in the system. Our system performs more than one order of magnitude better than a full-replication-based distributed event driven system of the same system scale.