Exploiting Sharing Opportunities for Real-time Complex Event Analytics

Complex event analytics systems continuously evaluate massive workloads of pattern queries on high volume event streams to detect and extract complex events of interest to the application. Such time-critical stream-based applications range from real-time fraud detection to personalized health monitoring. Achieving near real-time system responsiveness when processing these workloads composed of complex event pattern queries is their main challenge. In this article, we first review several unique optimization opportunities that we have identified for complex event analytics. We then introduce a family of optimization strategies that consider event correlation over time to maximally leverage sharing opportunities in event pattern detection and aggregation. Lastly, we describe the event-stream transaction model we designed to ensure high performance shared pattern processing on modern multi-core architectures.