Click here for full text:
ReCEPtor: Sensing Complex Events in Data Streams for Service-Oriented Architectures
Wei, Mingzhu; Ari, Ismail; Li, Jun; Dekhil, Mohamed
Keyword(s): data stream, complex event processing, SOA, query plan, EPL
Abstract: All mission-critical applications read or generate raw data streams and require real-time processing of these streams to collect statistics, control flow and detect abnormal patterns. A trend which has gained strong momentum in different industry sectors is the use of Complex Event Processing (CEP) over streams for all critical business processes, thus pushing its span beyond military and financial applications. A parallel trend has been the re-architecting of existing business processes with Service Oriented Architecture (SOA) principals to provide integration and interoperability. This requires use of middleware that incorporates web services, Business Process Management (BPM) systems or an Enterprise Service Bus (ESB), and in some cases Business Rule Engines (BRE). There is a gap between the new generation of business processes which desperately need CEP and the proposed CEP engines that were not built with SOA in mind. These engines either don't support the flexible, dynamic and distributed business applications deployed in SOA or they try to merge the middleware with the CEP engine as one big proprietary package. This paper describes the design and implementation of our CEP engine called ReCEPtor, which can sense complex events in data streams in real-time. Our modular architecture describes how to integrate ReCEPtor with different business applications in different platforms either directly or by using an off-the-shelf BPM system and a platform adapter. We address issues related to CEP system flexibility, interoperability, scalability and performance in this paper. We also discuss novel concepts such as sequence operators, query plan adaptation, and scheduling for progressive flow. We can already process ~15 requests/second through the 3- Tiers (web services, orchestration, and database) of our prototype end-to-end and without tuning and we are currently working on achieving higher speeds.
Back to Index