Chinese Journal of Computers   Full Text
Title	An Event-Driven Service-Oriented Computing Platform
Authors	LIU Jia-Hong WU Quan-Yuan
Address	(Institute of Network Technology & Information Security, School of Computer, National University of Defense Technology, Changsha410073)
Year	2008
Issue	No.4(588—599)
Abstract &   Background	Abstract SOA implementations based upon Request/Response model have deficiency in tight-coupled communication and poor cooperativity. Event-Driven Architecture(EDA) is one of the choices for implementing loose coupled communication with minimum delay and for responding rapidly and effectively to changing conditions, especially complex events derived from multiple events. By introducing Event-Driven paradigm into Service-Oriented computing platform, we can achieve loose couple communication and cooperativity. This paper presents a unified event-driven framework with design in delicacy for high-performance SEDA-based event concurrency and event algebra based event stream processing. To improve event stream processing performance, context semantic and corresponding detection algorithm are also presented. Experiment results prove that the framework is adapted to the dynamic large-scale distributed computing environment well and will be used widely due to its loose coupled communication, cooperativity and high-performance event stream processing and complex event processing properties. keywords service-oriented computing platform; event-driven architecture; event stream processing; event algebra; complex event processing background Many Request/Response based SOA implementations lack flexibility for communication and rapid response for changing conditions. Event-Driven Architecture(EDA) is the choice for implementing multistage processes that deliver information with minimum delay and for responding rapidly and effectively to changing conditions, especially complex events derived from multiple events. To get a high-performance event-driven computing platform for event concurrency and complex event processing, a unified framework for event processing, a concurrency management service and event correlation service including an expressive event language and a rapid detection mechanism on top should be designed with delicacy. Unfortunately the expressiveness of event language and scalability of framework is contradicting. Reactive systems, event notification systems and system monitoring applications have much appropriate solutions for separate aspects but lack a unified framework. Complex event can be application’s logic, like popular event notification systems’ suggest. Systems capable of processing complex events over stream in common use expressive declarative languages to model complex events and get them detected using Petri-Nets or automata based mechanism, some of which initiate auxiliary mechanisms to reduce event instances during event correlation, but their focus is to satisfy applications’ semantic, rather to get high-performance event stream processing. Data stream management systems are used for processing data of high arrival rate, but they lack capability to process time-series data in efficiency. This paper presents a unified framework to get high performance by a SEDA based model for processing event concurrency, a context policy based event algebra with well defined algebra properties for processing complex events over stream. The authors justify performance by experiment results.