Application Research On Event Stream Processing In The Wireless Content Delivery Platform

With the maturation of generic wireless content delivery, there is an urgent requirement of real-time processing of events generated in the delivery platform. Since event stream processing has to handle huge bulks of events occuring frequently in real time and it is so far impossible to put all event data into the limited memory, there then arise many fundamental research problems. In this dissertation, after the requirements are carefully analyzed, a brand-new event processing model is proposed, and the optimization and pattern matching algorithm in this model are then studied.The temporal logic based framework and the data stream processing model are combined in this dissertation for event stream processing. On the basis of formal definition of event, event stream and pattern matching for event stream, the event stream processing language is derived from continuous query processing. And, the language is very easy to use. The pattern matching is expressed by a given regular expression language. Since regular expressions have the same computing ability as finite-state automaton, the pattern matching can be done with DFA and hence the automaton construction algorithm is studied.To reduce the cost of accessing events and fill real-time requirements in event processing, an incremental approach is adopted to optimize the event stream processing. The single query processing evaluation algorithm of multiple input streams is given in this dissertation, with the cost analyzed. Since the multiple query optimization problems are NP-hard, greedy algorithm is used in processing multiple queries over multiple input streams.A prototype is implemented to prove the above models and algorithms in the real world, and a XML support extension is developed to process XML event streams, both of which are effectively applied in the wireless content delivery platform.Finally, a summary is made on the event stream processing. The dissertation points out that the event stream processing will have a wide application perspective, with the application scenario classified. Further research problems are also studied.