With the development of the monitoring technologies such as RFID, complex event processing over event streams has raised a lot of attention. Most existing complex event process-ing techniques only provide best-effort services to reduce the average response delay,which can-not guarantee that more detected composite events are detected under the deterministic response delay requirements. According to the demands of deterministic QoS for monitoring applications, this paper discusses the deadline-sensitive complex event processing resource allocation optimiza-tion over common Poisson streams. The event arrival and complex event processing are analyzed and modeled from the view of system service. Furthermore, this proposes the complex event deadline satisfying ratio model, multi-event stream processing disorder feedback correction model and furthermore resource optimized allocation model. By allocating the processing resource rea-sonably, the framework can produce more correct results under the real-time constraint, and the tradeoff of the real-time and correctness feature is considered. The experiments have shown the validity of the proposed estimation model and allocation strategy. The proposed models are believed to be valuable in theory and practice for improving the fixed real time complex event processing framework.%随着无线射频识别等监控技术的发展,事件流上的复杂事件处理技术得到了广泛的关注.现有的事件流处理技术主要提供尽最大努力(Best-Effort)式的服务来减少平均响应时间,这种方式并不能在确定的时间延迟需求下输出更多的结果.针对监控应用中的确定性服务质量需求,讨论了常见的泊松监控流上的截止期敏感的复杂事件处理最优化资源分配问题.从系统服务角度对事件的到达和复杂事件处理进行了理论分析和建模,提出了复合事件的截止期满足率模型和多事件流处理乱序反馈修正模型,进而给出最优化资源分配模型.通过合理的分配处理资源,保证了在实时限制下产生更多的正确结果,兼顾了复杂事件处理的实时性和正确性.实验结果表明,提出的相关的理论估算模型和资源分配方案是正确和有效的,对于进一步完善复杂应用下的同实时限制复杂事件处理框架具有重要的启发意义和实用价值.
展开▼
