Complex event analytics: online aggregation of stream sequence patterns
Complex event analytics: online aggregation of stream sequence patterns
复制标题
复杂事件分析：流序列模式的在线聚合
DOI：
10.1145/2588555.2593684
复制
发表时间：
2014
期刊：
影响因子：
--
通讯作者：
Elke A. Rundensteiner
中科院分区：
文献类型：
--
作者：
Yingmei Qi;Lei Cao;M. Ray;Elke A. Rundensteiner
复杂事件处理 (CEP) 是时间关键型决策应用程序中高性能分析的首选技术。然而，尽管已经开发出针对连续事件流进行复杂模式检测的有效技术，但此类模式的可扩展在线聚合问题却被忽视了。相反，聚合通常用作 CEP 模式检测后的后处理步骤，导致解决方案极其无效。在本文中，我们证明了 CEP 聚合可以被推入序列构建过程中。基于这一见解，我们的 A-Seq 策略成功在线聚合序列模式，而无需构建序列匹配。这将 CEP 聚合问题的复杂性从多项式降低为线性。我们进一步扩展 A-Seq 策略以支持并发 CEP 聚合查询的共享处理。与最先进的解决方案相比，A-Seq 解决方案在各种测试场景中实现了超过四个数量级的性能提升。
Complex Event Processing (CEP) is a technology of choice for high performance analytics in time-critical decision-making applications. Yet while effective technologies for complex pattern detection on continuous event streams have been developed, the problem of scalable online aggregation of such patterns has been overlooked. Instead, aggregation is typically applied as a post processing step after CEP pattern detection, leading to an extremely ineffective solution. In this paper, we demonstrate that CEP aggregation can be pushed into the sequence construction process. Based on this insight our A-Seq strategy successfully aggregates sequence pattern online without ever constructing sequence matches. This drives down the complexity of the CEP aggregation problem from polynomial to linear. We further extend our A-Seq strategy to support the shared processing of concurrent CEP aggregation queries. The A-Seq solution is shown to achieve over four orders of magnitude performance improvement for a wide range of tested scenarios compared to the state-of-the-art solution.