数据流上Ad Hoc查询的自适应处理算法

Adaptive Processing Algorithm of Ad Hoc Query in Data Streams

对数据流上的Ad Hoc查询进行自适应处理,需要保证已有查询计划快速在线更新和迁移,但现有方法实现新旧查询计划的更新需要大量的滑动窗口状态转换。为此,提出一种AdHoe查询自适应处理算法。该算法基于数据流概要分布特性和自定义评分模型,快速计算出现有查询计划的最佳增量更新,以实现新到达的Ad Hoc查询处理,降低新旧查询计划切换时间。在数据流benchmark Linear Road提供的高速公路数据集上进行实验,结果表明,与MS、PT方法相比,该算法可较快完成新旧查询计划的切换。

The adaptive processing of non-stop coming Ad Hoc queries over data stream concerns much on the fast on-the-fly plan updating and migration. Existing methods implementing the shifting from old plan to new plan need amounts of work to move the state maintained in operators＇ sliding time window leading to large time delay. This paper proposes a highly efficient Ad Hoc query processing algorithm called AQU, which can compute the best incremental updating based on the summary view of data stream characteristics and a slef-defined model, handle the newly arriving Ad Hoc queries. This algorithm reduces the time shifting from old plan to new one and is also adaptive to non-top incoming queries. Experiments on the high way data generated by data stream benchmark Linear Road shows the efficiency of this algorithm.