一种数据流上基于截止期的多查询过载预测模型

为保证数据流上查询处理的实时性要求,定义了截止期作为连续查询的实时性约束,提出了一种数据流上基于截止期的多查询过载预测模型.模型预测的过载点给出了能够保证数据流系统内所有查询满足各自截止期的临界情况.在多查询环境下,通过找到某一查询作为截止期瓶颈,使得该查询处理结束后剩余查询的处理能力正好大于系统的输入流速,从而计算出过载点.仿真实验结果证明,该模型预测出的过载点能够保证所有查询满足各自的截止期,并且预测算法具有良好的准确性和计算复杂性.

A Multi-query Overload Prediction Model for Deadline-Aware Data Stream Processing

To meet the real-time requirements for query processing over data streams, the deadline is defined as the real-time constraint of continuous queries. A multi-query overload prediction model is thus developed for deadline-aware data stream processing. The overload point predicted by the model indicates the critical conditions under which each and all of the queries can meet the requirements of deadline individually. In the circumstance of multiple queries, the overload point is estimated by finding a query as deadline bottleneck. After processing the deadline bottleneck, the processing capability of the residual queries is just higher than the input stream rate, thus working out the overload point. The simulation results showed that the overload point predicted by the model enables each and all of the queries to meet individually the deadline they required, and that the prediction algorithm is accurate and efficient.