摘要： 基于单个数据流的滑动窗口聚集查询降载技术和数据流连接技术，提出滑动窗口模型下的数据流连接聚集查询降载策略，给出判断系统是否过载的负载方程和使过载系统恢复到轻载状态的降载算法，使降载后的查询结果同时拥有较小的相对误差和最大的元组输出率。实验结果表明，该降载策略具有较好的可行性和适应性。

关键词: 数据流, 滑动窗口, 聚集查询, 连接, 降载

Abstract: Based on load shedding techniques for sliding window aggregation queries over one single data stream and join techniques over pairs of data streams, this paper proposes a strategy of load shedding for sliding window join aggregation queries. By minimizing the relative error for query answers and maximizing the output rate of tuples, it introduces a load equation which determines whether the system is overloaded and a load shedding algorithm which gracefully recovers the overloaded system via load shedding. Experimental results show that the strategy has better performance in feasibility and adaptability.

Key words: data streams, sliding window, aggregation query, join, load shedding

中图分类号: 

• TP311.13