共享连接结果的连续查询处理  

Continuous queries processing by sharing intermediate join results

作　　者：钱江波[1] 徐宏炳[1] 董逸生[1] 刘学军[1] 王永利[1] 杨雪梅[1] 

机构地区：[1]东南大学计算机科学与工程学院

出　　处：《东南大学学报（自然科学版）》2007年第1期5-8,共4页Journal of Southeast University：Natural Science Edition

基　　金：江苏省高技术研究资助项目(BG2004034);江苏省2004年度研究生创新计划资助项目(xm04-36)

摘　　要：深入研究了适合数据流连续查询处理的Shared PushDown,PullUp和Filtered PullUp三种可共享连接结果的策略.通过理论分析和实验证明,在数据流的连续查询处理中,PullUp策略性能较低,而Filtered PullUp和Shared PushDown策略各占优势.Filtered PullUp处理简单,消耗内存相对较少.而Shared PushDown增加内存使用,但在注册查询的选择因子覆盖较少时有一定优势.考虑到处理的方便,一般情况下Filtered PullUp是适合数据流连续查询处理的最佳方案.The traditional heuristic algorithm of pushing selection predicates below joins is possibly less efficient for continuous queries, because early selection destroys the ability to share subsequent high cost join processing. Three alternative selection placement strategies, which can share join results, are evaluated. Theoretics and experimental results show that PullUp strategy （selections are pulled above joins） has poor efficiency, and Filtered PullUp strategy （ data stream tuples are filtered by the union of the selection predicates, then to be executed by PullUp strategy ） is better than Shared PushDown strategy （ selections are pushed below joins） because the former is simple and uses less memory.

关 键 词：数据流 连续查询 窗口连接 选择操作 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]