条件过滤器的自适应排序调度策略  

Adaptive Ordering Scheduling of Selective-Filters

作　　者：刁静[1] 赵荣彩[1] 张铮[1] 邰铭[1] 刘勇[1] 

机构地区：[1]信息工程大学信息工程学院,河南郑州450002

出　　处：《信息工程大学学报》2011年第5期623-629,640,共8页Journal of Information Engineering University

基　　金：河南省杰出人才创新基金(0521000200)

摘　　要：自适应调度是以满足系统某一性能指标要求为目的,对数据查询中的算子进行动态调度的系统优化技术。针对选择查询中可交换顺序的条件过滤器的排序问题,设计了自适应过滤器排序算法,它以静态贪心算法为基础,通过滑动窗口机制模拟出过滤器之间的条件选择性特征,构建了一个条件概率关系矩阵,以指导过滤器排序的动态优化,得到一个当前条件下处理开销最低的过滤器排序,在较低的运行开销条件下,提升了系统的处理能力。Adaptive scheduling is a system optimization technique to dynamically schedule the operators in a query plan for the purpose of meeting certain performance specifications.An adaptive ordering algorithm based on the static greedy algorithm is designed to deal with the problem of sequencing the commutative filters in a select query.By modeling the characteristics of conditional selectivity between the filters with a sliding window,a relational matrix filled with conditional probabilities is built,and the sequence is optimized dynamically to realize a current fiter sequence with minimum processing cost.This technique can enhance system performance with reduced runtime overhead.

关 键 词：数据流管理 自适应调度 选择过滤器 滑动窗口 条件概率 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]