基于优先级的数据流数据库实时事务调度算法与实现  

A Scheduling Algorithm Based on Priority and its Implementation for Real-Time Transactions of Data Stream Database

作　　者：刘正涛[1] 毛宇光[1,2] 

机构地区：[1]南京航空航天大学信息科学与技术学院 [2]南京大学计算机软件新技术国家重点实验室,江苏南京210093

出　　处：《小型微型计算机系统》2006年第12期2212-2217,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(60473036)资助.

摘　　要：数据流数据库系统中的持续查询事务存在着依赖及算子共享的复杂关系.提出了基于综合事务空闲时间与价值的优先级调度算法.在算法中充分考虑了有依赖关系与共享算子的实时事务对事务优先级的影响.利用有关联的双链表给出了算法的实现,包括事务的接收策略与事务完成/夭折算法的实现.最后在GStream系统中,从累积实现价值率和差分截止期2个方面实验了算法的性能,实验表明LSV算法相对于RR算法、LSF算法、HVF算法都有很大的改进.There are two kinds of complex relations in the continuous query transaction in data stream database, of which one is that a transaction depends on the other transactions and the other is that many transactions can share one operator. A scheduling algorithm based on combining slack time and value is presented in this paper. The influence of real-time transaction complex relations is sufficiently considered. Furthermore ,The algorithm＇s implement including transaction acceptance policy and transaction completion/abortion policy ig given using three special double-linked lists. In the end,based on hit value ratio and differentiated guarantee ratio, an experiment which testes the LSV performance is completed in GStream. The experimental results show that the LSV algorithm can improve the performance compared to the classical RR.LSF and HVF algorithms.

关 键 词：数据流数据库 实时事务 调度策略 空闲时间 价值 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]