基于多核环境的并行性双向枚举连接  

Parallel Two-Way Enumeration Join Based on Multi-cores Environment

作　　者：陈永恒[1] 左祥麟[1] 

机构地区：[1]吉林大学计算机科学与技术学院,长春130012

出　　处：《吉林大学学报（理学版）》2014年第1期59-64,共6页Journal of Jilin University:Science Edition

基　　金：国家自然科学基金(批准号:60373099;60973040);教育部博士学科点专项科研基金(批准号:200801830021);吉林省重点科技攻关项目(批准号:20130206051GX)

摘　　要：基于多核处理器,结合自底向上和自顶向下两种算法,提出一种图遍历驱动的双向优化算法,该算法充分利用两种遍历算法的优点,并发挥多核环境的优势,实现了最优查询计划的高性能并行构建,解决了并行双向枚举连接问题.实验结果表明,该算法的性能优于已有算法,可明显提高数据库查询速度.Multi-j oin query optimization is one of the key problems in database searching.As multi-cores replace traditional single processor because of the advantages of their high performance and low energy consumption, researches on bottom-up dynamic query parallel algorithms have become popular.However,the advantages of top-down dynamic query algorithm can not be replaced by bottom-up dynamic query algorithms,such as cut algorithm.Based on bottom-up and top-down algorithms,this paper proposes a graph traversal two-way optimization algorithm on the multi-cores system.The proposed approach makes full use of the advantages of both top-down and bottom-up algorithms, gives full play to multi-cores environments, realizes high performance parallel construction of optimal query plans, thus resolves parallel two-way enumeration j oin problem. Experimental result shows that the proposed algorithm outperforms existing algorithms,and can effectively increase the speed of database query.

关 键 词：多核 查询优化 链接枚举 动态规划 

分 类 号：TP316[自动化与计算机技术—计算机软件与理论]