关系型数据库负载能耗预测模型.pdfVIP

Vo1. 44 No. 7 第 44 卷第7 期 计算机科学 2017 年 7 月 COMPUTER SCIENCE July 2017 一种基于 MBRC 值的关系型数据库负载能耗预测模型 杨德先孙华于炯国冰磊 (新疆大学软件学院 乌鲁木齐 830008) 摘 要 数据库负载的能耗解析与建模是构建节能的绿色数据库的基础。针对数据库负载的高能耗问题，将 SQL 语 句消耗的系统资源(CPU 和磁盘)映射为时间代价与功率代价，为数据库负载构建能耗预测模型。首先，根据负载的 系统资源消耗模式，计算负载的功耗代价;然后，根据负载资源消耗产生的时间代价，为负载构建动态能耗预测模型; 最后，利用 MBRC 值的设置对预测模型的准确度进行深入的研究。实验结果表明，所构建的预测模型能够对数据库 负载的能量消耗进行较准确的预测，预测模型的准确度研究有助于在不同的系统环境配置下提升动态能耗预测模型 的稳定性与精确度。 关键词 绿色计算，关系型数据库，绿色数据库，SQL 能耗建模，MBRC 中图法分类号 TP315 文献标识码 A 001 10. 11896/j. issn. 1002-137叉 2017.07.029 Relational Database Energy Prediction Model Based on MBRC YANG De-xian SUN Hua YU Jiong GUO Bing-lei (School of Software ,Xinjiang University ,Urumqi 830008 ,China) Abstract Power modeling and power profiling of database workload are the foundation of building energy-aware DBMS. To solve the high energy consumption problem of database workload ,this paper proposed a method of building energy prediction model for database workload by mapping resource consumption of the main components (CPU and Disk) to their power and time cost. Firstly ,power cost of workload is estimated according to their resource consumption pattern. Secondly ,energy cost of workload is estimated by combing the power cost and time cost derived from the re source consumption patter几 And finally , a deep research of the model prediction accuracy is conducted by using the MBRC settings. Experimental results show that ,the proposed energy prediction modeling method can make accurate es timation of workload energy cost ,and the research on