Peak Power Modeling For Query Plan

The energy costs of large-scale data centers have rapidly risen and quickly exceed the cost of the hardware itself.Energy consumption in data centers has become a growing concern.Energy management has become an important measurement and design standard for modern data centers.In a typical data center,most of the computing resources are dedicated to the database server.The DBMS is the largest energy-consuming component of the software deployed in database server.The research of energy efficiency database meets the growing demand for database management.The energy efficiency research of database system is an important part of solving data center energy saving problems.The core operation of database system query processing is joining.Therefore,this paper aims at studying the join algorithms based on asynchronous I/O,proposing the peak power prediction model and the trade-off mechanism between peak power and performance of join algorithms.The main contributions of this article are as follows:1.Using asynchronous I/O methods to achieve join algorithms in the database system,proposing an algorithm for the maximum CPU_Bound of the join algorithms in the peak power generation phase.2.By studying the relationship between CPU_Bound and peak power,we establish a peak power prediction model of join algorithms.3.Based on peak power and performance analysis of join algorithms,we propose a trade-off mechanism between peak power and performance of join algorithms.A lot of experimental verification is carried out for the proposed scheme in this paper.Experimental results show that the proposed peak power prediction model has a certain accuracy.The average relative error is within 6.64% at different CPU frequencies.At the same time,the relationship between the performance and the peak power of the join algorithms is studied so as to control the peak power of the system.