Towards Energy and Power Efficient Data Stores

Abstract

Over the last decade, use of data stores have increased considerably. From scientific research in different domains to marketing and analytics, the need for more data is becoming essential to understanding customer behavior and to push the frontier of research. This means that there is a need for more data stores which puts a lot of pressure on data centers to make sure that their costs with respect to energy and power are kept in check. We extend research in energy efficient database systems by analyzing SQL queries to save memory energy thereby saving overall energy of server systems and by optimizing query planning under peak shaving constraints thereby maintaining performance and in some cases show better performance by choosing a better query plan for an SQL query. As memory becomes cheaper, use of it has become more prominent in computer systems. This increase in number of memory modules increases the ratio of energy consumption by memory to the overall energy consumption of a computer system. As Database Systems become more memory centric and put more pressure on the memory subsystem, managing energy consumption of main memory is becoming critical. Therefore, it is important to take advantage of all memory idle times and lower power states provided by newer memory architectures by placing memory in low power modes using application level cues. While there have been studies on CPU power consumption in Database Systems, only limited research has been done on the role of memory in Database Systems with respect to energy management. We propose Query Aware Memory Energy Management (QAMEM) where the Database System provides application level cues to the memory controller to switch to lower power states using query information and performance counters. Our results show that by using QAMEM on TPC-H workloads one can save 25% of total system energy in comparison to the state of the art memory energy management mechanisms. Peak shaving is a common practice in data centers when overall power consump- tion has to be managed. Data centers send triggers to servers to reduce their CPU frequencies using DVFS mechanisms, which in turn reduce power consumption of servers, thereby reducing the power consumption of the entire data center. This reduction in CPU frequency of database servers have an adverse effect on perfor- mance of SQL queries executed. As database servers do not modify their internal query plan parameters under peak shaving constraints, they continue executing sub- optimal query plans. While there have been studies on incorporating query power consumption in Database systems for energy efficiency, only limited research has been done in creating better query plans when CPU frequencies are throttled. In this paper, we show that there exist better query plans for TPC-H workloads and that we can improve the performance of query execution by an average of 10% under peak shaving constraints.