III: Small: Collaborative Research: Making Databases Green - An Energy-Aware DBMS Approach

通过技术创新维持一个可持续发展的社会一直是计算系统设计的主要挑战。在这个项目中，我们关注的是一个重要类型的计算机应用的能源效率，数据库管理系统（DBMS），它经常消耗很大一部分的计算资源和能源在现代数据中心。这个项目的目标是设计和实现一个DBMS，它能够在降低查询处理性能的同时显著节约能源。该项目使用以下方法实现其目标：(1)动态地利用DBMS中的能源性能权衡以及硬件系统的低功耗模式，以提高能源效率并保证性能；(2)将节能DBMS设计表述为反馈控制问题，并采用适当的形式化控制技术，通过理论分析和保证达到预期的性能和能效；(3)应用先进的多阶段优化方法解决复杂的能量感知储能管理问题；(4)协调DBMS不同层中的各种控制和优化循环，以最大限度地节约能源和全局系统稳定性。该项目的更广泛影响有两个层面。在社会层面上，通过实施所提出的能源感知DBMS，可以减少能源消耗和二氧化碳排放。该项目可以使大量工业部门受益，这些部门的运营严重依赖数据库支持的软件，如在线零售系统、财务管理软件和社交网络平台，从而大大降低其电力成本。在教育层面，该项目在跨学科的环境中培养博士生，并通过提供丰富的应用示例、软件工具和项目机会来增强佛罗里达大学和田纳西大学诺克斯维尔分校教授的几门课程。该项目的出版物、技术报告、软件和实验数据可在www.cse.usf.edu/~ytu/EDBMS上找到。

Maintaining a sustainable society via technological innovations has been a major challenge for computing system design. In this project, we focus on the energy efficiency of an important type of computer applications, the database management systems (DBMS), which often consume a large portion of the computing resources and energy in modern data centers. The goal of this project is to design and implement a DBMS that enables significant energy conservation with graceful degradation of query processing performance. The project achieves its goal using the following approaches: (1) Dynamically exploit the energy-performance tradeoffs in DBMS as well as low-power modes of hardware systems for improved energy efficiency with performance guarantees; (2) Formulate the energy-efficient DBMS design as a feedback control problem, and adopt appropriate formal control techniques to achieve the desired performance and energy efficiency with theoretical analysis and guarantees; (3) Apply advanced multi-stage optimization methods to solve complex energy-aware storage management problems; and (4) Coordinate various control and optimization loops in different layers of the DBMS for maximized energy savings and global system stability. The broader impacts of the project are at two levels. At the society level, the reduction of energy consumption and CO2 emission by implementing the proposed energy-aware DBMS can be substantial. The project can benefit a large number of industrial sectors, the operations of which depend heavily on database-supported software such as online retailing systems, financial management software, and social networking platforms, by significantly lowering their electricity cost. At the education level, the project trains Ph.D. students in an interdisciplinary environment and enhances several courses taught at the University of Florida and the University of Tennessee at Knoxville by providing a rich set of application examples, software tools, and project opportunities. Publications, technical reports, software and experimental data from this project can be found at www.cse.usf.edu/~ytu/EDBMS.