CSR: Medium: Effective Control to Maximize Resource Efficiency in Large Clusters; Hardware, Runtime, and Compiler Perspectives

点击翻译按钮获取中文摘要

Current computer systems use energy and other resources inefficiently. To improve efficiency, this proposal rethinks the way computer systems are controlled. This proposal uses formal methodologies based on control theory, and an analysis that cuts across software and hardware, to obtain orders of magnitude gains in resource efficiency. The proposal designs controllers in the hardware, in the runtime system, and in the compiler. The expected result is that large clusters can deliver higher performance for the same energy consumption, or consume lower energy for the same performance.This work involves building a prototype running an experimental runtime system and a compiler. It develops a set of technologies that will improve the resource efficiency of general-purpose computers by one or more orders of magnitude. The work will enable current technology computers to continue delivering increased capabilities, providing economic benefits to our society. The research has four thrusts: adapting control theory to computer architecture, developing hardware controllers, extending runtime systems to actuate on the hardware, and compiler-driven control. This research will be performed in collaboration with AMD Corporation Research. This effort enhances multidisciplinary research and education on cluster and datacenter technologies at the University of Illinois. The material in this work will enhance the courses on computer architecture, runtime systems, compilation, and parallel programming that the investigators are teaching. This work will place undergraduates and under-represented students in research groups. The research will also have a major impact on industry, since it addresses a real, very timely technical problem: that of making our cluster and datacenter infrastructure more energy efficient.This project will produce various deliverables that will be placed in a website for the community to access. The project repository will include data, code, results, a hardware emulator, a runtime system, and a compiler. These materials will be made available for at least two years beyond the project's lifetime. They will be accessible from http://iacoma.cs.uiuc.edu.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.