The main objective of the proposed research project is to develop holistic approaches to the design of power optimization strategies for heterogeneous cloud computing environment. The focus will be on exploring the tradeoffs among the cloud computing physical infrastructure, including power-managed servers and cooling technology, virtualized infrastructure, and the middleware and service infrastructure. Contrary to cross-layer designs for energy optimization, where solutions are obtained through piecemeal approaches, the layered architecture is used to cast cloud computing energy optimization as an extended monotropic programming problem and derive a systematic analysis and design of global energy optimization strategies, whereby energy policy at a given layer uses subsets of the decision variables to achieve individual optimality and reduce the impact of carbon footprint on the environment. The research project provides a unique opportunity for graduate and undergraduate students to gain deep understanding of the basic science and engineering design principles of cloud computing and energy consumption in data centers. The outreach plan builds on existing minority and under-represented programs at the University of Pittsburgh to recruit and retain a diverse population of undergraduate and graduate students, and also seeks to establish collaborations with industrial institutions to leverage their educational programs and cloud infrastructure to advance the research project agenda.
