This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5).
Tremendous growth in hosted services and data sharing are motivating a new generation of storage virtualization technologies that encompass the entire data center or ensembles of data centers. Accurate dynamic resource provisioning is critical to achieving cost-effective and environmentally-sensitive operation of these service centers in order to (i) reduce fixed costs for floor space, server hardware, and infrastructure for power distribution and cooling, (ii) lower the operational costs for management and electricity, and (iii) mitigate the hidden environmental and social costs of excessive energy use.
The goals of this project are to develop effective dynamic resource provisioning and scheduling schemes for storage servers using novel workload shaping techniques to improve cost and energy consumption. The research involves the design, analysis, implementation, and performance evaluation of new algorithms for workload shaping with superior performance characteristics. This will result in significant improvements in resource provisioning, capacity prediction, and admission control, as well as energy and power consumption. Unlike traditional server scheduling techniques in which the negative effects of workload bursts affect even well-behaved portions of the workload, the proposed workload shaping approach isolates the bursts and confines their effects to localized regions using online workload decomposition and recombination.
The project will support the education of graduate students in an emerging sector of Information Technology, and will also involve undergraduate engineering students in research in a topical technical area.
