Abstract

No longer is faster always better in computer system design. Nowadays, across all levels of computer systems, speed costs power and power costs money—so performance must be balanced with energy usage. The most common approach for balancing energy consumption and performance is dynamic speed scaling, which adapts the processing speed to the current workload. The focus of this talk is to understand some fundamental questions about speed scaling. A key feature our analysis reveals is a conflict between designs that provide optimality guarantees on performance and designs that provide robust performance across varying workloads. This motivates us to use an analytic approach that combines techniques from queueing theory and online algorithms to attain worst-case guarantees for optimal stochastic control policies. This talk presents joint work with Lachlan Andrew, Minghong Lin, and Ao Tang.