Ph.D. candidate Bryan Donyanavard
Computer Science at UC Irvine

Abstract
As mobile battery-powered devices become ubiquitous, users and developers alike expect such devices to support a variety of demanding applications. In response, heterogeneous multicore processors have increasingly been deployed in order to provide a tradeoff between performance and power for workloads whose demands may vary over time. To effectively exploit these power-performance tradeoffs dynamically, we need to deploy intelligent runtime resource managers. Closed-loop resource managers are traditionally reactive as opposed to reflective, which is not always sufficient. In parallel, a number of disjoint research disciplines have been investigating computing systems that are inspired by self-awareness in humans. I will describe how I apply the high-level principles of computational self-awareness to design closed-loop resource managers for efficient execution of mobile systems in two scenarios: (1) for power management via DVFS, and (2) for QoS management within a power envelope.

Bio
Bryan is a Ph.D. candidate in Computer Science at UC Irvine. He is advised by Nikil Dutt, and his research focuses on runtime management of resource constrained systems. Website: https://www.ics.uci.edu/~bdonyana/