The proposed research will investigate how to build intuitive systems that manage the complexity of mobile computing. The approach, self-tuning parameter translation, inserts a software translation layer above the mobile system - this layer exposes only a small number of meaningful, well-calibrated parameters to the user. By drawing from a toolbox of techniques that includes on-line modeling of application workloads, off-line benchmarking, adaptation, feedback, application hints, and cost-benefit analysis. This layer transforms user input provided through the simple intuitive interface into policies that continuously tune the complex set of parameters exposed by the underlying mobile system.
Two case studies are planned to demonstrate the feasibility of this technique. First, a translation layer will be build that provides self-tuning power management for mobile computing systems. Second a layer will be built that guides the migration of mobile services. A mobile service is the resource-demanding, remote component of a partitioned, latency-sensitive application - it migrates between servers in response to user mobility in order to maintain high-bandwidth, low-latency connectivity to the client. A toolbox of reusable components that can be shared across translation layers will be developed, as well as a methodology that provides guidance about when to use each component.
