2008 Poster Sessions : Dynamic Task Fetching for Mobile Computing

Student Name : Dimitrios Tsamis
Advisor : Nicholas Bambos
Research Areas: Information Systems
As portable wireless devices have become commonplace to- day, the popularity and acceptance of a broad range of mo- bile applications is higher than ever. Acceptable user expe- rience warrants low latency of execution of computational tasks on the mobile terminals, which, owing to their porta- bility requirements, are typically constrained in memory. It is hence important to judiciously fetch new tasks from ap- plication servers while background applications are running on the device. In this work we use a dynamic programming (DP) framework to capture the tradeoff between congestion caused due to background tasks running on a mobile device and latency of execution of new tasks fetched from a cen- tral server over a time-varying wireless channel. Adopting a baseline model for wireless channel variations, rate of task execution, and congestion experienced at a mobile terminal, we establish the optimality of a switchover policy, which makes a decision to fetch or not depending on the number of tasks queued up for the mobile terminal at the central server and at the mobile terminal itself. We use the policy iteration methodology to develop an approximation to the optimal control and leverage it to design a low complexity heuristic algorithm FON (fetch-or-not). We illustrate our approach via numerical examples and also discuss severals modeling extensions to the baseline scenario.

Dimitrios Tsamis is PhD student in the Department of Electrical Engineering at Stanford University, working with Professor Bambos. His current research focuses on wireless networking architectures and network performance evaluation issues. Dimitrios received his diploma from the National Technical University of Athens, Greece where we worked with professor Venieris. Dimitrios is a recipient of the Stanford Graduate Fellowship.