ICAC 2008 Tutorial Session on

Model-based Predictive Control and Application to Computing Systems

Chicago, IL, USA – June 2, 2008

Sherif Abdelwahed and Naga Kandasamy

Control-based methods have recently emerged as a promising way to automate certain system management tasks encountered in computing systems. This approach offers some important advantages over heuristic or rule-based policies for performance management in that a generic control framework can address a variety of problems (e.g., power management, resource allocation and provisioning) using the same basic control concepts. Moreover, the feasibility or stability of the proposed control scheme with respect to the performance metrics can be verified prior to actual deployment.

This tutorial introduces the main aspects of model-based predictive control techniques and their application to automated system management. The following topics will be covered:

Dynamic models of computing systems: from finite state machines to hybrid automata.
Model predictive control: concepts, variations, and applications.
Feasibility and stability analysis: satisfying operating constraints and ensuring quality-of-service.
Forecasting and estimation techniques: ARIMA models and Kalman filters.
Application of model-predictive control to performance management in distributed computing systems.
Practical considerations in control design.
Future research directions.

The target audience consists mostly of the core audience of the autonomic computing conference, including researchers and engineers interested in the performance management of computing systems using model-based control-theoretic approaches.

Biography

Dr. Sherif Abdelwahed received his Ph.D. degree in Electrical and Computer Engineering from the University of Toronto, Canada, in 2002. He is currently with the ECE Department at Mississippi State University as an Assistant Professor. Dr. Abdelwahed research interests include control-theoretic techniques for designing self-managing systems, model-based diagnosis, modeling and analysis of distributed real-time systems, and model-integrated computing.

Dr. Naga Kandasamy is an Assistant Professor in the Electrical and Computer Engineering Department at Drexel University where he teaches and conducts research in the area of computer engineering, with specific interests in embedded systems, self-managing systems, reliable and fault-tolerant computing, distributed systems, computer architecture, and testing and verification of digital systems. He received his Ph.D in 2003 from the University of Michigan.