Collaborative Research: A Hybrid Systems Framework for Scalable Analysis and Design of Communication Networks

University Of Delaware, Newark DE

Investigators

Abstract

This collaborative project delineates an ambitious multi-disciplinary three-year research and education plan to develop a scalable framework for modeling, analysis, and design of protocols for heterogeneous communication networks. It encompasses the design of models for hybrid networks and will provide tools for protocol design and analysis. Hybrid models combine discrete events with continuously varying quantities. They allow some of the discrete events that occur in computer networks to be aggregated and, as a result, networks can be described in terms of continuously varying quantities (such as rates of occurrence of events). However, they can still retain discrete events of interest (e.g., transitions between modes of congestion control, blackouts in wireless networks, changes in routing, etc.). Different aggregation levels result in multi-resolution models, which scale to arbitrarily large, high-bandwidth networks and thus constitute invaluable tools for protocol design and evaluation. The research proposed will produce the scientific foundation for a scalable methodology to study heterogeneous data communication networks, including the development of multi-resolution models for large-scale, heterogeneous communication networks based on a hybrid systems formalism, development of model-based tools for formal analysis of data communication networks, and in particular, for evaluating and tuning network protocols, and, development of model-based control-theoretical tools for designing network protocols that satisfy application-specific (e.g., bounds on sending rate variation) and/or general (e.g., fairness) performance constraints.