DESIGNING COMPLEX COMMUNICATION NETWORKS: MODELING, ANALYSIS AND ALGORITHMIC METHODS
Prof. Mike Devetsikiotis
September 13, 2PM
We present an overview of our research work which is centered on the need for systematic study of complex, real-life user traffic and its interaction with high-speed networks of rapidly growing size and complexity. The objectives are to study the effect of long-range dependent or self-similar sources on network operation and the likelihood and circumstances of multi-stable or chaotic network operation due to congestion and feedback mechanisms, whereby the network experiences sudden and persistent degradation in performance. Furthermore, we aim to develop "self-sizing" mechanisms for fast internets under conditions of varying traffic in terms of characteristics and service requirements. Such mechanisms will allow future high-speed networks to automatically determine how much bandwidth to add and where to add it, as well as the partitioning of capacity among elastic network "bands", in order to meet the traffic requirements and required Quality of Service (QoS). Our techniques for predicting the values of QoS are based on on-line measurements and empirical effective bandwidth functions. Finally, in the area of the network-user interface, we formulate and solve minimization problems to determine the optimal parameters for "traffic shapers". In our approach, traffic shaping involves minimization of the connection cost in terms of the network resources, such that access to the network is obtained and the desired QoS is provided by the connection. We describe an algorithm for the proposed scheme and show its workability for self-similar traffic.
Mike Devetsikiotis received his Dipl. Ing. in Electrical Engineering from the Aristotle University of Thessaloniki, Greece, in 1988. He received the M. Sc. and Ph.D. degrees in Electrical Engineering from North Carolina State University, Raleigh, in 1990 and 1993, respectively. Since 1993 he has been with Carleton University, in Ottawa, Canada, where he is now an Associate Professor of Systems and Computer Engineering. His present research focus is on the design, system modeling, and efficient simulation of communication networks.
The Network Systems Colloquium is sponsored by the Network Systems Lab at Purdue University. For further information, please contact Kihong Park (email@example.com or 765-494-7821).