Applying Adaptive Nonlinear Congestion Control in Communication Networks
Andreas Pitsillides
Department of Computer Science
University of Cyprus Nicosia, Cyprus
Email: [email protected]
Abstract
In the early 1990’s control of congestion in communication networks ATM-Asynchronous Transfer Mode, Internet) was mainly tackled by ad-hoc and /or open-loop approaches, often based on probabilistic models, derived from simplistic to extremely complex suppositions. The derived models were either used as is, or simplified to make computationally appealing. Our work, spanning almost 20 years, constitutes one of the early efforts in motivating and introducing adaptive nonlinear control theoretic approaches to control congestion, starting with ATM congestion control, and later internet congestion control. Over the years a number of problems were successfully tackled, and solutions which are deemed implementable were developed. In the talk, I will provide a snippet of some early congestion control approaches, for both ATM and internet congestion control, and then present aspects of our work over the years. Our early efforts were based on the derivation of appropriate models for the combined control problem of connection admission, flow rate, and bandwidth allocation (capacity, service-rate) under nonstationary conditions. The derived models adopted fluid flow in the state variable form arguments, and were used to describe the time varying mean behaviour of the traffic in a congested path. They were used in ATM, initially for ABR (Available Bit Rate), using an adaptive nonlinear control approach to derive the control strategy for the finite buffer and server case, that is insensitive to propagation delays. Then, we extended this work to include congestion control for a differentiated-services network framework and formulated our nonlinear control strategy in the same spirit as IP Diff-Serv. Control design variables were adopted, which can be used to influence the delivered QoS. Later, we adopted the max-min resource allocation optimization based framework using utility functions, constrained by the additional requirement that queue sizes need to be bounded, a common approach during the early 2000s, to propose solutions for the internet congestion control problem. We proposed an Adaptive Congestion control Protocol (ACP) with learning capability, which enables the protocol to adapt to the highly dynamic network environment to maintain stability and good performance. This was based on a new estimation scheme for the effective number of flows utilizing each link in the network, originating from robust adaptive control theory design. The ACP protocol was analysed using a nonlinear network model to generate phase portraits which demonstrate that ACP is stable for all delays. Global asymptotic stability results in the decentralised max-min congestion control case, in the absence and presence of queuing dynamics, were also investigated. Apart from the practical significance of the results, our work also demonstrates the effectiveness of formal control theory techniques in general and adaptive control techniques in particular in delivering efficient solutions in a highly complex network system such as the ATM and the Internet.
Biography
Andreas Pitsillides is a Professor in the Department of Computer Science, University of Cyprus, and heads the Networks Research Laboratory (NetRL, http://www.NetRL.cs.ucy.ac.cy). Andreas is also a Founding member and Chairman of the Board of the Cyprus Academic and Research Network (CYNET) since its establishment in 2000. His research interests include communication networks (fixed and mobile/wireless), the Internet and Web of Things, and Internet technologies and their application in Mobile e-Services, especially e-health, and security. He has a particular interest in adapting tools from various fields of applied mathematics such as adaptive nonlinear control theory, nature inspired techniques, and computational intelligence to solve problems in communication networks. Published over 230 referred papers in flagship journals (e.g. IEEE, Elsevier, IFAC, Springer) international conferences and book chapters, he is the co-author with Josephine Antoniou of the book Game Theory in Communication Networks: Cooperative Resolution of Interactive Networking Scenarios (CRC, ISBN: 978-1439848081, 2012), he is the co-editor with Petros Ioannou of the book on Modelling and Control of Complex Systems (CRC Press, ISBN: 978-0-8493-7985-0, 2007), participated in over 30 European Commission and locally funded research projects with over 4.5 million Euro as principal or co-principal investigator, presented keynotes, invited lectures at major research organisations, short courses at international conferences and short courses to industry. He serves on the editorial boards of the Journal of Computer Networks (COMNET) and International Journal of Handheld Computing Research (IJHCR), served on international conferences as General Chair (MEDHOCNET2012, ICT2011, EuroMedNet’98), Vice General Chair (WiOpt’07), international co-chair (INFOCOM 2003), technical program chair (MCCS05, ISYC06), and on executive committees (e.g. INFOCOM 2001–2003, and ICT98), technical committees, guest co-editor, invited speaker, and as a regular reviewer for conference and journal submissions. He is also a member of the International Federation of Automatic Control (IFAC) Technical Committee (TC 1.5) on Networked Systems, IFAC TC 7.4 on Transportation Systems and the IFIP working group WG 6.3.
