NeTS: Small: Inter-Domain Traffic Engineering though ISP Cooperation and Competition

Rensselaer Polytechnic Institute, Troy NY

Investigators

Abstract

This project takes a comprehensive look at inter-domain traffic engineering, and how it can be done efficiently through cooperation as well as competition between Internet Service Providers (ISPs), and at different timescales. Using network flow decomposition and separate convex optimization techniques, the project seeks to design distributed inter-domain traffic engineering solutions that, along with effective intra-domain traffic engineering, converge to optimal traffic flows in the entire Internet. The project also involves modeling of traffic engineering - under competitive pricing of inter-domain forwarding by ISPs - as a two-timescale pricing-forwarding game, and the study of existence, computability, and efficiency of the equilibrium. The project explores dynamics and convergence of distributed pricing updates by ISPs, and compare different pricing strategies and models like point-to-destination pricing vs. point-to-anywhere pricing, linear vs. non-linear pricing, and cascaded vs. non-cascaded pricing. The project considers inter-ISP interactions and decision-making at multiple timescales that involve traffic forwarding (fast timescale) and pricing updates (slow timescale), where decision-making on a slower timescale may depend on that at a faster timescale. Furthermore, the project considers both cooperative and competitive approaches through which the Internet as a whole (and not individual ISP networks or Autonomous Systems) can be utilized efficiently. The models and approaches considered take into account the control and operational structure of the Internet that consists of multiple providers seeking to maximize their individual benefits, and accordingly considers the relevant issues like competitive efficiency. Using these models, the project seeks to develop competitive pricing-forwarding solutions that can converge to efficient and stable traffic flows in the entire network (Internet. Broader impact: The issues that the project investigates are timely, and fundamental to the efficient operation and economic viability of the Internet in the long run. Solutions that the project is developing will improve the overall use of the Internet bandwidth resources through optimal/near-optimal traffic cross-provider engineering in the Internet. Pricing strategies developed and investigated in this project will improve the Internet's economic efficiency and viability through inter-ISP competition. For providers, this would imply more efficient and flexible management of their available resources, better investment in infrastructure, and higher revenue overall. For end-users, it would imply better service at lower cost. Research results will be used in advanced graduate coursework on network optimization and algorithmic game theory. The PIs will also enhance the research and educational experience of undergraduate students, by supervising undergraduate research projects.