End-users rely on Internet Service Providers (ISPs) for access to the Internet. Given the limited geographic coverage of these ISPs, the establishment of interconnectivity between ISPs is crucial for global connectivity. Internet Exchange Points (IXPs) are equipped with extensive network switches that enable ISPs to establish peering connections facilitating direct traffic exchange. In this talk, we will present results from the first systematic theoretical study of selfish peering by ISPs at these traffic exchange points. Analyzing traffic exchange between ISPs requires modeling the interplay between traffic pricing and traffic flows at the IXP, and consideration of the notion of pairwise equilibrium. We will first establish the efficiency of congestion-proportional pricing at IXPs. We will then extend the model to consider port capacity purchases, where queuing delay effects at the switch ports are also modeled. This model explores a unique game where port capacity purchase decisions are made by ISPs unilaterally, while traffic exchange rates are decided bilaterally by pairs of ISPs. For both models, we show that the worst-case efficiency at equilibrium, measured as the Price of Anarchy (PoA), is within a small constant factor of the optimum traffic exchange efficiency. Finally, we will consider the problem of automation of the two steps of the peering decision process: i) if two ISPs should peer, and ii) where they should peer. For this problem, we analyze the insights obtained from applying machine learning tools to routing and peering datasets available in public repositories such as PeeringDB and CAIDA. Towards the end of the talk, we will also give an outline of how some of our game theoretic analysis methods can be utilized for selfish flow routing, scheduling and pricing in other applications such as power and heat flow networks.
Dr. Koushik Kar is a Professor in the Electrical, Computer & Systems Engineering department at Rensselaer Polytechnic Institute, Troy, NY, where he has been a faculty member since 2002. Dr. Kar's primary research interest is in developing and analyzing network algorithms that can be implemented in a decentralized manner, through local coordination, only requiring limited information about the network topology and conditions. His recent work is mostly on the study of resource pricing and economic efficiency questions related to use of shared resources (like bandwidth and energy) by self-interested agents, with applications to the Internet and the smart grid. Dr. Kar received his B.Tech. degree in Electrical Engineering in 1997 from the Indian Institute of Technology, Kanpur, and his M.S. and Ph.D. degrees in Electrical & Computer Engineering from the University of Maryland, College Park, in 1999 and 2002, respectively. Dr. Kar received the Career Award from the National Science Foundation in 2005, has won multiple best paper awards in international journals and conferences, and has been on the editorial board of several top journals in the field of communications and networking.