Congestion on the Internet : operator responses, economic analysis, and improving the network architecture

Abstract

Our thesis presents the results of monitoring traffic management techniques of network providers. We find that in addition to complete blocking of traffic flows on some transport layer ports, providers are also actively managing traffic based upon application layer content. We document how providers are engaged in not just deep packet inspection, but deep flow inspection - tracking, shaping, and actively managing the state of individual flows. We explain why network operators attempt to manage traffic in these manners by looking at both the technical and economic forces influencing their decisions. We argue that inefficient congestion management and contractual practices of network providers are not unexpected given the limitations of the existing Internet architecture. Indeed, managing network traffic based upon transport and application layer content are likely to become more prevalent as high bandwidth demanding applications such as video and peer-to-peer increasingly dominate the traffic mix on networks. The existing Internet architecture leaves network operators with little architectural support for making more efficient traffic management decisions or contractual relationships. We therefore consider how making the architectural and protocol changes proposed by Briscoe et al [14, 17, 16], might help better align the economic and technical interests of traffic senders and traffic carriers. Briscoe's proposal would enable traffic management and/or contractual relationships to be based directly on congestion information. Our contribution is to further the analysis of the economic implications of this proposal. We apply both game theory and elementary economic analysis to better understand the proposal's advantages and limitations.

(cont.) Aligning the congestion responses of end-users with the traffic management and contractual decisions of network operators is part of a larger challenge of designing the next generation Internet architecture and protocols while taking into account the economic interests of different stakeholders. Accepting a proposal such as Briscoe's requires the research and wider networking community to be able to balance both technical and economic arguments in a design process. We examine how this integration might occur. Establishing a means of discourse is vital to the objective evaluation of what constitutes good architecture and good research for these techno-economic problems.