ISP-P2P Collaboration

Peer-to-peer (P2P) systems account for more than half of Internet traffic today, and an increasing number of user applications, e.g. Skype, Bittorrent, Joost, GoogleTalk, and P2P-TV rely on P2P methodology. P2P systems build overlays at the application layer, independently of Internet routing and ISP topologies. This leads to measurement traffic overhead and routing inefficiencies for P2P users. While P2P applications spur broadband access, they also take customers away from traditional telephones and pose significant traffic engineering problems for ISPs, thus putting them in a dilemma!

Instead of having network operators impede P2P traffic, which has anyway been unsuccessful so far, we propose a simple, general and unique solution based on active ISP-P2P collaboration, where an ISP hosts a server, called the oracle, that helps P2P users choose optimal neighbours. A P2P user sends the list of potential peers to the oracle, which ranks this list based on a number of factors that each ISP can decide individually. For instance, the ISP can prefer peers within its network, to prevent traffic from leaving its network. Further, it can pick higher bandwidth links, or peers that are geographically closer. Intra- and inter-ISP routing policies can also be considered by the oracle while ranking peers.

We have analyzed the scheme on various models of P2P systems, with both application and network layer simulation frameworks. We have also conducted Testbed experiments and Planetlab deployments. Our results show that P2P nodes, on consulting the oracle, are able to keep most of their peerings within ISP boundaries, without adversely affecting the overlay graph structural properties. Not only does P2P scalability improve, we also show that there is no adverse effect on the content search phase of P2P networks. The P2P users are still able to locate all available content, and the download times decrease considerably. The ISPs are able to save costs by keeping large amount of traffic within their network, perform better traffic engineering, and provide better service to all customers. We have demonstrated that the scheme maintains its benefits across various models of P2P user behaviour (churn, free-riding, query patterns), as well as different ISP and P2P topologies.

The concept is being extended to build a global coordinate system, based on collaboration between multiple ISPs on the one hand, and between ISPs and P2P systems on the other hand. Another extention of the oracle aims at reducing pollution and malicious content in P2P systems. Besides, the oracle is being adapted to work with P2P-TV and other media streaming applications prevalent in Web X.0.

We firmly believe that as the Internet evolves from a client-server model to a user-generated-content model, where different users generate, search and download content at the same time, and where the content ranges from low negotiation traffic in social networking sites to heavy multimedia content in file-sharing systems, our scheme based on active ISP-P2P collaboration will be a boon to both network operators as well as Internet users!

• Vinay Aggarwal [17] (contact person)
• Anja Feldmann [18]

• Obi Akonjang [19]
• Rumen Tashev
• Rüdiger Wolf-Sebottendorf
• Shailesh Gupta

• full list of publications on ISP-P2P collaboration [20]