As bandwidth management has become an important topic, the IETF has held a handful of meetings to discuss the issues it involves and possible IETF work items that might address them. One of those meetings was a one-day workshop organized by the Real-time Applications and Infrastructure (RAI) area and held at Massachusetts Institute of Technology at the end of May 2008. Following that were two BoF (birds-of-a-feather) sessions, held at IETF 72 in Dublin.
RAI Peer-to-Peer Infrastructure Workshop
The one-day workshop began with a discussion that was designed to frame the issue and that included both the Internet service provider (ISP) and peer-to-peer (P2P) service operator viewpoints, as well as input with regard to additional scoping.
Jason Livingood and Rich Woundy provided some technical perspective from their vantage point at Comcast, the cable service provider. Participants learned that Internet service providers are observing enough P2P application traffic in their networks to impact customers’ delay-sensitive applications and services, such as VoIP. The unacceptable customer experience occurs when one or more of a subscriber’s delay-sensitive applications are noticeably degraded in performance because other subscribers’ P2P applications are consuming all available bandwidth-that is, in excess of network planners’ expectations. While the ideal solution might be to ramp up the bandwidth for each customer, there are both technical and business reasons that such a solution is not generally feasible. For one, some P2P systems literally know no bounds, thereby making aggressive use of all available bandwidth.
Jason and Rich’s goals for a better future include the following:
- Optimize to provide the best possible network experience for the broadest set of customers, which means minimizing or eliminating cross-customer service quality impacts.
- Enable continued Internet evolution in a manner that avoids a game of cat and mouse-that is, detection and mitigation of specific protocols.
On the other side of the issue were Stanislav Shalunov and Eric Klinker of BitTorrent, who offered the P2P perspective. After walking through some of the complexities they face in order to deliver the reliable and quick file sharing for which BitTorrent is known, they said there is a point where P2P and network operations sometimes run at odds with each other. When attempting to locate sources of file chunks, P2P networks look for the least common chunks, and they figure out where to get them. These are the less available chunks of what you’re downloading. However, from a network perspective, the location of these chunks may not be the most desirable. Stanislav and Eric observed that efficient choosing of peers could reduce transit traffic by more than 50 percent.
The afternoon presentations showcased proposals for moving forward. Some of the proposals focused on P2P technologies and oracles, such as P4P: Provider Portal for (P2P) Applications, by Haiyong Xie and Laird Popkin; Traffic Localization, by Yu-Shun Wang; and ISP-Aided Neighbour Selection in P2P Systems, by Vinay Aggarwal. Others focused on network and bandwidth management approaches, such as Bob Briscoe’s Solving This Traffic Management Problem . . . and the Next, and the Next, which can be found here.
The end of the day brought no immediate conclusions for IETF action items, but it did increase awareness and set the stage for BoF sessions to be held two months later at IETF 72.
The RAI area held the Application-Layer Traffic Optimization (ALTO) BoF in Dublin at IETF 72. Chaired by Enrico Marocco and Vijay K. Gurbani, the discussion focused on (1) possible approaches to exposing certain aspects of network topology to applications and (2) how P2P technologies might be enhanced to take advantage of that information.
According to draft-marocco-alto-problem-statement-02.txt, “Many of the existing overlay networks are built on top of connections between peers that are established regardless of the underlying network topology. In addition to simply achieving suboptimal performance, such networks can lead to congestions and cause serious inefficiencies. . . . [T]raffic generated by popular P2P applications often crosses network boundaries multiple times, overloading links which are frequently subject to congestion.”
Both the presentations and the discussion at the meeting focused on issues surrounding caching and peer selection (P2P application questions), as well as on bandwidth costs. A draft charter for a working group was proposed and discussed. However, while there was support for moving forward with this topic area, the sense of the room was that the draft charter did not adequately capture a clear problem statement that participants could support as IETF work. It was sent to the mailing list for refinement.
The Transport Area held the Techniques for Advanced Networking Applications (TANA) BoF, chaired by Stanislav Shalunov and Gorry Fairhurst. The session focused on some of the transport-layer possibilities for supporting bandwidth-intensive applications.
According to www.ietf.org/internet-drafts/draft-shalunov-tana-problem-statement-01.txt:
“The TANA BoF is held to explore the problem space, to gauge the interest in the problems within the Transport area, and to see if the community and the area directors believe that it makes sense to form a TANA working group within the Transport area chartered to work on
- standardizing end-to-end congestion control that enables advanced application to minimize the delay they introduce into the network and a protocol using it and
- a document describing the current practice of peer-to-peer apps’ use of multiple transport connections and recommendations in this space”.
Discussion between those in the group included Jason’s review of ISP requirements and Laird’s review of P2P application requirements. A number of issues and angles were also discussed, and at the end of the meeting there was clear support for moving forward with work on the first item.