poster notes

FeedTree: Distributing Web micronews using peer-to-peer multicast

Authors

Dan Sandler, Peter Druschel

Abstract

Original: (from paper)

Syndication of micronews, frequently-updated content on the Web, is currently accomplished with RSS feeds and client applications that poll those feeds. The recent surge in popularity of this technology has caused RSS providers to become concerned about the escalating bandwidth demand of RSS readers. Current efforts to address this problem by optimizing the polling behavior of clients sacrifice timeliness without fundamentally improving the scalability of the system. In this project, we argue for a micronews distribution system called FeedTree, which uses a peer-to-peer overlay network to distribute RSS feed data to subscribers promptly and efficiently. Peers in the network share the bandwidth costs, which reduces the load on the provider, and updated content is delivered to clients as soon as it is available.

Tweaked:

Widespread distribution of "micronews", the frequently-updated content that characterizes today's Web, is increasingly accomplished with RSS feeds and client applications that poll those feeds for updates. The recent surge in popularity of this technology has caused RSS publishers to become concerned about the escalating bandwidth demand of the growing RSS user community. At the same time, users interested in more timely updates have every incentive to exacerbate network problems for publishers by polling feeds more frequently. Current efforts to solve these problems, including restrictions on clients' polling behavior and centralized caching schemes, do nothing to fundamentally improve the scalability of the system. We have developed a micronews distribution system called FeedTree which replaces polling with peer-to-peer multicast, which delivers feed data to subscribers promptly and efficiently. Peers in the network share the bandwidth costs, which reduces the load on the provider, and updated content is delivered to clients as soon as it is available.

Tweaked 2:

Widespread distribution of "micronews", the frequently-updated content that characterizes today's Web, is increasingly accomplished with RSS feeds and client applications that poll those feeds for updates. The recent surge in popularity of this technology has caused feed publishers to become concerned about the escalating bandwidth demand of the growing subscriber community. At the same time, feed subscribers interested in more timely updates have every incentive to exacerbate network problems for publishers by polling feeds more frequently. Current efforts to solve these problems leave intact the polling architecture of RSS and thus do nothing to fundamentally improve the scalability of the system. We have developed a micronews distribution system called FeedTree which uses a peer-to-peer overlay network to deliver feed data to subscribers promptly and efficiently. Participants in the network share the bandwidth costs, thereby reducing the load on publishers, and updated content is delivered to clients as soon as it is available.

One Year Later: (this is not very good/useful)

In recent months RSS has gone from obscure acronym to pervasive Internet movement, as many popular news sites and top-tier technology firms now proudly advertise support for this This phenomenal adoption continues to reveal the shortcomings in the conventional RSS service model, whose polling architecture places a substantial network strain on publishers while asking users to endure update delays of 30 minutes or more. We have developed a micronews distribution system called FeedTree, which uses a peer-to-peer overlay network to distribute RSS feed data to subscribers promptly and efficiently. Peers in the network share the bandwidth costs, which reduces the load on the provider, and updated content is delivered to clients as soon as it is available.