ISPs & P2P, Getting Along Without Getting Cozy

penguin-geek writes "Researchers at Northwestern University have discovered a way to ease the tension between ISPs and P2P users. As we all know, there's been a growing tension between Internet Service Providers (ISPs) and their customers' P2P file-sharing services, and this has driven service providers to forcefully reduce P2P traffic at the expense of unhappy subscribers and the risk of government investigations. Recently, some ISPs have tried to fix the problem through partnerships with certain P2P applications. The Ono project represents an alternative solution: a software service that allows P2P clients to efficiently identify nearby peers, without requiring any kind of cozy relationship between ISPs and P2P users. Using results collected from over 150,000 users, they have found that their system locates peers along paths that have two orders of magnitude lower latency and 30% lower loss rates than those picked at random by BitTorrent, and that these high-quality paths can lead to significant improvements in transfer rates. In challenged settings where peers are overloaded in terms of available bandwidth, Ono provides a 31% average download-rate improvement; in environments with large available bandwidth, Ono increases download rates by 207% on average (and improves median rates by 883%). Ono is available as a plugin for the Azureus BitTorrent client, an open tracker and an standalone service you can integrate into any P2P system."

Comments

Well, that took long enough

[Animats]

That's been the trouble with these "peer to peer" protocols. The routing algorithms have been horribly inefficient. It's quite possible to have the same data flowing in both directions on the same pipe. Multiple copies, even.

It might be cheaper for the telecom industry (which is big) to buy out the music industry (which is tiny) and just cache the RIAA's entire output on local servers. Just cacheing the top 100 releases or so might cut traffic in half.

(This won't scale to movies, though. Movies are bigger and more expensive to make.)

Re:Well, that took long enough

[JustinOpinion]

That's been the trouble with these "peer to peer" protocols. The routing algorithms have been horribly inefficient. It's quite possible to have the same data flowing in both directions on the same pipe. Multiple copies, even. Seems to me that is an artifact of a protocol being designed to operate on a hostile network.

Distribution could be wildly efficient if the users and the network operators were on the "same team." If they wanted to, they could design a bit-torrent variant where chunks are cached by intermediary servers, so that they can always be delivered quickly from a local node. Further, servers could maintain accurate models of network topology, and clients could then use this data to pick the best path. Chunks from popular files would almost always be available from a nearby server cache or a nearby peer.

The problem is that the network is either indifferent to user activities, or actively trying to prevent user activities (throttling, etc.). The end result is that the protocol is tweaked not for efficiency, but for circumvention (e.g. encryption).

I like the idea presented in the summary, since it is in principle a net benefit to both the users and the network operators. However even if it works, it may not last. For instance, ISPs may use even more aggressive tricks (maybe even exploiting this proposed variant), forcing the protocol to become even more inefficient (e.g. switching to a multi-hop TOR-like protocol).

"Nearby peer" mechanisms are anticompetitive

[Brett Glass]

One thing that many people do not think about at first (but realize when it's pointed out to them) is that mechanisms which try to identify peers on the same ISP's network are anticompetitive. (That's why only the biggest carriers, like AT&T, support them.) Here's why. The cable and telephone monopolies have so many customers that the odds are there will be someone else on the same provider's network with the requested files. Small ISPs, on the other hand, will rarely if ever have someone with that file and so will still experience a great impact from the cost shifting and congestion caused by P2P. Hence, you can see why the big guys are cautiously embracing schemes like "P4P" as an anticompetitive weapon to block new entrants -- particularly wireless ones.

Re:The problem is that it is stupid.

[CountZer0]

Except they aren't only looking at the physical location of the machines. They are basically merging both network and physical location to come up with a hybrid location mapping that provides the lowest latency route.

From the FAQ:
Does this really work? In a paper pending publication, we show that our lightweight approach significantly reduces cross-ISP traffic and over 33% of the time it selects peers along paths that are within a single autonomous system (AS). Further, we find that our system locates peers along paths that have two orders of magnitude lower latency and 30% lower loss rates than those picked at random, and that these high-quality paths can lead to significant improvements in transfer rates.

Re:Hot Potato for ISPs

[billstewart]

ISPs don't actually care about copyright infringement, except possibly the cable modem companies which are also selling television and might have their advertising revenues impacted. Back when Napster and @Home were still around, @Home had two positions on Napster - officially, they'd say "Evil Copyright Infringers are Bad! And people generating upstream bandwidth from home are Bad!". Unofficially, the people who worked there mostly said "Well, duh! The reason people are buying broadband at home is to download music - Napster's really great for us!"

ISPs care about money - buying more upstream costs money, and upgrading peering links or internal distribution networks costs money. They also care about customer perceived performance, and if P2P uses their networks inefficiently, and swamps a neighborhood's upstream in ways that interfere with TCP performance, that's bad. For the most part, this technology will reduce their costs by reducing exterior bandwidth, and that's good, as long as it doesn't do it in ways that the improved P2P performance finds other bottlenecks in their system to step on. The better the P2P paths can match the structure of the ISP, the lower the impact on their network will be.

This approach doesn't actually require the ISP to install anything, or to do anything, or expose them to participating-in-P2P-themselves infringement conflicts; there are other approaches that do, such as putting P2P caching servers in their network. So it's pretty much all gravy for them, especially since they know that some large fraction of the bits they're carrying are P2P. (The Akamai caching servers here aren't being used to cache the P2P - they're web caches used by traditional content providers, and what this tool is doing is using their location to identify some of the structure of the ISP network to do better P2P peer matching.)