How BitTorrent wants to save the Internet

Janko Roettgers — Fri, 02 Dec 2011 19:20:30 +0000

This week’s talk about usage-based billing, spurred by the growing adoption of streaming services like Netflix and their impact on ISPs, has been a bit of a deja vu for BitTorrent CEO Eric Klinker. BitTorrent Inc. was at the center of the last big hubbub about clogged pipes back in 2007, when Comcast was caught throttling its customers’ P2P transfers. And just like back then, Klinker thinks the solution to ISPs’ current woes isn’t to charge or slow down customers. Instead, he believes smart technology can solve our problems and save the Internet in the process.

Klinker’s optimism isn’t just based on wishful thinking. BitTorrent file sharing is still responsible for 16.5 percent of all residential fixed-line data traffic in the U.S., according to recent Sandvine data. That makes the P2P protocol the second largest data service after Netflix, and twice as big as YouTube. But no one is really talking about BitTorrent clogging up the pipes anymore, and Klinker believes that has a lot to do with some tweaks the company made to its file transfer algorithms.

The history of µTP

How µTP works
BitTorrent clients used to exchange files via TCP, the predominant Internet data protocol. The problem with TCP is that it views any congestion as an error, and the only remedy it knows to that error is to resend data packages. Combine that with BitTorrent’s core sharing design, which rewards people for uploading as much of a file as possible, and you are in for a vicious cycle: If packets slow down somewhere between user A and user B, user A will just send more and more data, in turn leading to more congestion. BitTorrent’s µTP protocol, on the other hand, is based on UDP, which comes without TCP’s urge to constantly repeat itself. Instead, it uses quality-of-service algorithms built on top of UDP that can sense whenever there’s a congestion anywhere between user A and user B. Once an Internet traffic jam is detected, µTP yields until the pipes are cleared up again. For a more technical analysis, check this submission to the IETF.

BitTorrent Inc. started its work on traffic optimization back in 2006, when it acquired a small company called Plicto. The startup that had been founded by engineers who had previously worked on the Internet2 infrastructure, which is the next-generation, high-speed data network that connects universities and other research facilities. Plicto had figured out a way to optimize file transfers in a way that wouldn’t negatively impact the overall performance of the network, and BitTorrent thought this was a great asset to sell P2P to corporate clients.

BitTorrent’s business plan at the time was to help media companies with the distribution of large video files, and the thought was that a TV network would never buy into a solution that would impact people’s VoIP calls, or bring their web browsing down to a crawl. BitTorrent rebranded the technology as µTP to match the branding of its flagship µTorrent client and began to deploy it as part of its P2P-CDN to enterprise clients.

However, it quickly became apparent there was little money to be made with undercutting CDNs. The company scaled back its B2B offerings and instead began to incorporate µTP into its consumer file-sharing software, where it is now used by more than 100 million clients in the market.

At the same time, it began to pitch the solution to potential partners and standards bodies, and in 2010, BitTorrent finally open-sourced µTP. The technology has since been adopted by all the major vendors of BitTorrent-based file sharing software, including Vuze and Transmission. Klinker estimates that today, 80 to 90 percent of all torrent traffic is µTP-based. This would mean up to 15 percent of all residential U.S. traffic is now congestion-aware: an astonishing number.

Why BitTorrent loves declining numbers

Sandvine's most recent numbers show BitTorrent declining during peak hours. The company believes that's a good sign.

So what kind of practical impact has µTP had on ISPs and their network congestion issues? Klinker believes the proof is in the numbers. Sandvine’s data has widely been seen as proof that file sharing overall is receding. However, the folks at BitTorrent were ready to open the champagne when the network management company released its most report. The reason: Sandvine concentrates on peak bandwidth, and BitTorrent is now only responsible for 7.62 percent of all residential peak download bandwidth. “That’s a good indicator that it’s working,” Klinker told me, explaining that µTPs biggest goal was to steer clear of times when bandwidth usage peaks.

BitTorrent Inc. is now looking to make µTP more popular, both with ISPs and other companies that could use congestion-aware data transfers. There’s some interest in the IT world, and the company is working together with heavyweights like Microsoft to turn µTP into an officially recognized standard through the IETF. Klinker believes cloud backup services, companies that have to distribute big software updates and possibly even VOD vendors that don’t depend on real-time data transfers could all benefit from µTP.

How about congestion-based billing?

But on the ISP front, Klinker still has to fight some misconceptions. “It’s astounding how few of them understand how their business works,” he told me. The latest example is the recent debate around usage-based billing. Klinker doesn’t believe it would do anything to alleviate network congestion, and he wishes ISPs would be a little more innovative. One idea he floated to me was to make apps aware of caps. If BitTorrent could query a user’s data quota, then it could help users to avoid going over their caps, and eventually, ISPs could reward users for utilizing apps that are aware of network congestion and help to avoid bandwidth spikes. “You should really do congestion-based billing,” he suggested.

Image courtesy of Flickr user kevinspencer.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

AT&T’s New Bandwidth Cap Is Bad News for Netflix

Janko Roettgers — Mon, 14 Mar 2011 14:53:46 +0000

AT&T will soon cap its DSL bandwidth at 150 GB per month, the company confirmed yesterday. Customers who use more data during at least three months will have to pay $10 for each additional 50 GB bucket of data. That’s bad news for Netflix and its users, who could get dangerously close to the cap.

How much Netflix video does 150 GB get you? Not that much, actually: If you watch a movie like Moulin Rouge in HD, you’re going to use around 3.5 GB of data. A single episode of Weeds equals about 800 MB when watched in HD. If you were going to use all your 150 GB of AT&T bandwidth to watch HD video from Netflix, you’d only be able to watch about three hours per day — and that’s without doing anything else.

Nielsen recently estimated the typical customer is streaming around 11 hours of video from Netflix’s website per month. However, Nielsen’s data is based on PC and laptop usage only and doesn’t include any streams accessed via iPads, Roku set-top boxes, Blu-ray players or any of the other 250 devices Netflix’s streaming service is now available on. These devices have arguably been the biggest driver for the company’s online video growth, and they’re likely to also have a significant impact on many people’s bandwidth consumption.

Granted, all of this is pure back-of-the-envelope math. Real-life usage involves data transfer overhead, which eats up additional bandwidth. Then again, only a portion of the Netflix catalog is actually available in HD. Many TV shows are, but a good number of movies can only be watched in SD, which doesn’t eat up quite as much bandwidth.

Still, AT&T’s bandwidth cap could have a significant impact on the future of the service. Netflix currently only offers 720p HD. An update to 1080p would close to double its bandwidth impact, meaning that you’d suddenly only have 90 minutes per day to watch before you’d be billed extra by AT&T. Competitor VUDU is already offering 1080p streams, and YouTube has been offering 1080p for over a year. It’s technically possible; there’s demand for it; but bandwidth caps could prevent Netflix from upping the ante in terms of HD quality.

Bandwidth caps could also spoil Netflix’s attempts to position itself as an alternative to traditional pay TV. U.S. households watch more than five hours of TV per day. The average American would burn through his monthly AT&T bandwidth allotment in just 18 days if he’d cut the cord and replace all of his TV viewing with HD streams from Netflix.

The biggest issue for Netflix, however, could be the psychological effect. People will think twice about using Netflix if they think it will lead to extra ISP charges. The company is well aware of these issues; Canadian users, who often have to deal with much lower bandwidth caps, have the option to disable HD streaming entirely as part of their account settings. That’s right; Netflix offers the option to make your video streams look worse so you won’t give up on streaming entirely. There’s no word yet on whether a similar option will be introduced in the U.S. as well.