BitTorrent’s Swarms Have a Deadly Bite On Broadband Nets
By Leslie Ellis -- Multichannel News, 5/8/2006
In case you missed it, the annual compilation of technical papers from the National Show is out. As usual, it’s worth a look, especially if you’re into immersion learning from the industry’s top tech thinkers.
If your copy got lost in the stack, or you’d rather read the Cliff’s notes, this week’s translation focuses on my top pick from the 2006 collection: A meticulously researched look into how the file-downloading scheme known as BitTorrent affects broadband networks.
The short answer: Badly. Based on the research, conducted by Terry Shaw, of CableLabs, and Jim Martin, a computer science professor at Clemson University, it only takes about 10 BitTorrent users bartering files on a node (of around 500) to double the delays experienced by everybody else. Especially if everybody else is using “normal priority” services, like e-mail or Web surfing, which is what tech people tend to call “best-effort” traffic.
(One quick side note about this year’s papers: “The Superbly Nerdy Title” award goes to Scientific Atlanta Inc.’s Bryant Best, for this doozy: “Management of Simulated Raman Scattering in CATV WDM Reverse Path Systems.” And no, he’s not talking about the packaged noodles that boil up fast and cheap.)
The Shaw/Martin paper starts with a few handy stats to pop into the frontal lobes, when thinking about services like BitTorrent: Some 60% of American teenagers equipped with (their parents’) broadband have downloaded audio and video files over peer-to-peer networks; some 18% of all broadband traffic carries the torrents of BitTorrent.
BitTorrent is uniquely more powerful than other peer-to-peer techniques, the authors said, because it uses a technique called “swarming.” They called it a “radical enhancement” to prior efforts, like Gnutella and Kazaa.
Here’s a simplified view of how swarming works. Say that’s you prowling around for a copy of Napoleon Dynamite over your cable modem, using the free BitTorrent software you downloaded onto your PC. You pick the title from a list. In the background, you get what’s called a “tracker,” which watches every other BitTorrent downloader of that title.
In the BitTorrent lingo, the thing you want is the “torrent.” Those who have the whole thing are the “seeds.” They dole out chunks of the digital thing you want. Your tracker interacts with seeds and other BitTorrent trackers, to fetch the chunks you need.
For a popular movie, it’s not unusual to see 30,000 downloaders pulling from each other, and from 200 seeds, the authors found. Once you get swarmed with the whole file, you, too, can become a seed. The more you share, the better download performance you get.
It was that last part that juiced the curiosity of the authors, who wanted to know: Are broadband networks vulnerable to that improved download performance?
A test ensued. Using BitTorrent, the authors repeatedly downloaded an undisclosed but popular-at-the-time movie, sized at 4.3 Gigabytes. They used “packet sniffing” techniques to periodically trace the packets of the download.
At this point, the paper goes into a crazy amount of detail that just won’t fit here. The condensed version: They took the raw data from the download traces, and plugged it into a simulation model. That way, they could figure out how the little picture (repeated downloads from three locations) mapped into the big picture (repeated downloads by zillions of cable-modem customers, worldwide.)
In the simulation, the authors looked at traffic patterns across several hundred cable modems, some set up with BitTorrent, some not. Some of them ran normal Web traffic; some ran voice services.
The not-so-shocking conclusion: BitTorrent uses tremendous amounts of bandwidth, especially in the upstream (home outward) direction. How much? Try 10 BitTorrent users gumming up 55% of the upstream signal path, per neighborhood node.
Other fun facts: The average download speed surpasses 500 kilobits per second. The PC used in the download test interacted with about 40 peers, to collect the chunks of the requested file. After a file was fully ingested, the test PC stayed up as a seed for 13 more hours.
In the paper, the authors avoided recommendations.
In the National Show technical session where the paper was delivered, though, the term “bandwidth management” came up several times.
So did talk of prioritization techniques for premium services.
Short of that, if you own and operate services on broadband, there’s probably one practical reality that emerges from peer-to-peer traffic like BitTorrent: It bogs down every other third-party provider of “best effort” services on the link, like Vonage and Skype, of the voice variety, and Slingbox, of the place-shifting variety, and any of the video hopefuls.
But that’s a morbid blessing, because it carries piles of blame about why such-and-such service isn’t working right. Behind the blame is the finger pointing — “I’m not slowing you down, he is.” Translation: More phone calls from cranky customers.
Visit Leslie Ellis at www.translation-please.com.