The original Cringely frog
mascot, circa 1998. March 2, 2006, 2006
Peering into the Future:
Why P2P Is the Future of Media Distribution Even If ISPs Have Yet to Figure That Out
By Robert X. Cringely
"If you build it, they will come."
Residential broadband use is still growing in America, but the rate of growth has fallen significantly according to a pair of studies covered last week in this column. This is nothing dreadful and nothing permanent, but under the current system of solely personal computers using the Internet, it will take another generation (20 years) to reach near-100 percent market penetration comparable to telephones. Ironically, by that time homes with traditional hard-wired phone lines will have dramatically decreased, showing us the other end of the bell curve as technologies become obsolete. But it doesn't have to be this way, as technology companies well know. The way to kick broadband growth back into top gear is to change the nature of the network and its interfaces, adding phones, television, and home automation to the mix. And that's why we see huge efforts in all these areas. But the point I made last week that inspires this week's column is the idea that only through peer-to-peer data distribution can this new network operate efficiently.
It is very hard to get your mind around the enormity not of the Internet, but of the Internet-on-steroids we'd need to absorb most other forms of communication and media distribution, but let's try anyway. "Desperate Housewives," in its puny 320-by-240 iTunes incarnation, occupies an average of 210 megabytes per episode. A full-resolution version would be larger still. In theory, it would be four times as big, but practically it would probably come in at double the size or 420 megabytes. But let's stick with the little iTunes version for this example.
Twenty million viewers, on average, watch "Desperate Housewives" each week in about 10 million U.S. households. That's 210 megabytes times 10 million downloads, or 2.1 petabytes of data to be downloaded per episode. Fortunately for the download business model, not everyone is trying to watch the show at the same time or in real time, so iTunes, in this example, has some time to do all those downloads. Let's give them three days. The question on the table is what size Internet pipe would it take to transfer 2.1 petabytes in 72 hours? I did the math, and it requires 64 gigabits-per-second, which would require an OC-768 fiber link and two OC-256s to fulfill.
There isn't an Internet backbone provider with that much capacity, much less excess capacity. Fortunately, it wouldn't have to all go over a single link and could, instead, be injected centrally into the network and fan out to viewers all over the country, in which case the OC-48 and OC-192 links used by Global Crossing, Sprint, MCI and others just might be enough.
But that's just one popular show. What will we do, then, with American Idol?
Ah, but remember Moore's Law, which is going to increase our bandwidth dramatically over time! It doesn't matter. Throw 250 million viewers watching 180 channels up on the Net, raise the resolution to full broadcast then raise it again to HDTV, and even Moore's Law won't catch up. Just carrying all the viewers of "Desperate Housewives" at the current iTunes resolution won't be economically viable for another decade according to Moore's Law.
I am no Luddite. IP is the future of global communication on all levels. But adding video to the mix is so bandwidth intensive that using current techniques will push back total IP conversion for decades.
Still, there is incredible incentive to push this digital conversion and a heck of a lot of money on the line. So we'll just have to cheat.
And that brings me back to the peer-to-peer schemes I discussed a little last week. By using excess upload capacity of client nodes as repeaters, putting together 64 gigabits-per-second actually isn't that hard. Stealing 256 kilobits per client would require a total of 256,000 participating clients to do the job, which is only 2.5 percent of the total "Desperate Housewives" viewer population.
There are some who believe Bit Torrent alone can do the job, but I feel that a true media market is going to require more components than are currently offered in Bit Torrent. There have to be payment systems, rights management systems, and some underlying quality-of-service layer that can save the day just in case you are the only person in the world who wants to watch that particular episode of "The Green Hornet."
Last week, I wrote about Seattle-based Grid Networks, which has many of the networking components in place. Grid, which already has more than 100,000 active nodes, can reliably deliver massive amounts of data at bit rates as high as clients are capable of receiving. That's a key differentiator -- being able to aggregate bandwidth to deliver not only lots of data, but lots of data FAST. Grid also appears to be the only major player in this new space that plans to offer Windows, Linux, and Macintosh clients, while most of the other services seem to be Windows-only.
Another emerging player in this space is Network Foundation Technologies (NFT), a startup in Louisiana that is taking the very non-Bit Torrent approach of using a binary-tree (one parent) distribution model rather than Bit Torrent (or Grid's) multi-parent distribution. Binary-tree distribution means that every node in the network has a single parent and no more than two children, which limits the total available bandwidth per connection. So where Grid and Bit Torrent can dump a lot of data very quickly on a given client, NFT is limited to half the capacity of the node above it in the distribution tree. While this would appear to be a disadvantage, there is one very good reason for doing it: NFT's binary distribution trees support live video.
NFT's binary trees form and re-form dynamically as client nodes appear and disappear from the network. The system is based on TCP, not UDP, and has been shown to support 14 or more levels, which is enough for tens of thousands of nodes, all fed from a single master connection equivalent to DSL. That means if you are willing to live with sub-megabit bandwidth, you could start an NFT video channel from your cable modem. Every man a TV network. System delay varies from 30 seconds to two minutes, but you could watch a live football game on NFT, where you couldn't easily do so on the multi-parent systems.
The merging Big Kahuna in commercial peer-to-peer seems to be Wurld Media's Peer Impact, which has similar technology to Grid Networks (though Windows-only), but where Grid is a networking company, Peer Impact is a media company and actually has a pretty compelling business model.
Peer Impact is up and running right now, though most of what the network has available isn't TV or music, but video games. About 1,100 video games from most major publishers except Electronic Arts are available through Peer Impact. The network has also announced it is adding video content and movies from NBC/Universal, and says it will have all major film studios and all major record companies onboard by the end of this year.
Peer Impact is similar to iTunes in that Apple sets the price ($0.99 per song and $1.99 per show). Where Peer Impact is different is in its use of Microsoft DRM, Windows client software, and a peer-to-peer distribution scheme. But where the company is REALLY different is in its relationship to participating nodes: Peer Impact pays users.
Ten percent of gross revenue for Peer Impact goes back to the participating nodes through two different programs. The first program is simple carriage: If someone downloads a song through your node, you get paid based on what percentage of the total bytes you provided, with five percent of gross revenue devoted to that task. The other five percent is for people who actively promote certain content through fan sites, for example. If you run a web site honoring the works of John Cassavetes, it could include links to his movies. If someone downloads "Minnie and Moskowitz" through your site, you get five percent of the download fee even if you don't actually carry any of the bits. If you carry some or all of the bits, you can get up to another five percent. All this is handled through PayPal and Peer Impact assumes most participants will spend most of that on more movies, which is part of their reason for doing it this way.
Here, finally, is a business model for video distribution that makes some sense. Nobody gets rich, but members are at least partially compensated for their participation, which ought to make the network robust and explains all those content deals with studios and record companies.
The best way to make money with Peer Impact, it seems to me, is through video games, which cost up to $30 or more and can be downloaded as 30-day demos. The trick, then, is to have a game fan site, stock it with demos that establish you in the distribution system without having to actually buy the games, and collect $1.50 to $3.00 every time someone downloads through your site.
Of course, there is an argument against all of this, that the ISPs won't allow it. But understand that it is my job to look into the future, and as explained above, there simply is no alternative to this kind of solution if the Internet is to grow and subsume these more traditional media types. That means IT IS IN THE INTEREST OF THE ISPs FOR THIS TO HAPPEN. It helps them, too, by creating a distribution system that jumps 20 years ahead of Moore's Law yet not putting a significant impact on the most expensive part of their operation, their Internet backbone links. Intra-ISP bandwidth is cheap, especially if it is within the same Point of Presence or data center. Inter-ISP bandwidth is expensive. By having enough peering nodes inside the ISP, connections to the greater Internet can be minimized and costs kept under control.
In short, if this technology didn't already exist, the ISPs would have to invent it, even if they don't yet understand that concept. In time they will. In fact, it would be to their advantage to take it upstream and put the client software in the DSL or cable modem, and effectively become a reseller for peer-to-peer firstname.lastname@example.org
That last bit is a frigging brilliant idea and somebody should run with it.
I've mentioned three companies here, so which will be the winners and which the losers? Right now, they all look like winners to me. Peer Impact has the business model and the distribution deals. Grid Networks is cross-platform and more flexible so it will appeal as a second standard as well as having a deep IP portfolio that could become quite strategic. And Network Foundation Technologies plays the live TV card, which means sports and news events can be watched live just like we are used to doing in the current cable and broadcast TV systems.
They could succeed on baseball, basketball, and football alone.