When Down Becomes Up

Silicon Alley Insider has a short interview with BitTorrent CEO Doug Walker about plans to entice the media Big Boys to use BT's peer-to-peer delivery service. Walker claims he can undercut the likes of Akamai and Limelight Networks.

P2P does a great job of file transfer. Streaming? Not so much.

Streaming via P2P doesn't use any less bandwidth that streaming directly. It just uses somebody else's bandwidth. So yes, it can be "cheaper", but this is virtual savings, not real savings. Once P2P streaming delivery gets big enough--if it ever does--the ISPs will step in and demand their share. This is what the whole net neutrality thing is about. In fact it's already happening, with Comcast selectively blocking some P2P clients. Even if there's no direct royalty to the telcablecos, sooner or later their bandwidth gets chewed up, and then they raise prices to consumers.

This thing won't scale. It runs up against the fundamental problems of the Internet in the U.S.: lack of edge capacity and asymmetric bandwidth.

Most internet connections are designed to be timeshared. Your advertised 2 Mb/s (or 10 Mb/s) link only gets that kind of speed if no one else in your neighborhood is using theirs. Cable systems often have a 500:1 share ratio. Even DSL is shared in a way, limited by capacity at the DSLAM in the central office (and it's typically slower to start with). Which all works fine for web pages where it's a quick download between idle times. But video streaming of any kind runs into real problems with enough simultaneous users, because there's a minimum sustained rate that must be achieved to avoid jitter and/or buffering.

What makes the problem worse for P2P is that almost all internet connections are asymmetric; the upstream bandwidth is an order of magnitude slower than downstream. This is because the phone and cable companies--with their heads firmly in the sand--never envisioned the internet as anything other than a way to shove increasingly expensive media down consumer throats. That is, after all, the model they were founded on. But P2P and User Generated Content are turning that notion on its head.

So we have too little bandwidth, and it's increasingly pointed in the wrong direction anyway.

Streaming via P2P works fine on a small scale (Joost's beta wasn't half bad; but notice how the more users it gets the less you hear about it?) And P2P is great for file transfer since there are no latency/buffer issues. Just don't expect it to be the answer to video delivery for consumers. Conversations with executives at the likes of Akamai (who has their own P2P technology) have confirmed peer streaming will likely be limited to private networks where there's greater control over the protocols and the bandwidth.

Don't get me started on streaming vs. download, and why the latter is better because storage is cheaper than bandwidth. That's a subject for another post.