The $7 TV Network
The
architects of TCP/IP, Bob Kahn and Vint Cerf, knew exactly how they'd
bring television to the Internet. They would use multicasting, a
particular IP service that allows many hosts to share a multicast
address and, through that address, receive the same content whether
that's rocket telemetry or Wheel of Fortune. Multicasting was seen as
the most efficient way for one computer to talk simultaneously to a
million others and that capability has been built into every router
pretty much since the beginning, though generally not enabled by the
network administrator. A decade ago Cisco Systems bought Judy Estrin's
Precept Software and its IPTV video product specifically to throw
television on the web and hopefully encourage admins to turn on
multicast support. No such luck. That is until next week's Video on the
Net show in San Jose when a start-up from Evanston, Illinois will
introduce Neokast, which is effectively multicasting for the masses.
Soon every computer will be able to broadcast to the world.
Multicasting hasn't broadly succeeded before now primarily because
it places a large burden on the routers, which are responsible for
caching and retransmitting video. Multicasting is generally turned off
in routers to save bandwidth and keep the network running as fast as
possible. Cisco wanted to turn multicasting on for IPTV specifically so
the routers would slow down and have to be replaced. With Cisco it
always comes down to routers and how to get people to buy new ones.
That's evident in Cisco's purchase this week of WebEx, where we can
expect Cisco to strongly push video services on those two million WebEx
customers, straining the system and forcing hardware upgrades. It's not
about Microsoft; it's about the routers.
Precept Software found that video worked well on a LAN but poorly on
a WAN, where lack of multicast support required creating VPN tunnels
that were just too much overhead for last-century PCs and networks.
Much the same applied to the Mbone, or Multicast Backbone experiment
from the late 1990s. Academics will argue that Mbone was a success, but
if that were truly the case, why aren't we watching TV over Mbone today?
Neither our PCs nor the Internet were ready for multicasting in
1997, but today they are, the trick being to somehow enable an
efficient multicast-type experience without turning on multicast
support in the routers, where multicasting remains switched off.
Enter Neokast, the brainchild of a PhD candidate from Northwestern
University, Stefan Birrer. Neokast uses peer-to-peer technology to
effectively emulate a multicast experience.
Neokast presently operates as a .NET application, meaning it is
limited for the moment to Windows computers. The player can operate as
a stand-alone application or a browser plug-in. And as far as the user
is concerned, connecting to a video stream is a matter of going to a
web site and clicking on a link. The viewing experience is very much
like cable or broadcast TV because with Neokast you aren't initiating a
video stream, you are joining a broadcast in progress. There are clever
ways to use Neokast for video-on-demand, but right now the company is
emphasizing its broadcast-like features.
The way the P2P components work is simple to describe but hard to
accomplish. I watched a Neokast demo from my office in Charleston and
the stream began playing in about two seconds, which is close to
instant-on in the world of Internet video. If there is buffering
happening, as there simply must be, it isn't a very big buffer. Had
there been no peers up and running other than mine, the video would
have streamed straight from the server in Chicago, but with enough
peers operating, the load on the originating server is several orders
of magnitude less than for typical one-stream-per-user distribution.
For content creators this is key: the more people who watch your
Neokast the more efficiently will your server bandwidth be utilized.
According to Birrer, under normal circumstances the server bandwidth
should plateau at 3-4 times that of a single stream NO MATTER HOW MANY
VIEWERS ARE BEING SERVED. With a per-stream bandwidth of 700 kilobits
per second, this means that Neokast would never require more than a
continuous three megabits per second of server bandwidth per video
channel.
Let's put that in a real-world context. Three megabits per second is
almost precisely 1000 gigabytes per month, which is half the allotted
monthly throughput for a $6.99-per-month web site at 1&1. So if
Neokast's claim is valid, it would be possible to broadcast American
Idol or the Super Bowl or friggin' CNN worldwide for $7 per month.
Heck of a deal, eh?
Neokast is not a video technology, it is a networking technology, so
there is no proprietary video codec. For the moment Neokast uses Xvid
because its free licensing appeals to a bunch of grad students
programming out of several apartments in suburban Chicago. But Neokast
could just as easily use another codec like H.264, possibly improving
its video performance, which I found already quite acceptable.
Here's what makes Neokast different from other peer-to-peer
products, especially BitTorrent. Unlike BitTorrent where you are
downloading a file, with Neokast you are intercepting a video stream.
Under normal circumstances the client does not end up with a local copy
of the complete video, the P2P caching aspect of the product being
limited to a few seconds or at most minutes of content. This makes
video rights management much easier and ought to appeal to TV networks
and movie studios.
Also unlike BitTorrent. Neokast is scrupulously polite. According to
Birrer, the key technology is Neokast's flow-control algorithms that
help the program to play well with networks and with other
applications. Neokast gives a preference, for example, to local peers
wherever possible, allowing a single copy of the show to come through
your Comcast or AT&T gateway then ricochet around inside the local
subnet, thus limiting the local ISP's Internet bandwidth hit to not
much more than the native 700 kbps. This trades inexpensive intranet
bandwidth for much more expensive Internet bandwidth.
Neokast uses a varying combination of TCP and UDP packets and looks
to the network a lot like web surfing, so it is difficult for ISPs to
limit the program with traffic shaping. But why would ISPs even want to
limit Neokast? They ought to encourage its use as a more efficient
protocol.
There may well be other products that can duplicate Neokast.
Certainly there are others that claim to. But the ability to stream
live TV is rare on the Internet, and Neokast's claim to carry live
programming with no more than a 10 second total end-to-end delay might
well be unique.
Neokast will be shown for the first time in public next week at Jeff
Pulver's Video on the Net conference in San Jose, but for a sneak peak
you can watch a short video I made about Neokast this week,
interviewing the founders. That video is among this week's links.
Neokast comes from a company mysteriously called Metis Enterprise
Technologies, LLC when they ought to have simply named the company
Neokast. And no, I don't have any financial interest in Neokast, but
thanks for asking.
The people behind Neokast have the goal of making everybody a
broadcaster. I can see Neokast appealing to established TV networks or
local stations that could put their present products - commercials and
all - straight up on the net. It might appeal, too, to movie studios
that like the idea of viewers watching without being able to
necessarily copy the film. But the news implications of somebody
setting up a webcam from their window in Baghdad or Darfur and serving
a truly global audience is what appeals to me.
Of course it will eventually be harder to do than I have described
here. 1&1, for example, would balk at having their bluff called on
that loss-leader bandwidth allotment. Just finding what you want in a
million-channel world would be hard, too, though less hard than
figuring out what to put in 24 hours of programming per day.
All Cringely, all the time? I don't think so.
No comments:
Post a Comment