news | need to know | home

Need to Know

Describing the technology and practices of television and new television technology.

August 2004

A Network Open for All the Video
An open garden is far more attractive to consumers than one that limits customers to a few program choices. Doing it right requires some thought and supporting technology.
· Plenty of bandwidth to the home, which would seem to leave the shared cable network far behind. Cable systems, by devoting additional channels to data, can easily offer 100 meg in peaks, and plenty of room for individual video, however. Cisco is shipping that gear in 2005.
· Sufficient capacity in the central unit, the DSLAM, switch or CMTS. The first generation Bell fiber, as well as most DSLAMs, can't keep up if many users are watching video at the same time. New non-blocking units solve that problem, at virtually no additional cost.
· Fiber throughout your network, with switch capacity that rarely overloads. Masayoshi Son at Yahoo BB in Japan connects each office with fiber and lights 2 Gig-Es. Most cablecos use similar strategy, while older DSL networks at 45 mbps and 155 mbps will create bottlenecks. A provider with fiber (purchased or owned) can usually increase the capacity at moderate cost, which is why solely implementing QOS is rarely the right choice. It doesn't add peak capacity, so has the effect of slowing down the network under heavy use. The practical effect is to make unreliable the basic net connection. Carriers should generally be able to move video within their own network at close to the speeds they are selling customers, such as 3 megabits. Adding the necessary switches and high-speed ports isn't free, but it a modest part of the cost of running a network. Slightly degrading your own network performance is a politically correct method to block other company's video coming over the net, while telling the regulators that's not what you are doing. So is refusing to peer, discussed below.
· A major server farm/video headend and/or an open peering point with good connections. The best networks offer both
· Server on net
· BT is the leading exponent of the massive headend server, at least as theyve described their plans. They'll provide all the computers, switches, software, format conversion, security and everything else required, including billing if you want it. Shared equipment, located at efficient points in BT's network, will be the most cost effective for everyone concerned. The devil's in the details of the pricing, of course, and they haven' made that decision. Video providers would have just have to send the programs to BT servers.
· Open peering
· The internet works by transferring data at peering points. If each broadband network connected at a public peering point, a large video provider could maintain their own servers and manage the cost and quality of their connection. Smaller producers will have a choice of commercial services, or could develop their own inexpensive system. Unfortunately, many carriers including most telcos don't offer a simple connection, forcing most video providers to pay expensive transit fees to backbone carriers like MCI, Level 3, or AT&T to deliver the programs. The better system would have each large carrier offering a connection at public peering points like MAE East or Equinix, accepting video traffic at reliable speeds. Providers would buy capacity to those peering points of the quality they need, and share the (modest) actual costs of the facility and switches. There's no law requiring a carrier like Time Warner Cable to peer, one way they can raise the cost of other video services. If they peer, there's no rule that prevents them from creating a bottleneck with a low capacity connection.
Freely forward this publication, and urge colleagues to take a free subscription as well. We will be supported by ads, although those in this issue were provided as a courtesy to DSL Prime advertisers.