Rumours persist about AT&T's U-verse service
IPTV catastrophe - hype or horror?
Comment There are still stories emerging about the technology in the AT&T IPTV system U-verse, which suggest that the design of the Microsoft middleware will make something cataclysmic happen one day to all of the AT&T IPTV customers.
The scenario that is always put forward is that during the Superbowl the final deciding score might be missed, due to some glitch or other that affects everyone watching it on IPTV.
Really the industry has to "just get over it". I’m not fond of a monolith like Microsoft coming along and taking over an exciting technology like IPTV and making it mundane to write about because most telcos will opt to use it. But that happens to all technologies at some stage.
The previous stories all came about because Microsoft has failed to explain exactly how its systems works, so in the absence of an explanation competitors are willing to offer one.
Fast channel change, reliant as it is in the Microsoft TV IPTV Edition software on D servers that live inside the network, was thought to be potentially disastrous, until you understand that if it fails, the system just reverts back to the process of subscribing to an IGMP multicast, just the way every other IPTV system works.
The scenario was put to us was that everyone would change channel when the adverts came on during the Superbowl, and that this would overload the channel change servers, which could indeed happen. But the disaster scenario of the server rebooting itself, the routers that provide the tables that the server relies on crashing and the entire network bringing down the head ends, was all fantasy.
The same thing was last week put forward by Light Reading, as a major exclusive, which suggested that the AT&T trial installation has been blighted by lost packets. Somehow this would become disastrous as the system scales because Microsoft does a server based packet resend using a resilient version of UDP, which should arrive before the lost packet is needed, and avoid a blip on the screen.
The Light Reading sources said that if the packet was lost far enough down the network then everyone would have the same problem and they would all have a packet resend, and somehow this would happen in the middle of the Superbowl and the whole system would crash or under-perform. Of course, this is exactly how Fastweb is built, and it has almost half a million customers and it has never had a catastrophic failure.
It is the same old story again and again, and Microsoft's rivals should be pushing their own visions of how to make IPTV more valuable.
People on the AT&T trial report that they have had pixelation on the screen for a few seconds, once or twice in three months, and that not only is it not a problem, but they are more of them on existing cable TV systems. Apparently Forward Error Correction (FEC) may be added by AT&T to make it super resilient.
Microsoft sticks by its belief that resilient UDP is the way to go, describing it as a "patient, resilient retry of packets that takes place in a very short time". It points out that FEC works well when a pattern of errors is well-characterised, but added that packet loss in IP can't be characterised so succinctly and that both approaches are not in conflict with each other and can be used together if appropriate.
FEC is used in most wireless systems and is a key component of mobile TV, and 3G phone services. But the AT&T system is supposed to sample the bitstream at varies points in the network and repair lost packets as they go, and put the packets back in the correct order. That would be a major re-architecting of the network and we can't see this being added retrospectively.
But if a customer cannot notice these packet losses, then the catastrophic failure during the Superbowl theory, really needs to be put permanently to bed.
Light Reading went on to insist that there is already a 15 to 30 second delay to live video streams to allow some time for dealing with packet loss. But every pay TV system has delay imposed at the encoding end, to allow for compression time and it is worse on mobile TV systems.
Copyright © 2006, Faultline
Faultline is published by Rethink Research, a London-based publishing and consulting firm. This weekly newsletter is an assessment of the impact of the week's events in the world of digital media. Faultline is where media meets technology. Subscription details here.
Sponsored: Hyper-scale data management