The interesting question was what we could do in the network to improve the delivery of video. I'd spent a lot of time on the server and client side and much less time in the network. It was interesting to me to step into Redback and look at video from another point of view: How do you design the network so all of these applications can work well, and what capabilities from the server and client world would you try to bring into the network to enable new applications.
Some of the product plans include bringing server functionality in to the router. It's a cool place to do it because you have tremendous amount of bandwidth available at that point and a tremendous amount of data going through the system. You can achieve some can see some interesting cost and performance efficiencies there. From your point of view, where are service providers in terms of video service development?
We're at a place where consumer demand is growing quite rapidly and where a bunch of applications are out there that already work. So we have some idea of what people want and how they're consuming bits. We're feeling stress on the network caused by some of those applications, and we're trying to figure out how to design networks that can serve both predicted and unpredicted needs, as well as plan for predicted revenues. What kind of timeline are you working on?
I'm looking at fixing known technology problems, while at the same time future-proofing
There's a set of known issues with IPTV that everyone seems to encounter. Channel-change times aren't as fast as you'd like, for example. Packet loss in an IP network really degrades video quality, so there's the question of what to do about that. And there's the looming issue of more and more high-definition content in a bit-restrained network. If you're competing with other broadcast media that don't quite have the same issues, you want to be able to compete on the number of HD channels a household can view at the same time. With IPTV, there are a lot of unicast issues when people can choose when to watch something. It becomes another leg on the chair in terms of how people interact with media. But unicast delivery puts stress on the system because it's a one-to-one stream for every user out there. Something like a server is sending bits to that user and interacting with that user and knows the user has those session rights and privileges, like hitting fast forward. What are the Internet video issues?
On the Internet video side, there are some known problems like P2P traffic using up a lot of bandwidth, or content in the house or on the loop occasionally stepping on other content. If you're a provider and you want to enable a great Internet video experience, but if your network is oversaturated, you can end up providing a less than ideal streaming experience because there might be P2P content eating up your network bandwidth. If you can't tell which content is which, you'll deliver all of those bits with equal priority, but those for streaming video will look bad. Video is a very dependent process. If we're talking about megabit video, we're talking 50 to 80 packets a second, and you want them all of those packets to arrive on time. If even one doesn't, you'll see some sort of glitch, and most of them won't be pretty. The unfortunate thing about video, as in life, is that the least sensitive thing is the one you're least likely to lose. The thing whose loss would cause the most pain is the thing you're most likely to lose. Given the technology issues, how do you at Redback start helping service providers future-proof their networks?
Part of my role is to raise thoughts about things we could build now and how they would impact future services, as well as what new services we might want to offer. We talk to service providers about that and make a list of things that we want to keep in mind as we design today's solutions so those future possibilities aren't blocked in any way. Is IP video all about the consumer?
Flexible networks that can handle video are also good for business services. One thing you want to do for video is guarantee high quality of service on a dedicated pipe. So the same ability that provides a high-definition pipe for a consumer can be a great thing for a business that wants a certain amount of connectivity. So with video, you can think of business applications coming in at the same time and helping to monetize the network. It feels very clear to me that carriers are figuring out how to handle consumer and enterprise services. In terms of technologies or network elements needed to enable more video on demand or IPTV, do you have a list?
On the IPTV side, a lot of scalability issues need to be addressed to give people the experience they're used to. IPTV is a very challenging thing because customer expectations for TV are set so high. Viewers expect fast channel change; they expect no loss on the line. IPTV comes out of the box with a very high bar, and when you have that, you have to build a very good infrastructure from the beginning. IPTV has scalability challenges, so there are certain gotchas that happen for carriers.
The perfect storm for carriers could be channel changing during the Superbowl when everyone changes channels at the same time. With IPTV, you have to build for that technologically, and it could mean that a very expensive infrastructure will be required to handle it. How can you lower the cost of video infrastructure?
One of the things we've been starting to work on is building some of the technology that's currently in the servers into the routers. So basically, bring into the router the technology you might need for channel change or packet resend. If we do that, we believe we can massively reduce costs. Otherwise, carriers could end up with billion dollar server infrastructures where they have a lot of normal-looking servers scattered all over the network all trying to handle basic packet resends or channel changes. Is mobile video more complicated than the wireline?
In mobile, there's more uncertainty about the actual applications. I don't know how big a market there is for watching video on a mobile device. I've seen estimates that it could grow to a $6 billion market by 2012 or 2011, but I don't know exactly what that means.
Ericsson, Redback's parent company, is big in mobile infrastructure. We're seeing HSPA [a software upgrade that helps carriers upgrade subscribers to true broadband speeds over existing GSM Networks] evolving now. So in this country, we're going from cells that can broadcast from 3.6 megabits to 7.2. Then there will be some hardware tweaks, and we'll jump to 14.4 megabits, and then the next generation starts at above 100 megabits per cell. That's shared by people in the cell, but you can also have multiple radios if you need them. So in five years, certainly a megabit or two would be reasonable to pull from a crowded cell. In that world, you'll have enabled video that would look ok on the big screen. So does mobile broadband become the way we distribute television? In some places like rural China, it may be the best thing to do. There's just going to be a tremendous amount of change happening in five years. What does that mean for Redback in the next five years?
In the next 5 years, something like $10 billion will be spent in our part of market -- the edge router market -- to help accommodate these changes. That's just the edge router market, but it's all about these converged and new services, and bringing broadband connectivity to mobile devices. Those bits have to go through the edge of the mobile networks, so we end up having a strong play there.
One additional strength at Redback, which makes the video problem more interesting to think about is that Redback has a unified device. So when you're routing all of this traffic around, all of the different flavors of traffic can go through one box. This is helpful because you can then do a lot of load balancing of load and traffic prioritization. . In a converged network where all traffic going over same pipes, you're virtually controlling it, which can have lots of flexibility. What are your thoughts on the concept of "personalized" ads for IPTV?
There has been talk about micro-segmentation for a long time, like presenting a custom ad to each household, and I honestly have no idea how to do that. Other creative guys I know don't think they know how to do that, unless maybe they could hook into your brain stem.
But an idea that fits really well with IPTV is serialized advertising. If we record what ads have been watched, then create multi-part ads within a show or within a couple of hours of evening television. You could then make ads that tell a story in three or five parts that could follow you across channels, and the advertiser could actually tell you a story, so they could increase engagement with the consumer. Brands like that, and it's then an inventory you can sell for a bit of a premium.