OTT content will only scale with seamless integration of unicast, multicast

With most pay-TV operators now delivering some over-the-top (OTT) content via the Internet, or are at least considering doing so, the questions of quality of service (QoS) and scalabity arise and quickly become urgent as demand increases. That has been shown in several European countries, notably in the UK, where the BBC’s iPlayer catch-up service has generated a boom in OTT content that has sometimes led to deterioration in quality. One problem is that the network provider gets no revenue from all this extra traffic, and worse, has to cope with irate customers phoning their help desk to complain that the video is of far lower quality their access bandwidth should be capable of delivering.

Sorting this out will involve business negotiations between the various parties, including providers of content and the carriers as well as broadcasters and pay-TV operators. But, the broad technical solution is clear: Overlay the Internet with dedicated content delivery networks (CDNs), in effect replacing private operator networks with a series of publicly available services guaranteeing QoS between two points and potentially as far as the user’s point of access. There is much talk of cloud-based content delivery services as well, but whatever name is given to them, there will be an expectation that bandwidth and delay are as stated on the tin and will be provided continuously, not just as an average.

Most importantly, however, with OTT services providing an increasing amount of longer-form and live content, there has to be seamless integration between the unicast mode of delivery used over the Internet and multicast as applied for cable TV and walled-garden IPTV services. The resulting hybrid infrastructure could be called “real-time multicast” because it incorporates live unicasting while meeting the twin multicast requirements of delivering content only to people who want to watch it, as well as only downloading or streaming the content once across any given point-to-point link within the network.

Unless this step is taken, existing broadband networks will choke under the weight of proliferating unicast content, according to Nick Fielibert, Cisco’s European CTO for video technology. This means that the network must be capable of converting a unicast stream into a multicast delivery on the fly, with the aid of suitable caching out on the network. When there is just one person on a network watching a program, that content would be unicast all the way from the source to the destination. However, the moment a second unicast starts within a given service area, this would be detected and an instruction fed back to the last common branch point leading toward the two end points. Just a single stream would be transmitted to that branch point, and from there, the video would be unicast to each of the two users. This principle, which underlies the multicast idea anyway, can be extended to any number of users.

The consequences, as Fielibert pointed out, become increasingly profound as the OTT service scales up. In the case of 1 million people watching a live event in HD, each at a download speed of 5Mb/s, the core network load would be 5Tb/s if the video was all unicast. That, Fielibert said, would indeed swamp most infrastructures. But if the unicast can be converted to multicast, then the load would be only a few multiples of 5Mb/s, reaching out to the nodes serving each group of subscribers, depending on the network structure.