Your browser is not supported. Please update it.

3 November 2011

Broadcasters call for CDN tweaks, to globalize online video delivery

When the BBC and Sky agree on something to do with broadcasting it may still not be right, but we do at least take notice. So it was at the CDN World Summit in London last week, when there seemed a general consensus, spearheaded by the BBC and Sky, that CDNs were doing well in helping content owners and broadcasters deal with an increasingly complex and fragmented ecosystem for video delivery. Steve Bennedik, head of technology at Sky News, even described CDNs as the unsung heroes of digital media, which sounded strange given how much publicity CDNs have enjoyed of late. CDN vendors themselves have done a great deal of singing their own praises.

But after paying homage to the CDN industry, the conference identified some clear gaps in their armory, highlighting a need for better guarantees over quality of delivery, cleaner handoff, and support for the common file format defined by the Digital Entertainment Content Ecosystem (DECE).

This last point was perhaps most interesting since DECE members include the major Hollywood studios along with various pay TV operators, technology vendors, and digital content retailers, whose remit is to create an ‘open’ specification for the secure delivery of digital entertainment. The body’s most prominent initiative is the Ultraviolet digital locker or ‘digital DVD’, but underlying this is the common file format designed to ensure that content can be played via any service or on any device conforming with DECE specifications.

Among delegates at the CDN Summit calling for DECE Common File Format support was Maria Ingold, head of technology at FilmFlex, a UK pay-per-view movie service, which is a joint venture between Sony and Disney . Ingold reckoned widespread adoption of this format by CDNs would transform the whole video industry by avoiding duplication of content within the global delivery infrastructure. At present different wholesale or retail content providers, including the likes of Amazon subsidiary LoveFilm and FilmFlex itself, store and serve the same content. With a common file format, content could reside in a single set of distributed locations around the world and be served from there irrespective of the content provider. This would certainly cut storage requirements by avoiding duplication between different distributors, and reduce complexity, but comes with the big caveat that it would have to satisfy studios’ security requirements.

It is this factor after all that has hindered network PVR, where the idea was that an operator would hold just one copy of each piece of content in principle, even when multiple subscribers had ‘recorded’ it. But rights holders insisted that when a subscriber records a program it must be stored separately in a ‘personal locker’. However in the case of CDNs the fact that DECE is actually being largely driven by Hollywood studios should give the Common File Format a greater chance of success.

The other technical aspect of CDNs over which there was considerable concern was quality and the ability to guarantee it. A big idea of CDNs after all is to do away with the need for operators and content owners to have their own dedicated infrastructure, and so CDNs must be able to provide similar guarantees of quality. This requires some agreement over handoff between networks, which must be smooth and support specific Service Level Agreements covering bit rate, delay and jitter. Indeed the BBC pointed out that blips do occur over the links between CDNs and the ISP providing the final delivery to the consumer.

The continuing trend towards HTTP delivery for CDNs should help resolve this issue, although this too presents a temporary challenge for CDN providers, because it requires breaking video up into small chunks in an optimal way to ensure smooth delivery at the end while maximizing efficiency. Over time this should bring a big advantage in that network traffic can be smoothed out, while ensuring that all the video chunks are delivered quickly enough with relatively small amounts of caching in the network. This is because if the video is delivered in small segments, any build-up of congestion at a given router or link in the IP infrastructure can be tackled by rerouting subsequent chunks, without having to cache a large file while waiting for traffic to clear. This also reduces delay.

Given the theme of migration towards a common architecture for CDNs, it was not surprising the question of fixed/mobile convergence arose. At present there is a clear distinction between CDNs serving premium TV and those dealing with mobile video content, each having different requirements for quality. However the arrival of 4G LTE mobile services combined with the tablet boom, and the fact mobile video is growing so fast, is starting to encourage thoughts of convergence. Within a few years it is likely that the same CDNs will serve both fixed and mobile services, with a declining distinction between them.

This begs the next question, which is what CDNs will look like in a few years, and what the business model will be. CDNs will increasingly comprise the logistics infrastructure of video, with the twin purposes of containing delivery costs and maximizing quality of experience. In this guise CDNs will operate on different geographical scales, ranging from global to local delivery. The BBC would like to see a federated model where content owners do not have to negotiate directly with CDNs but operate through a broker that would obtain the best prices for the right service. This would stimulate competition but does require common standards for handoff and QoS negotiation.

The BBC has already adopted this model in a loose sense, having originally built its own CDN to carry iPlayer traffic but is now using three others as well. In practice at least two are active at a given time to provide redundancy, which the BBC sees as another advantage of the federated model. Even if CDNs are progressing better than expected, they still usually lack sufficient inbuilt redundancy to meet the growing reliability requirements of services such as iPlayer for which ‘best effort’ is no longer good enough.