Build or Buy: The CDN Dilemma

Online video is growing, and it's just going to get bigger. Video distribution over the Internet is rife with challenges and yet consumer expectations of QoE are steadily increasing. So how can content distributors and operators create content delivery systems that can satisfy rapidly growing usage?

A combination of broadcasters and operators addressed this issue in a panel discussion at TV Connect last week. Andy Burnett, director of online technology and operations at UK commercial broadcaster ITV plc (London: ITV), said that he was seeing more traffic online than on the linear channel for some shows, which made him think that online video was going to grow to a factor the industry can't even predict today.

Ian Parr, head of design at BT , said that 70% of traffic on BT's network is video, so it was already a concern. Now, as a sports broadcaster, video delivery had become even more important. (BT purchased exclusive rights to the European Champions League soccer for £900 million, and for English Premier League matches for an additional £960 million in January 2017. It runs BT Sport, a broadcast network and multiscreen video distributor -- see BT Sport's COO Discusses UHD.)

"You have to leverage CDNs," according to Jeff Webb, principal streaming architect at Sky , to manage this onslaught of online video. But should you work with existing CDN vendors or create your own CDN, optimized for your needs?

Interestingly, responses differed widely, even among companies you would think had similar needs. ITV's Burnett was a clear advocate of using a single, third-party provider. He said he felt no pressure to build a CDN, saying it would take a significant engineering effort to do so and he would rather use that effort to create something that differentiated the service. (See ITV's Griffiths Looks to the Cloud for Multiscreen Delivery.)

Rather unusually, ITV only used one CDN provider, while most online service providers work with at least two. Burnett conceded that there were advantages to a multi-CDN approach, in terms of redundancy and potential efficiency in some situations. But he felt that the trade-off with higher costs made these benefits questionable.

He felt that providers that put a lot of effort into working with their CDN, sit down with them and go through in great detail what is needed can create a reliable, trustworthy relationship with their CDN provider. This gives them the benefits of high performance without the associated costs and complexity of managing multiple providers. Essentially, he wanted to think of his CDN as a utility or commodity, and having just one provider was a lot easier to manage.

Sky's Webb disagreed somewhat, in that he uses multiple providers. He felt that as a provider of multiple genres of content, his services could "break" any single CDN. "Even guys like Akamai can struggle," he said. "Also, we can have problems [with a single CDN] in some parts of the country even if other parts are working."

He is concentrating on developing more intelligence at a higher layer, because he feels it's more important to have a client that can enable flexibility.

Not all of Sky's efforts are going into client development, however, as the company recently announced it is building its own CDN using the Velocix CDN product from Nokia Corp. (NYSE: NOK). (See Sky Deploys Nokia's Velocix.)

Why would SKy build its own? Webb explained that the rapid growth in video volume meant that service providers had to look at a different, more cost-efficient model for content delivery. He felt that independent CDNs had their own margins to worry about, and at some point you just couldn't push their pricing any further. At that point it made more sense to just build your own, where you don't have to worry about a third-party protecting its margins.

But Burnett saw a different solution: He felt that this was the stage to push CDNs to develop a new pricing model. He felt all service providers needed to scale rapidly now, and couldn't do so with the current model. CDN companies needed to shift to a "per bit" or "per subscriber" pricing model.

BT's Parr felt it was difficult to fix on a single model, as the market kept changing that equation. The optimal approach would vary based on the provider's specific needs and the technology and pricing options available at the time. He pointed out that BT now has a variety of approaches on its network. It has Netflix Inc. (Nasdaq: NFLX) and Google (Nasdaq: GOOG) caches embedded into its network, and also uses Akamai Technologies Inc. (Nasdaq: AKAM) and Limelight Networks Inc. (Nasdaq: LLNW) to distribute its sports service online. But he acknowledged this is a change of strategy for the operator, having built its own CDN using Cisco technology some years ago.

Weijun Lee, member of the CTO group at vendor ZTE Corp. (Shenzhen: 000063; Hong Kong: 0763), said that building their own CDN allowed operators to scale more easily, cost-effectively and deal with latency better, than if they work with a commercial CDN provider. He also felt operators could customize their own CDN more effectively.

Jamie Panagos, senior director of video operations at US cable operator Charter Communications Inc. , agreed. He felt that for Charter's needs, building a CDN was the only option. "It's IP video delivery, and it's on-net traffic primarily," he said. "It's easier to decide to build your own CDN [in that situation.] Scaling a CDN is hard; Akamai, Level 3 all have had problems."

Charter has opted for a very simple CDN, built to scale for IP video. According to Panagos, it can scale almost indefinitely for pennies, compared to traditional third-party CDNs. He also noted that the customization work Charter would have had to do for a commercial CDN was almost as much work as building its own. Panagos also said that open standards were critical in Charter's ability to develop its own CDN. It would not have been possible in the past, in his opinion.

He felt that IP video started as a companion product to mainstream television, but was evolving into the main product now. At some point, viewership would shift from the small screen to the large screen. This raised QoE requirements for online service delivery, and while CDNs were critical for this, he felt there were also other issues at stake: different encoding techniques and formats; DRM; authentication and authorization; and digital ad insertion, among others. The industry would need to take all this into account as well.

From the panelists' comments, it seemed the build-versus-buy decision was dependent more on the provider than anything else. Both Panagos and Webb talked about needing their own solution to scale efficiently, but BT's Parr is using commercial CDNs for soccer coverage -- a very highly scaled service. And Burnett talked about developing at a higher layer, allowing for more differentiation rather than concentrating on the delivery element, which he now sees as a commodity.

But as Parr pointed out, today it seems that the best solution is dependent on shifting technologies and economics, as well as the changing needs of the individual provider. For now, every provider of an online video service will need to evaluate specific requirements and in-house capabilities, and then take a decision on the right approach, while recognizing it may need to change its approach in the future.

— Aditya Kishore, Practice Leader, Video Transformation, Telco Transformation