Monday, December 15, 2014

Building a Mobile Video Delivery Network?

Part II
Part III
In 2014, mobile video is a fact of life. It has taken nearly 5 years for the service to transition from novelty to a growing habit that is quickly becoming an everyday occurrence in mature markets. Nearly a quarter of YouTube and Netflix views nowadays are on a tablet or a smartphone. Of course, users predominantly still stream over wifi, but as LTE slowly progresses across markets, users start to take for granted the network capacity to deliver video.

Already, LTE networks start to show signs of weariness as video threatens the infrastructure and the business model of mobile content delivery.
For those who are familiar with my blog, I have been complaining for a while that mobile carriers are not doing enough to make their networks more video capable. You would think that with anywhere between 40 to 70% of the data traffic, video would warrant more interest and effort than what we see today. Many studies show that although video is the dominant and fastest growing application in mobile, its service quality is mediocre. Conviva claims that about 15% of videos in wifi and cellular networks never actually start, while Skyfire shows that close to 50% of consumers experience video problems “often” or “all the time” in the US.

Of course, part of the issue here is that 85% of these videos streamed over mobile networks are from OTT properties. In many cases, network operators and content providers are at odd when it comes to managing the service. Mobile carriers essentially see these services as non-paying passengers on their transport networks and are either looking at encouraging the offloading of this traffic or to at the very least limit the space that they occupy, particularly in congested areas.

Content providers are predominantly designing services for the internet. It just happens that some of its delivery (increasingly) occurs on mobile devices in cellular networks. The technology and economics of their service is based on the internet model, where bandwidth is plentiful and they are already paying for reach (CDNs) and access (transit and peering). Paying wireless carriers for essentially the same services was  a no-starter until a significant part of their customer based started accessing their services wirelessly on smartphones and tablets. As multiscreen and mobile becomes an important use case, content providers are downloading a streaming player into your devices when you start playing web video on your browser or are enjoining you to use their apps. These are defensive moves aimed at extending the control of the user experience. The reality today is that there are too many players with diverging controlling interests in the delivery of mobile video to make it a good experience. Soon, one will hope, the actors will recognise that no one can control the mobile delivery service end-to-end, forcing cooperation. We are starting to see signs of this with announcements such as Vodafone UK and Netflix exclusive partnership.

We are now at the crossroads where the penetration of mobile devices, the ubiquitous access to fixed and mobile broadband have redefined how video is produced and watched, but not yet how it is delivered.

What would be the attributes of a Video Delivery Network?

Well, ideally it would be designed for both mobile and fixed IP delivery. If we look first at the services it will enable and the business models it is likely to foster, such a network will need to be able to accommodate both live linear video, as well as on demand streaming. It will have to be designed to unlock advertising in a contextually relevant manner and provide frictionless compensation and service level agreement (SLA) management between the actors. Furthermore, models such as pay per use, duration passes, service vouchers, gift cards and sponsored usage will also have to be built in. The corollary from these assumptions is that, in essence, a collaborative service management method is necessary between consumers, announcers, networks and content providers.

What would this network look like, from a technology standpoint?

We have some examples today of partial implementation of these services, in a disjointed, vertical manner. Netflix has transitioned from using commercial CDNs to implementing their Open Connect network. Google Global Cache is extending the content provider’s reach into carrier networks. If we draw this trend to its logical conclusion, a well managed video network will need to have end-to-end managed quality of experience. The only way to achieve this is to integrate player/app/browser/user experience with Radio Access Network (RAN) congestion management, which itself provides explicit data to the Core network for active traffic management that is policy-managed by a negotiated SLA/QoE between content provider, announcer and network. Effectively, this would force network operators to open APIs for announcers and content providers to control the delivery of the content from a quality/speed standpoint. This is the carrier’s contribution to the bargain. The resulting quality of delivery for premium services will be a negotiation in real-time between the demand (content provider and announcer) and the supply (network conditions) at this point in time, for that service, for this user in a specific location. The quality rating at the end or throughout the session should be used as a metric in the calculation of the transfer price of the service. All this can be arbitrated and managed by SLA as it is the case on the internet today.

For freemium, free to air and advertising based services, privacy and regulatory provisions would warrant that each party involved in the ad targeting would retain the use of the data they collect and provide a geographic / demographic / contextual abstraction layer to determine the ad selection. As a result, carriers will need to fundamentally change the way data is collected and analysed, transitioning from operational to marketing view if they wish to monetize the user segmentation. The ad insertion itself should occur as close to the user as possible to enhance contextual and individual granularity. This requirement implies that for encrypted traffic, encryption as well occurs at the point of ad insertion and not before to enable targeting. Technologically, the delivery method should rely on adaptive bit rate DASH to make best use of the network resources, but the encoding should occur in the carrier’s network, with mezzanine files pre-cached and controlled by the content providers.
That ad insertion, encoding and encryption location has been a moving target in the past years because it is where the control point is from a content provider’s perspective. They have allowed CDNs in the past to perform these tasks because they had no other choice, they will need to allow carriers to perform the same to unlock this jigsaw. This is the content provider’s contribution to the bargain.

Inevitably, announcers will have to create an inventory of ads that are mobile specific, not only targeted at devices but at contexts of mobility. Measured quality, high engagement rate and hyper targeted segmentation should help raise CPM in that market.
At last, at the device and radio level, there is no reason that content that is popular would have to go all the way to the content provider’s origin servers to be delivered. An intelligent video service would be able to detect if the service requested is live and linear and watched by others in the area and switch to a broadcast delivery. If the service is on demand, but the content exists closer to the user’s location that is where it should be served from, being from someone else’s device, a network PVR or a cache in the RAN or the core network. There is where network virtualization will take its full capacity, when virtualized storage and networking function can be pushed down to the device level, peer-to-peer transmission will become possible.

What these trends indicate is that a video delivery network will need to be vertically integrated. The boundaries between devices, radio, core and content provider networks will subside, with automation, programmability and virtualization enabling the efficient delivery and management of highly reliable and profitable video service. These questions and more are reviewed in details in my latest reports "Video Monetization and Optimization 2014" and "SDN - NFV in Wireless Networks".

Originally published in The Mobile Network in September 2014.