The growth of unmanaged, data-heavy video on the Internet is presenting a major issue to broadband service providers.

 

Analysts say that this category of Internet traffic will grow at a 28 percent compound annual growth rate. Some BSPs have suggested a higher growth rate of 40 percent, while Cisco (News - Alert) recently suggested that video would represent 90 percent of all Internet traffic by 2013. This begs the question, “how will networks evolve to bring management to this tidal wave of media?”

 

These trends suggest that the majority of all traffic, specifically video, will be consumed from outside the BSP’s walled garden. Today it is within the walled garden that management and traffic engineering can be applied to broadband services. Traffic coming in from the Internet and CDNs is almost exclusively grouped together as best effort traffic. The looming growth numbers, however, already preclude “more bandwidth” as the answer to this growing flood of video. High-definition video is simply too bandwidth hungry to not make a dent into any network capacity planning exercise.

 

 

Broadly speaking, there are two approaches that bookend the management question. On the one extreme, the BSP can ingest all media coming from the network, transcode the input into a BSP-centric format, and distribute through existing IPTV (News - Alert)/VOD infrastructure. Essentially bringing the mountain to Mohammed. The other extreme is to revive end-to-end signaling techniques highlighted by IETF’s IntServ standard work group. In this model, each end device could signal into the network and request capacity. However, this model has failed to scale in the past and has yet to develop appropriate policy and billing mechanisms.

 

Pursuing the first approach requires that the BSP has 1) an existing IPTV/VOD footprint, and 2) that somebody has the wherewithal to map current and future OTT content into that footprint. The trouble with this approach is that OTT video does not follow any one technology. For example Amazon utilizes SSL connections to address digital rights management while Swarm Cast utilizes multiple servers to stream media snipits that converge on a player. There are perhaps a dozen or so discrete media players on the market today and they continue to evolve. If OTT video continues to grow as a market place, it is entirely plausible that the types of media players will also continue to evolve.

 

A key objective for this approach is not just to bring OTT media to the walled garden managed environment, it also to breath new life into deployed IPTV/VOD set top boxes. Many of which only support very limited CODEC types. Holding onto this footprint is important for BSPs who have a significant CAPEX investment in the subscriber’s home. However, over time this will become an insurmountable burden to carry forward. Where newer media players will be able to offer HD video at perhaps 2 or 3 Mbps, the BSP may be forced to transcode this feed up to a much higher data rate, so that the legacy devices can render the image.

 

The second approach may never be a real option, because there is simply too much shared state information contemplated in the standards, as they exist today. That is not to say they won’t evolve, but for the time being there is a lot of work that needs to done before the IntServ model can be widely deployed.

Finding the Management Middle Ground

 

There are really two objectives the BSP has to bring management to OTT video. The first is to manage customer satisfaction against ever increasing operating costs (all those bits cost money to move). The second, and more important, is to derive revenue from the delivery of this growing wave of traffic.

 

It would be incorrect to suggest that there is no hope, and that the goal of managed OTT services can never be realized. In fact there are many developments that have chipped away at the problem. At the forefront is the entire content delivery network industry. CDNs offer content providers the ability to host content in a highly distributed manner. Traditionally CDNs have measure quality of service in terms of click speeds. In other words, the fewer the router hops between the end user and the content, the faster the click speed. As a result, most content is already hosted near the edge of most major networks. So it’s not really an end-to-end equation any more. It’s an edge to end user equation.

 

The other important attribute CDNs have brought to the industry is an acknowledgement that quality delivery of services over the Internet is not free. Content providers pay significant fees to CDNs to host their services. The industry simply needs to understand there is one more player in this equation; namely the BSP.

 

If the management question can be simplified down to “what makes up this media transaction, and what does it need to work,” the CDN business model can be extended through the BSP network, providing true end-to-end QOS. Managing OTT video, in some ways has a lower barrier of precision than other managed service. By and large, OTT video already exists today, and is running unmanaged (from a BSP perspective).

 

A 90 percent management solution is a significant improvement over the status quo. As a result, it may well be overkill to have a fully check-pointed guarantee. A mode where nothing can happen until the network says so is both idealistic and arguably unnecessary. Certainly CDNs never brought that much certainty to the equation, yet they have become an indispensable facility in the creation of services (not to mention a Billion dollar industry).

 

 

The biggest hurdle for the network to participate in the management of OTT video is understanding where and what constitutes a session or transaction. For every content storefront, there are different CODECs, encoding rates, and physical media flows. Each of these parameters can fluctuate based on business decisions made by the content storefront. With an ever-growing list of storefronts, having prior knowledge of past, current, and future business decisions is a near impossibility. The only scalable solution is for the media players to indicate to the network what is happening.

 

While this sounds a lot like signaling, there is a difference between a signaling protocol and a notification message. RVSP, as a case in point will initiate a request that must be acknowledged. Once acknowledged, all media must be bound to the RSVP enabled session. This state is shared with a number of devices from the media player through to the storefront media servers. If this methodology fails, then the media play must retry the whole procedure again assuming a best effort network.

 

In contrast, a notification through the network will be picked up if 1) the network is capable of “seeing” the notification and 2) the network is able to grant the request. If either condition fails, the media player will simply act as it does today – with best effort results. On the other hand, if the network responds, the media player notices surprisingly “managed” behavior within the network.

 

There are key control points where this type of behavior can be implemented, as discussed in previous articles. Most notably would be the subscriber edge, where most if not all subscriber traffic must traverse. This is typically the first hop that CDNs interconnect with and the last layer three hop before the end users home gateway. The subscriber edge is where BSPs traditionally develop services and their associated business cases; hence the ability to develop a revenue model. Dynamic traffic engineering methods discussed in previous articles can become a powerful tool in the creation of managed OTT services when paired with the notification method described in this article.

 

 

It a world where OTT video is rapidly becoming the dominant source of traffic, BSPs face the imperative that it must somehow be managed. This article suggests a simple approach is available to BSPs that achieves the desired objective of managing subscriber expectations and operating costs, while providing the BSP with a mechanism for additional revenue. It is not perfect, but is a significant improvement over the status quo, and can be deployed without the massive CAPEX and OPEX (News - Alert) associated with alternative 100 percent solutions.