SDV or Software Defined Video looks like being one of the top buzzwords at this year’s forthcoming IBC in Amsterdam as it was to an extent earlier at NAB 2014. It is the latest concept to enter broadcasting from the general world of IT where it is known as SDN (Software Defined Networking), being closely related to cloud and virtualization. It is associated with commoditization both of underlying hardware and the lower level networking functions of routing IP packets from source to destination.
For video especially its objectives include reducing hardware cost by replacing dedicated ASICs (Application Specific Integrated Circuits) and FPGAs (Field Programmable Gate Arrays) with cheaper generic processors that enjoy greater scale economies, and making the whole infrastructure more responsive to changes in traffic levels, while also being simpler and faster to upgrade for new services or extended range.
The concept is particularly well suited to broadcasting, which has traditionally relied on expensive dedicated hardware and has the exacting requirement for video of low latency and high bandwidth. As video services migrate to IP based IT infrastructures they need a new layer of control to ensure QoS. By separating overall network control from lower level routing it is easier to manage video traffic effectively at the service level and prioritize video streams when and where necessary.
Some infrastructure vendors have been ahead of their time in anticipating the coming of SDV, such as Sweden’s Net Insight with its Nimbra MSR (Media Service Routers), which effectively separate video traffic control from routing, with the result that congestion occurs less commonly and can be anticipated before it occurs.
Yet it is the software encoding vendors, such as Envivio and Elemental, which have been trumpeting the SDV label in association with their own products. Ericsson has been muttering about it too in relation to its Virtualized Encoding, a software package that can be implemented on a combination of dedicated programmable hardware, like its own video processing chip as well as GPU (Graphical Processing Unit) based servers, that are on premise or deployed in the cloud. This then is rather a hybrid in still allowing use of dedicated hardware, but does have the software abstraction layer and the ability to allocate resources for encoding on demand wherever they are.
Ericsson is actually right in suggesting that SDV is not about whether hardware is generic or dedicated, but the ability to use whatever is available in a flexible way on demand. SDN in the IT world is not considered synonymous with generic hardware, but it is true that is the direction most processing is going. That is because generic chips from Intel and others are becoming much more powerful and tending to embrace the processing building blocks needed for most applications.
This trend began in the 1970s with incorporation of the first math co-processors in general CPUs to handle floating point arithmetic and has continued ever since. While there will always be some specialized tasks best done in dedicated silicon, for most cases the only advantage of ASICs and FPGAs may come to be greater power efficiency. This incidentally is the only concession given to dedicated silicon by the likes of Envivio, which claims that even this gap is narrowing through development of intelligent on-chip power management regimes which enable silicon compartments to be shut down when the associated function is not required.
The general case for SDV made by Ericsson is that it ‘eliminates complexity, enables more efficient resource utilization and bandwidth management and bridges the gap between broadcast and IT infrastructures, allowing deployment of enhanced TV services much more quickly.’
Elemental’s Chief Marketing Officer Keith Wymbs makes the important added point that SDV is a key enabler for multiscreen because it allows the complexity of dealing with different target platforms to be offloaded and separated from the overlying service. Wymbs naturally argues that SDV brings scalability, since responsibility for providing hardware capacity on demand is offloaded to an external infrastructure provider, and for the same reason enables upgradeability, although the latter also relies on the appropriate technologies being supported.
The Elemental argument then is that SDV frees video providers from the constraints of dedicated equipment by enabling the best architecture and processor combination to be used for a particular application, evolving as requirements change over time. Support for new services and video formats can be integrated relatively painlessly almost through configuration in software, without any hardware changes being required, by the operator at any rate.
Yet if this all sounds too rosy, there is an important caveat here, which has applied ever since businesses relied on external providers of utility computing in the 1970s, through the whole era of outsourcing right on to the contemporary cloud. It is that the customer, whether this is a broadcaster, content owner or pay TV operator, is reliant on a third party for the infrastructure, who has to be trusted to have the scale and agility to ensure that peaks in demand can be met, and the technological capability to cope with the complexities of multiscreen delivery. Just because a function is virtualized does not mean it disappears into a black hole ‘ it still has to be performed.
Nevertheless in practice most operators will be able to rely on their SDV partners providing they perform due diligence and then they can enjoy the benefits, which include being able to respond on the fly to changing consumer demand and being able to implement new features through upgrades and API integration of third party software. Avoiding having to wait for new custom video chips to be released to market is also a big potential advantage, since operators and broadcasters can then continually update and enhance their platform in line with emerging consumer demands and technology advancements, taking advantage of revenue generating opportunities as they arise.