Grass Valley, a name synonymous with broadcasting hardware for 60 years, epitomizes the software revolution overtaking video infrastructure and especially playout, with the claim that it is at least holding onto traditional customers as a result. One of the latest such repeat orders came from Pike Media Lab, a provider of playout services to cable and OTT operators in the Russian Federation primarily. This was for its ICE SDC software-defined playout platform, which inherits the code base of the hardware appliance ICE with the same integrated features. It supports uncompressed 2022-6 video streams and enables the Russian media services provider to deploy new operational models more readily than with the existing hardware playout system.
The immediate benefits of the Software Defined Networking (SDN) approach are greater flexibility and lower costs through avoiding need for dedicated hardware, instead being able to use COTS (Common Off the Shelf) components, which can also accelerate deployment of new features or services. But it also brings challenges, such as security risks associated with having less knowledge of what is going on beneath the bonnet and reliability, given that hardware-based playout systems had already surpassed the five nines level, meaning on average less than five minutes downtime a year. The likes of Grass Valley have worked to bring their SDN levels of security and reliability up to at least the same levels and Pike Media Lab at least is convinced that has been achieved, given that it is offering the technology to its customers.
One advantage for broadcasters is that they have come to the table quite late compared with other sectors of enterprise computing and so SDN techniques are now quite mature and well proven. That being the case, we can agree with Grass Valley’s assertion that most broadcast infrastructures will inevitably migrate to SDN technologies over the next few years, allied with the related Network Functions Virtualization (NFV).
SDN and NFV are sometimes confused and they do often come together as distinct yet complementary parts of contemporary virtualized cloud infrastructures. NFV is the network equivalent of traditional virtualization, separating higher level functions form underlying hardware and the immediate operating platform. NFV allows traditional hardware components such as IP routers, load balancers, firewalls, intrusion detection devices, Deep Packet Inspection (DPIs) and traffic accelerators, to be implemented on general purpose COTS hardware. It uses established server-virtualization techniques already proven in enterprise IT, but goes further by allowing the functions to be decoupled more and distributed loosely across a cloud computing infrastructure. In this way, it divorces the functions further from the original notion of having dedicated hardware appliances for each network function.
SDN then operates at a higher level of abstraction by separating the overall control of a network from the data plane housing those specific underlying functions. The SDN then avoids applications from having to be aware of underlying network functions at all, which are accessed through function level APIs. The functions can and increasingly are provided via NFV but could be dedicated devices. So SDN does not require NFV and similarly NFV can be deployed, as it often has been, ahead of SDN to take advantage of COTS hardware.
SDN also provides a convenient layer of management which can help with deployment of network monitoring and data collection, acting as an intermediary between the network and the key applications in analytics, recommendation and quality assurance. The architecture can also help reduce latency by allowing functions to be relocated to minimize length of data paths or the amount of processing entailed in their execution. This is achieved by divorcing functions from the underlying physical infrastructure so that workloads are defined by virtual addresses. This of course means that tables associating these virtual addresses with underlying physical locations must be kept up to date.
SDN and NFV evolution has also been driven by their growing use in cellular communications starting with 4G, where it has simplified telco data centers and allowed use of more commodity servers. Recent extended NFV concepts such as CORD (Central Office Re-architected as a Datacenter) allow telcos to address the latency problem by migrating some of the core functions to their customers’ premises, running on the infrastructure there to cut data transit delays.
In the case of playout, the software architecture has to replicate the original chain of hardware devices that typically started with content stored on video servers, culminating in output either as Serial Digital Interface (SDI) or IP. For a standard TV channel this would typically involve a video router to switch live sources, with Audio Overs for mixing in voice overs or announcements.
The playout is also separated into contribution and distribution. Contribution feeds, typically live feeds from events or studios, have firstly to be inserted into the playout process, with the primary requirements at this stage being low latency so that the pictures remain as close to live as possible, and then of course high quality. The preferred method therefore is an uncompressed feed which is simply encapsulated into IP for transport, using either SMPTE standard 2022-6 or increasingly SMPTE-2110 following final ratification.
Then there is the output stage as content is delivered either directly to the consumer, or to other parts of the organization in remote locations. Here bandwidth requirements dominate, with H.264 often still being used although being superseded by more advanced codecs such H.265 or AV1. The tradeoff here is between bandwidth and processing, given that uncompressed video places the lightest load on CPU resources, while heavy compression consumes more CPU but reduces the distribution bandwidth costs.
The SDN must replicate these processes, whether in an external data center, cloud or within the broadcaster’s own facilities. In each case it should enable cost efficiencies through the use of COTS hardware and exploit the whole IT industry’s continuing investment in R&D, which will ensure the channel platform footprint becomes smaller and more flexible. As this migration proceeds, consideration must be paid to the broadcast output, which will remain the principal source of revenue for broadcasters and content owners for the foreseeable future. As Grass Valley points out, a major benefit that then follows from the dynamic nature of the data center is the ability to launch new services like event based or pop-up channels. Additional services can be built around major sporting events and then taken down afterwards with little capital investment or upheaval. This of course all assumes that the right choice of technology and cloud provider have been made, because SDN/NFV is not a magic wand, but more a conductor of the underlying infrastructure.