Test, QC & Monitoring Global Viewpoint – April 2023
ST2110 is proving its worth as broadcasters throughout the world adopt it to provide the bulk of their stations heavy video and audio shifting. But will we need to relax the timing constraints to truly take advantage of flexibility, scalability, and resilience?
Transitioning to IP requires a new way of thinking as IP networks are fundamentally packet switched topologies that rely on bursty data to optimize their efficiency. When networks were first posited, they fell into three distinct camps; circuit switched, packet switched, and time division multiplexed. Networks such as token ring were examples of time division multiplexing, whereas ethernet were packet switched, and SDI are circuit switched. Ethernet is now prevalent but relies on connected devices not using all the bandwidth of the connection for all the time.
A desktop computer connected to a 100BaseT ethernet is not going to using 100Mbits/s of data bandwidth all the time. Instead, the data is going to be bursty due to the asynchronous nature of the underlying operating system, and the standard deviation of the datarate average is often huge. However, broadcast equipment, due to the cyclical nature of video frames and audio samples, is highly synchronous and when used in its uncompressed format exhibits a constant average datarate with a very small standard deviation.
As we progress on our IP journey it’s inevitable that we must move away from the synchronous nature of ST2110 distribution that represents the highly cyclical and constant bit rate of uncompressed video. Engineers the world over have proved they can make ST2110 work reliably but they also rely on highly tuned networks with sub-nanosecond timing. ST2110 networks are to broadcasting what Formula-One cars are to motoring. Yes, you could drive a Formula-One car on a domestic road, but doing so relies on the road being highly tuned and maintained.
In effect, we need to make broadcast IP video and audio bursty, or less constant and evenly gapped, thus making video and audio distribution more efficient over industry switchers and routers.
One of the challenges broadcasters have with IP networks is that the IT solutions use protocols such as ICMP to automatically route packets and this proves ineffective for baseband video and audio. However, a new breed of network topology is appearing in the IT world and that is Software Defined Networking (SDN). SDN is incredibly powerful as the control layer and data-plane are physically separated. In the IT world, this is a major change, but in the broadcast world we’ve been doing this for thirty years with SDI, AES and analog signal switching.
SDNs are opening a whole new method of automated working for broadcasters but the real win becomes apparent when we use monitoring pods to understand the data flows and communicate this to the SDN controllers. From here, the SDN controller can provide the optimal routing for the uncompressed video and audio streams. However, the network would benefit greatly from bursty data to aid statistical type multiplexing and further optimize the packet switching.
I believe, we should now consider breaking away from the tight packet transmit timing constraints as this will have two benefits; there will be less need for evenly gapped packets that rely on custom hardware solutions, and the data will be less temporally constrained so SDNs with appropriate monitoring pods will be able to automatically determine the optimal routing throughout the network thus taking advantage of the scalability, flexibility, and resilience modern high-speed networks provide.
To make the most of IP we really need to remove some of the outdated and historic timing constraints that were defined back in the 1930s. Yes, latency must be low, but predictable latency is more important than just ‘low’ latency. And this in turn will facilitate greater intelligence in the network through monitoring pods and SDNs to further improve scalability, reliability, and resilience.