Hardware Infrastructure Global Viewpoint – September 2021
Industry has been using metadata to streamline workflows since Henry Ford built the first Model T, all that has changed is the amount of data available and our ability to process it faster and more efficiently. With audiences making ever-increasing viewing demands, the need to gather, process and understand metadata is more important now than ever, especially if broadcasters are to keep their churn rates under control.
One of the challenges we face in modern broadcasting is that viewers have more choice now than ever. They don’t even need to reach for the remote control as their viewing devices are often in the palm of their hand. Although they may be watching their favorite programs on the move, or in a congested café, they still demand the same quality of service they are used to from static linear home televisions.
Internet delivery is bidirectional, that is, the broadcaster can receive data from the viewer when streaming programs allowing them to understand very quickly who is watching their service, when, and how. They can determine quality metrics such as bitrates and network congestion and see when viewers have switched to another service.
Going back through the delivery chain network probes gather a wealth of information about how the network is performing. The increasing TCP round trip delays indicating congestion and switch buffer blocking can all be measured and used to improve the system.
Moving back to the broadcaster’s infrastructure we can determine equipment failure and quickly find network breaks. Disc drives, power supplies and motherboards can all be monitored, and their replacement accurately predicted.
However, all this data is meaningless unless it’s used coherently. There is no point in just looking at the datarate of a network probe in isolation to determine the reliability of a network link. Unlike synchronous SDI and AES systems, where the bit rates are highly predictable as they operate in closed networks, data throughput in IP infrastructures is highly dependent on seemingly unrelated agents.
Misbehaving TCP or UDP processes could easily flood a link depriving another service user of valuable bandwidth. Or a DASH manifest file could be incorrectly populated by third party service leaving multiple viewers without their programs, even though the media streams are working perfectly.
The point is the viewer doesn’t care. All they know is that their service is failing, and if it happens too often then they can easily vote with their feet and switch to another provider.
Metadata is one of the most valuable tools modern broadcasters have at their disposal as it conveys deep insight to how a system is behaving. Not just individual components, but more importantly, the whole system.