I was talking to the CTO of an IBC exhibitor today, and he came up with the expression “file-based 2.0”. The point he was making was that we have been through the first phase of moving content to files, where we did little more than replace a video format with a data format.
File-based 2.0 is about using content files as files. It brings in topics like virtualisation for efficiency, and it moves us towards using IP for realtime delivery as well as shipping fixed files from place to place.
The realtime issue cannot be underplayed. In the broadcast business, we are very difficult customers. Basically, we want a new picture every 40 milliseconds. Not most 40 milliseconds, but every time. Not roughly 40 milliseconds, but precisely 40 milliseconds.
The original purpose of IP was to make sure the file got through, and it is brilliant at that. But we want more than that: we need it to arrive on time. “’Best effort’ is not good enough,” Brick Eksten of Imagine Communications told me. “To get the critical timing we need at the frame level, we have to be critical at the packet level.”
If we can crack the timing issue, then we have the prospect of handling all content – realtime streams and packaged files – over an IP network. That first means we can use commodity IT hardware for much of the routing and processing, which will transform the economics, not just in capital expenditure but in hard costs like power and air conditioning.
It also means we can use IT techniques to manage the workflows. Imagine Communications is talking a lot about software-defined. This does pretty much what it says: rather than build a fixed, rigid workstation, you can have an architecture which is as flexible as you need it.
Virtualisation is a great example. One installation of IT hardware – typically blade servers – can run many different processes, shifting from one to another as the workload increases and priorities demand. It means hardware is not lying idle – but still consuming standby power, rack space and cooling – until another job for its task-specific application comes along.
That, in turn, needs management. Some process has to decide which tasks will go on which processors in which order. You need a virtualisation management layer.
But it does mean commodity off the shelf hardware is going to be important. Most standard processes, along with the signal routing and virtualisation management which Software-Defined relies on, will be on IT hardware. It seems clear to me that this is inevitable.
“We took a long hard look at this business,” said Imagine Communications’ Eksten. “We want to be clear what we can bring to the game. And it is clear that companies like HP, Cisco, and Juniper have been doing this for years. So we partner with them. They do the physical side, and we layer the management, the control plane. We beat the operating system into submission.”
We need deterministic ethernet so we can rely on the timings we need. We need an excellent virtualisation layer, which is going to provide the elasticity broadcast processes need. And we need a transition path, because no-one is going to move from one platform to another overnight. With those three in place, Software-Defined looks like playing a central role in the media world of the immediate future.