One hundred and seventy media functions, and zero interoperability.

That’s what CBC/Radio-Canada was offered in 2024 when it asked the broadcast vendor market what it was proposing in terms of software-based platforms. It proved that while there was an appetite for these kinds of workflows, the wider broadcast community wasn’t quite on the same page yet.

Software-based infrastructures are no longer a nice-to-have; it’s as if compute has finally caught up with ambition, and today they are all very doable.

But while these more flexible architectures promise scale, flexibility, and the ability to deliver more efficient usage models, it’s also clear that it is a widely held view that no single vendor can own the whole chain, which means interoperability is not optional. And that’s a much more contentious point.

The JT-DMF And The Media eXchange Format

The basis of this series is to find out more about where we are on this journey and how we’re all going to get to the destination safely. To do that we’ve spoken to five people who are all heavily invested.

All five have taken a very visible lead in the JT-DMF, a joint initiative between the EBU and AMWA that is designed to accelerate the introduction of Dynamic Media Facilities (DMF) to promote wholly software-based broadcast environments. Borne out of the initiative is the Media eXchange Layer (MXL) project, also the result of extensive collaboration between the EBU and a raft of organizations from the broadcast and IT industries.

What Are Software-Based Infrastructures?

Director, Global Collaborations / Innovation Hub at CBC/Radio-Canada, Felix Poulin is right across all of it. Not only is he the co-chair of the Technical Steering Committee of the MXL open-source project, but also the co-chair of the JT-DMF workgroup. And he has a very vested interest.

“We use the term Dynamic Media Facility in the EBU and in the JT-DMF, because when you use the term ‘software-defined infrastructure’ the focus is really on the software aspect,” he says. “But the software is simply the means to achieve the dynamic and flexible facility, not the end in itself.

“The real requirement is to achieve an infrastructure that can evolve with your business needs, where you don’t need to make major changes to your physical infrastructure every time you have a new idea or a new project or a new way of producing content. That’s the business agility, or rather the digital agility, that we as infrastructure providers need to deliver to media businesses. To achieve that dynamic media facility, we need to adopt software-based, cloud-fit architectures using standard IT hardware and technologies at the bottom of the stack, and then really concentrate on media applications running on top of that IT infrastructure.”

Launched in 2021, Lawo’s cloud-native HOME architecture has been exploring this flexible software-first concept for some time.

“The broadcast industry has been very used to selling proprietary boxes to do specific functions, and software-based infrastructure is about using the physical underlayer to perform a number of functions,” says Lawo CTO Phil Myers. “The real benefit is the ability to overlay across that physical infrastructure anything you may need now or in the future. Software means different things to different people, but it’s really about functional blocks and abstraction: decoupling function from a physical piece of hardware.

“If you look at a traditional TV studio, there will be hardware and software there anyway, but they’re most likely coupled together; a box with a piece of software running on it doing a specific function. Where we want to go is to remove that coupling, have the software layer sit on top, and let the user decide what they want it to do. And when they’re done with it, convert it into something else.”

Across the board, nobody on the panel disagreed with this concept, although Matrox Video Product Manager Daniel Robinson expressed a note of caution concerning how these concepts are presented. These aren’t just semantic differences; they reflect different strategic framings, and whatever you call it will shape what you build. And with the introduction of any new technology, these things matter.

“Software-based infrastructure is an interesting term, but personally I think it is an overloaded one,” he says. “People often think of software-based infrastructures as something like AWS, but to me an actual infrastructure typically means a network switch, a computer, a virtual machine…a physical infrastructure.

“In this context, what we’re really talking about is software modules that combine to build a workflow for a customer. To me, it’s about replacing what were traditional devices and connectivity – like SDI, Character Generator boxes and playout servers – with software modules and connecting those together using deployment or configuration tools like Infrastructure as Code (IaC). While the term ‘infrastructure’ can be somewhat overloaded, in this context it’s very much about that shift from traditional to pure software.”

A Natural Evolution

With more than 70 years in the broadcast industry between them, Riedel’s Chief Executive Officer (Product Division) Jan Eveleens and Appear CTO Andy Rayner both feel like this is a natural evolutionary step.

“It used to be that broadcasters defined where the world was going in terms of AV technology, and for a while the consumer industry defined that,” says Eveleens (Riedel). “Now I think there is an incredibly complex environment with new entrants, sports rights owners and professional sports people all vying for attention and influence. It’s very volatile.

“For me, the development of software-defined infrastructures is part of a very long journey towards a new future. I’ve seen things develop from analog to digital to IP network-based systems. The next logical step is everything running on general compute where you can run processes in a public or a private cloud. It is inevitable that more and more things will move to general platforms and shared infrastructures because these systems bring more flexibility and better utilization. It is the reason why our clients are looking at more and more software-based infrastructures.”

Meanwhile Rayner (Appear), a SMPTE fellow and a Director of VSF in addition to his role at Appear, wholeheartedly agrees, demonstrating how the industry has learned to leverage IT infrastructure incrementally in a way which makes the current developments feel earned rather than sudden.

“Over the last 15 years there have been several attempts at making this happen; these attempts have failed or haven’t gained any momentum because they’ve relied on bespoke hardware components to make it work,” he notes. “The reason this is happening now is because the compute and compute fabric infrastructure is capable of supporting what is needed for the live production chain.

“The broadcast industry has never been very good at fully leveraging wider IT industry infrastructure, but with ST 2110 we managed it with switch fabric, and switch fabric manufacturers rose to address issues like PTP propagation and buffer depth.

“Today we have a system in which Ethernet switch fabric is fully capable of supporting everything we need, even in a fragile UDP world. Today, the compute itself is at a point where not only are the CPUs, GPUs, memory and bus throughput able to handle what we want, but the underlying technologies that deliver high speed and guaranteed data transfer in a time bound manner are all in place. So it’s a perfect time for us to be realizing applications that can sit on top of standard infrastructure without having to be bespoke engineered.”

The Economics & The Real-World Case

So how far along are we with leveraging IT as a core part of the broadcast stack? Although there’s an awful lot of talk about it, not that much of it seems to be going on in the real world, but it can help broadcasters to hedge their bets in an uncertain environment. It’s already happening in Toronto.

“A few years ago we refreshed our Montreal Broadcast Center with a brand new building and took the opportunity to adopt ST 2110 technology throughout,” says CBC’s Poulin. “It gave us greater flexibility because with ST 2110 you get rid of specialty connectivity, pool resources and enable anything to talk to anything. During the Olympics it was straightforward for us to repurpose devices for special workflows and then route back to business as usual without rewiring.

“The use of software was limited but it was a good first step into full flexibility. Our Toronto project is very different. It’s not a brand new building; it’s a brownfield project so we need to refresh the technology while we remain on air and do it progressively over many years. The media business is evolving so quickly at every level that it’s very hard to create an eight-year plan and understand where the business will be in eight years’ time. The idea is to go step by step, and to do that we need the agility that DMF and software-defined architecture gives us. We can start small and we can iterate; we can grow the system and update it with new components as they become available.”

Growing and developing the system is one thing, but another advantage of software-based technology is that abstraction becomes a tangible benefit. Cloud compute means the infrastructure no longer has to be physically with you; it can be located in a data center or in a public cloud, but wherever it is, it provides flexibility.

“The main benefit to broadcasters and content owners is that they don’t have to go through a crazy CapEx investment process of investing in very costly and significant hardware infrastructure every few years,” says Appear’s Rayner. “The beauty of having a software infrastructure is that you only need to build to a default level of usage, and you can flex your peak demands on top of that with other compute infrastructure.

“The mindset a lot of broadcasters are moving to is to self-host their base load of production compute, but anything over and above that, including any events-based content, will use outsourced public compute because they don’t need it continually. There’s a massive difference in content production between the 24/7 channel stuff and the levels of sports production which are only being utilized one or two days a week. In this environment, the bulk of the traditional hardware infrastructure is sitting there doing absolutely nothing. Look at the level of sports and light entertainment on a Saturday compared with the rest of the week – why invest in all of the kit you need just to run Saturday afternoon and evening? That idea of being able to flex is incredibly attractive.”

This agility is even more relevant in today’s fast-paced broadcast environment, and it has additional benefits that go beyond flexibility.

“The benefits are clear; it reduces cost,” says Riedel’s Eveleens. “It’s not simply that the upfront costs are reduced; it also helps protect against changes in requirements over time. The world is more dynamic than it was 20 years ago, which means that even if you make a significant upfront investment that you can financially justify at the time, that justification might go out the window if something changes.

“A software-based infrastructure that relies on subscription models enables users to change or add to it, and in that way it’s much more flexible. None of this would have been possible even 10 years ago, because the compute was not there. The compute was not fast enough, but modern GPU and CPU power, memory size and bandwidth on interconnect, are all enabling these developments. If we could have done this 10 years ago, we would have done it, but at that time it was simply not possible — it would have been way too expensive.”

According to Lawo’s Myers, this way of thinking also has a knock-on to the total cost of ownership (TCO); as the normal buying pattern for broadcast customers is defined by project, that under-utilization of assets is common, but it’s not just the capital cost; it’s the running costs too. Power, electricity, and the SLA costs of maintaining its health over that period are all hidden factors.

Agility Under Crisis

Crucially though, it also gives you agility under crisis.

“None of us sitting here in 2019 would have anticipated what came in 2020 with the pandemic,” he says. “We all saw how organizations had to move to different models and reconfigure their infrastructure at a time when not only could they not easily buy equipment, but the world was effectively locked down from a trade perspective. A lot of companies had to cobble things together just to stay alive. One of the big benefits of software is if you need to reconfigure to do something else at the last minute, it’s easy to do. You don’t have to relocate physical equipment and you can instantiate software wherever you need it, whether it’s in your own building, in another building, or in the public cloud.

“With software-based infrastructure, you don’t need a huge workforce to physically move things around and reconfigure devices because you can use automation to define what you want and go and build it. Not only does that give you scaling, but it gives you consistency – very important when building large-scale infrastructure – and it gives you documentation. I’m always very clear to say that this is not about replacing people. It’s about using new tools to help them do their jobs more efficiently, more consistently, and in a more sustainable way.”

But Robinson (Matrox Video) added a note of caution which runs contrary to the narrative from many vendors in the broadcast space; just because you can move to a software model doesn’t suddenly make everything cheaper.

“In the past if you built a hardware box you had to use a waterfall design methodology, and after three years of R&D you’d release it and hopefully someone would buy it,” he says. “With software, you can iterate and adopt much more agile development and the net benefit for customers is more rapid feature development. More generally, being software-based allows for much greater flexibility; you can run your software in the cloud or on-premises and it allows for better scalability because you can quickly provision new software and your limiting factor is compute, which in the cloud you can just spin up on demand.

“The real cost in broadcast equipment these days isn’t the equipment; it’s the software license. Just because you move to a software model doesn’t suddenly make everything cheaper, because in reality the value has always been in the software. But having the ability to move and transfer licenses flexibly means you’re not tied to a hardware dongle that has to be plugged into the back of a machine.”

The Walled Garden Question

As Poulin (CBC) has already observed, “It’s typical that with new technology you start with proprietary solutions that prove the concept, and it’s faster for companies to develop that way. But at some point you reach a limit in terms of what you can achieve in a single-vendor environment.”

This is exactly what happened when Poulin asked the market what it had to offer in terms of its software-based functionality. CBC didn’t approach its tender with any preconceptions about walled gardens, but the reaction from the market was clear that it was an issue.

“We had a list of around 30 media functions, like processors and mixers, that we needed to cover,” he adds. “The market came back with around 170 different media functions, but there was no interoperability between any of the vendors. We couldn’t just pick from all of them because we needed to choose one ecosystem, and we could also see that not one of those vendors had everything we needed.”

But that’s not the way any of our panel want things to go – everyone is agreed on that, and everyone supports openness and interoperability, even if everyone’s approaches are not always the same.

“The universal cry from end users is that they want an open and interoperable future,” says Rayner (Appear). “As an industry we have engineered that in the hardware world with SDI and ST 2110 and lots of other bits in between. But at the moment there are a couple of vendors that are creating proprietary walled gardens; they are trying to pull other vendors in, but trying to do so in a way that subsumes other vendors and closes them into an ecosystem, which doesn’t work for the end user.

“Where I think we have to get to as an industry – which is very much the DMF vision – is for end users to have the ability to deploy different software components from different vendors on their compute infrastructure, irrespective of whether that is a private or public compute infrastructure. Because no single vendor is going to be able to provide the totality of what’s needed.”

It’s a compelling argument. But achieving it is far from frictionless, says Robinson (Matrox Video). Simple things need to be thought through too; like who to call when things go wrong.

“Multi-vendor solutions always sound extremely appealing to a customer. The idea that you can swap one vendor out for another and everything just works is an amazing idea, but it raises a real question: who is ultimately responsible?” he asks.

“Customers want to be able to pick up the phone and speak to one person when there’s a problem. What they don’t want is vendors pointing fingers at each other. As we move more to the software world, I think there’s more work to do on orchestration at a platform level.”

It’s things like this, things like the accountability of these new and dynamic environments, that will take time to work out. But there seems little doubt that it’s the right way forward and gives customers the most options to pick the technologies that best suit their needs, and the ability to build on what they already have.

“Firstly, the world will not be a better place if there is one dominating company or entity that delivers all the infrastructure,” says Eveleens (Riedel). “That approach stifles innovation, but it is also not cost effective for end users as there is no real competition. Secondly, there is nobody in the industry large enough to be able to do that.

“We will always be in a situation where there are multiple vendors involved because not only can nobody deliver a full broadcast workflow, but customers also have legacy products. For all that to work, we need to have some agreement around open standards to ensure that we can all work together.”

That’s not only very doable, says Myers (Lawo), but in the IT industry, it’s already being done.

“AWS, Google and Microsoft are each ecosystems in themselves but they provide very strong capabilities to interconnect and allow multiple vendors to come together,” says Myers. “What we should avoid as an industry is telling vendors they have to come and live inside one particular ecosystem. Vendors can choose to participate in an ecosystem if they want to, but we should be allowing ecosystems to come together.

“The approach has to be open APIs with interfaces that don’t only work with ‘friendly’ vendors. It allows ecosystem vendors to innovate and address functionality that you can’t get in other ecosystems, and it gives customers maximum choice – not just one vendor, but the ability to choose components from different ecosystems with the reassurance that they can be interconnected.

“If the industry takes a position of closed walled gardens, what happens if there’s only one walled garden left? That single vendor would then need to accommodate every possible user and customer requirement, across wildly different products, regions, and workflows. It’s just not sustainable.”

The Commercial Model

While there is consensus about motivation there remains an unsettled note, as well as an admission that the industry hasn’t yet solved the commercial model. And it’s one which might prove an even greater challenge.

“At the JT-DMF kick off meeting there were four tracks, and by all accounts the business track was by far the most interesting,” says Robinson (Matrox Video). “There needs to be a healthy ecosystem for vendors and for broadcasters that is mutually beneficial. Vendors need to be able to make money to deliver good solutions and support their customers, but there’s obviously a demand from broadcasters to be more flexible.”

Rayner (Appear) does not disagree: “There is a real question on the economics and the charging methodology that vendors and users expect when they move into software-based environments. The end user expects some cost savings because they’re not running services all the time, but the vendors are still using development resources and they still have business models they have to fulfill.”

While everyone agrees on what the benefits of software-based workflows are, it seems we’re not there yet.

Next, we look at how microservices and cloud compute infrastructures adopted across other industries are not based on standards, and whether the media industry should try to form standards as part of its own adoption pathway.