The complexities of moving to the cloud

IBC

Broadcasters and content owners are moving more of their operations into public cloud but this global trend masks a range of complexities. 

https://www.ibc.org/trends/the-complexities-of-moving-to-the-cloud/7287.article

Media organisations are moving at different paces and there is no one size fits all technical approach. While some have already moved lock stock and barrel to the cloud, others face investment, training and technology dilemmas about how best to proceed.

“Cloud is a one-way street and something broadcast CTOs have to embrace,” says Baskar Subramanian, Co-founder, Amagi. “It’s a question of what to move and how fast.”

The poster child for this is Discovery which began its wholesale move to AWS Cloud in 2016.

“Previously we’d have to buy a load of servers and install them, we’d need a file transfer system, and we wouldn’t know what the return on investment would be over time,” explains Simon Farnsworth, CTO Broadcast Technology & Operations at an SDVI hosted webinar. “Now, we can very accurately cost things like major new projects. It has become a lot less emotional and more binary since we can accurately predict cost.”

He estimates Discovery’s cloud-based supply chain has already saved the company $100m. Cloud is also claimed to have shaved $1bn in synergies from Discovery’s 2018 acquisition of Scripps Networks.

“Historically, [when Discovery entered] new territories we had siloed ops teams with siloed tech stacks and siloed workflows but we’re able to standardise that now,” Farnsworth says. “All the content for Discovery+ is in the cloud and it’s just a question of feeding it through to our own operated platforms or to affiliates. We need to be fast. What [cloud] has allowed us to do is generate the same amount of content while investing a truck load in new product.”

Comcast Technology Solutions, perhaps the largest service provider in the world, is about to make a major acceleration in moving its own supply chain (though not yet including Sky) to the cloud.

“We want to move into the cloud for flexibility and speed,” explains Bart Spriester, VP and GM of Content and Streaming. “To provide services to spin up and down and we need it to be usage based. We need to remove the integration lead time of on-prem solutions and remove capital approval cycles and slow software deployment.”

In 2020 the company syndicated 66 million minutes of content to partners like Cox and Rogers and more than 170 affiliates. “With this volume we need to take a lot of friction out of the system,” Spriester says. “We think there will be a huge benefit to moving this out to public cloud infrastructure.”

The benefits of moving the supply chain to the cloud are clear. This includes the ability to build up and down rapidly and only pay for resources when required. Enterprise scale operations can be run with greater efficiency and accuracy than before.

“Changing from a custom on-premises environment where different processes are done on different vendor’s kit to using common tools in an open source environment gives a much more consistent view of the operational state of the platform,” explains Tony Jones, principal technologist, MediaKind. “The way you build and configure systems is declarative meaning that you instruct system components what you state you want it to be and the system executes how to get there.”

It is the deterministic behaviour of systems in a cloud environment that means broadcasters can predict with far greater certainty exactly what operating a service should cost.

Add to that a microservices approach to development and deployment and broadcasters can upgrade equipment and introduce new features far faster, more economically and more flexibly than before.

“If broadcasters want to have a healthy future in competition with SVOD vendors they have got to think along those lines,” urges Jones.

One destination, many paths
These arguments may be well known but getting there is not straightforward for the majority of broadcasters. In the negative column are cost, complexity and cultural inertia.

“People are on different pathways,” says Peter Sykes, Strategic Technology Development Manager at Sony Europe. “At one end you have more traditional organisations making SDI to a IP as a first step while others are now moving to combine IP with cloud. Media companies know they have to reach new audiences but can’t increase resources and in some case are having to reduce capital outlays.”

This financial squeeze is one reason for a phased migration to cloud. Many broadcasters put their toe in first by moving disaster recovery operations. This has sped up since Covid-19 underlined the necessity for business continuity.

Another step might be to take less critical workflows like media processing and VOD to cloud. For others it makes sense to move complete sub-systems into a public cloud environment rather than a component-based approach. These systems are typically operated as-a-service by external providers.

“They could move a complete broadcast chain encompassing playout, compression and multiplexing or ABR packaging as a one functional unit,” Jones says. “There’s not really any value to the broadcaster to build that themselves but if they choose to take it prepacked it’s an operationally easier environment and there’s just one [vendor] to talk to if there’s a problem.”

The pace also differs depending on delivery technology. “The traditional DTH anchor of broadcast delivery is moving slower than OTT DTC service launches which are more likely to be cloud deployments,” says Richard Mansfield, MediaKind’s Steaming Director. “Broadcasters not ready to migrate their entire infrastructure are making this their first step.”

Arguably, the biggest issue hindering broadcaster moves to the cloud surround skills and mindset.

“The primary issue is cultural mindset more so than technology,” says Subramanian. “It’s a question of being comfortable with a particular way of doing things and a reluctance to doing things differently.”

Broadcasters used to plugging-in individual components using SDI or its IP version SMPTE 2110 face difficulties in working out how to apply that to the cloud.

“Imagine you picked a handful of vendors, one of whom deploys into AWS virtual machines, one deploys into a Kubernetes environment and another one into a Kubernetes service in a cloud provider,” posits Jones. “How you integrate that as a complete system is a nightmare and probably beyond most broadcast engineers.

“Not only do different vendor applications need to interface together but you also need to consider whether the deployment environment they work in are compatible with each other,” he adds. “A lot of legacy software apps that were built to run on premises have been adapted for the cloud but are not cloud native. There are no standards for this deployment. It’s a wild west.

“We have seen some big network operators that been able to grasp that change – but it does take quite a big investment.”

IT training needed
Related to this is the need for a whole new set of IT skills required of broadcast engineers.

“Broadcasters launching OTT services in the cloud are often doing so using an IT team,” says Mansfield. “In the long run, this separation is insane. They are essentially doing the same thing as the broadcast team but delivering to a different output medium. To be successful those teams need to be merged together as one operation.”

For Subramanian the answer lies in better education about the total ownership cost of cloud workloads. “The finance team, the operations and tech departments are all used to a capex model which, when suddenly taken to opex-driven model, catches them off guard. In some senses the cloud complicates life because there are so many different pieces of the puzzle.”

In one simple illustration, buying a server for on-premises versus putting a server on the cloud cannot be compared like for like. “With the cloud model you need to consider the networking gear, the data centre, the air con power and performance,” he says.

Aside from Kubernetes, which is an orchestration layer adopted by all major cloud platforms, Subramanian agrees that the internet is fracturing away from the broadcast safety net of unified standards. He doesn’t think this a problem. “There will be a plurality of standards that we all need to support including NDI, SRT, RIST and Zixi but this multiplicity breeds innovation.

“Fundamentally what is missing from the whole ecosystem is better education to create business models. We have seen customers cross that bridge once they understand the significant benefits.”

Somehow, Discovery seems to have done this. “We managed to flip the [internal] conversation from finance looking at the bottom line to looking at metrics,” explains Farnsworth. “How much volume is flowing through? What is the reliability like? What is the cost so we can start delivering KPIs? It’s a much more straightforward conversation and means we can concentrate on creating a better consumer experience rather than how we make it work technically.”