Live from the D2C OTT Frontline: Part 2

Live from the D2C OTT Frontline: Part 2

March 17, 2022 | 3 min read
AI, Artificial Intelligence, Cloud, Coding

By Olie Baumann, Head of Innovation, MediaKind

In the first part of this blog, I explored some of the takeaways and experiences brands and technology partners might face when building out OTT direct-to-consumer (D2C) services. While the D2C market is growing rapidly, pop the bonnet on any D2C streaming platform, and you’ll discover it requires a rich, comprehensive, and often complex ecosystem of technology innovation.

MediaKind is at the forefront of D2C development, and we have proven this through our work with leading sports rights-holders, brands, festivals, and other events organizers over several years. We’re continually testing the waters on exciting and often ground-breaking innovations to help bring incredible new services to market in rapid time. This blog offers a round-up on some of these technological breakthroughs and new operational models propelling us to the frontline of the D2C battleground.

Embracing Managed Kubernetes

The shift towards managed Kubernetes services is an important change MediaKind has undergone to streamline our operational performance for D2C customers. We no longer sit on Virtual Machines (VMs) and manage the Kubernetes ourselves; instead, we have handed this process over to cloud providers. The benefits of this are significant – it allows us to spin up and deploy a cluster within minutes, managing all the back-end configuration while the customer can rest assured that their code runs smoothly in the cloud. It also enables us to deploy updates and new features rapidly without taking things offline while also scaling quickly and responding to peaks in demand.

There is a trade-off with this process in that it becomes harder to be cloud-agnostic. Not all Kubernetes services are born equal! There are also peripheral services that operate entirely differently. For example, Google’s Global Load Balancer is configured one way, while Microsoft Azure’s equivalent product, Front Door, is configured entirely differently. We need access to all these services when delivering our cloud applications, therefore we keep a core, common part running on the Kubernetes service and a cloud-specific part running around the edge. It’s not always easy to make sure that those peripheral bits perform in the same way. It’s therefore important to move towards the cloud provider’s managed Kubernetes service.

The importance of Machine Learning

MediaKind leverages Machine Learning (ML) in numerous ways to heighten our technology innovation and improve our service to customers. One way we it is to improve video compression, using ML to infer coding unit splitting for HEVC. This ensures that we make optimal use of compute resources to improve video quality which can be particularly important when deploying in the cloud where sometimes nodes can suffer from noisy neighbors. Yet the applications of ML within media extend into many areas, most pertinent to the D2C space being its ability to power metric forecasting. As we look to turbo-boost the quality of service we give our customers, we’ve sought new ways to scale our operations – and this means monitoring and managing a whole range of metrics across a series of dashboards.

To achieve this, MediaKind uses Grafana Cloud’s metrics forecasting capability to provide the observability needed to ensure our systems are consistently up and running. Previously, we had a central controller that collected all the data from the various components within the D2C workflow, as part of a Prometheus database. We’re now decentralizing this process and making all the products natively export their Prometheus data. This is brilliant because it distributes the load for metrics and data collection across all the components rather than having a single bottleneck solution.

By relying on third-party SaaS providers like Grafana Cloud, we can reduce the burden of running D2C services and streamline the whole data pipeline. We offer the ability for our customers to use Grafana Cloud to collect the same data as we use for monitoring and services. It comes down to allowing software vendors who are experts in their field – whether it be high-speed data database management or metric collection and data storage – manage what they’re experts at, while we manage what we’re experts at: media processing and distribution.

New media economics

As touched upon in the first part of this blog, one of the most significant shifts in how MediaKind operates with our D2C service customers is the migration towards Managed Cloud Applications (MCAs). The ‘cloud’ term here can often be misunderstood, as MCAs can take the form of either a managed cluster or on-premises. We can also manage them at different levels – either from a support contract, through managing the hardware, or managing all the software running all the time. There’s also a fully managed SaaS-style service, whereby we provide the infrastructure, software, and management on top.

The key here is giving our customers the choice of running as much or as little of the cloud infrastructure with us as they need. No one customer is the same. Having the agility and flexibility – and on-demand resources – is table stakes for D2C streaming service development. It’s a field rich with experimentation and observability. Working with customers who are continually changing the game and pushing the boundaries of fan engagement is a thrilling experience. As the D2C market and relevant technologies continue to thrive, fans worldwide can expect to enjoy even better, more compelling content than ever.