“The cloud is the holy grail of efficiency and cost savings,” is something you might hear said. It’s full of promise and more and more companies are eyeing it up to help simplify everything from content acquisition to distribution and delivery.

There are some very compelling use cases, but it’s not an easy transition and no one has ever managed it overnight. Content distribution is a good example, and even the world’s biggest streaming service had to work at it for a while to get it right.

Netflix started working with AWS in the cloud way back in 2008. It took eight years to migrate all its facilities into the cloud, but that same year it expanded to an additional 130 countries. Today it serves more than 190 countries and caters for more than 30 languages, and it uses the cloud for almost everything, from databases and analytics to recommendation engines and video transcoding.

It's the ultimate example of how the cloud can deliver efficiencies for Content Delivery Networks (CDNs) and is an inspiration for many content providers looking to transition away from on-prem and traditional satellite environments. 

The cloud is many things. It’s flexible and hugely scalable; it delivers robust, low latency streaming, centralized control and management; it’s geographically agnostic. But getting there takes a lot of planning, and it starts long before you start pricing up equipment and server space. 

Of all the things to plan for, the biggest thing is more cultural than technical. Although it’s also a bit technical.

1. People & Culture

Many companies dip their toes into the cloud with a good old proof of concept (POC), and most of those POCs are based around existing SDI or IP networks. While not without their challenges, these networks are in many ways a known quantity. But in the cloud, network designers need to start from the ground up, creating databases, user accreditations, stream interfaces and workstations.

And all these things are beyond the scope of most established broadcast engineers.

A skills matrix of existing staff is a good first step that content providers can carry out, but identifying what additional skills are required for the project is a better first step, because the design and implementation of a cloud solution requires an incredibly broad range of IT skills. An understanding of core video and audio theory is still very much required. But the deployment of any cloud-based solution also requires the adoption of Infrastructure as Code (IaC) and a set of Continuous Integration and Continuous Delivery (CI/CD) pipelines for a business to assure itself of speed, reliability, accuracy and cost control.

CI/CD pipelines introduce a DevOps requirement into a project and uses tooling that is found in a more traditional software engineering environment. Cloud networks need full-stack developers and cloud engineers. They need people who can ensure version control to enable them to carry out iterative changes without rebuilding the whole thing. 

Outside of the engineering teams, there are also cultural shifts across the entire company, with operational challenges to everyone from financial teams through to operations. Sensible companies will plan for significant change management across all teams to encourage staff training, fill skills gaps, provide 24x7 observability, and support and manage expectations.

But planning what you need rather than what you have, is a good place to start.

2. Costs

How many times have we heard how cloud reduces capital expenditure costs? Loads. But there are also many areas of streaming delivery that bear cost and having a good understanding of where these costs can be optimized is key to success. 

In addition to staff training and skills acquisition (see above) some example areas include:

• Bandwidth and Network considerations: data transfer in and out. Not just egress, but also source ingress connectivity using managed connectivity services.
• Security and Compliance: content protection and encryption, Digital Rights Management (DRM), regionality and sovereignty requirements.
• Technical Complexity: integration of an existing system, formats, codecs and protocols, not to mention end client support if required.
• Scalability: fixed or variable viewing audiences can greatly influence solution architecture and design, so some research into this is recommended.
• Vendor Lock-in: the modularity of a system can help to mitigate this, so adopting industry standards can limit proprietary technology becoming blockers.

3. Security

Security is a big worry for many media companies when thinking of moving from a legacy distribution model to a cloud-based system. Providers like AWS practically wrote the book on cloud security (even if it is an eBook that can only be read on a screen reader); their integrity as a service provider demands it. But there are always security risks associated with transitioning from traditional broadcast models.

Securing the content throughout the pipeline is paramount. Interception of the stream is the greatest threat, and while a broadcast production environment provides numerous security measures to make access to the streams more difficult, it is still possible.

Utilizing technology like SRT with 256bit encryption helps to maintain that upstream security posture. Further downstream, closer to the edge and to the eventual consumer, each stream subscription is fully encrypted but also tokenized insuring that only permitted consumers with integration to the token generation source can access the streams.

4. Bandwidth

With all broadcast services, the end client experience is the key to success and if streaming bandwidth is limited for that client, adopting an Adaptive Bitrate (ABR) helps by delivering lower bandwidth variants of the content and allowing the client to overcome temporal or permanent network restrictions. This is especially relevant when near real-time streaming is required, such as for live sports broadcasting, gaming or gambling.

However, from the content or platform owner's perspective, intentionally limiting the bandwidth of content leaving the cloud or CDN is also desirable because every byte of content has a direct cost attached to it. Efficient distribution encoding with formats such as h.265 or AV1 can lower the stream bandwidth and reduce cost, while at the same time improve the consumers’ experience by using higher framerates or resolutions.

There will always be a trade-off between cost, quality and latency, but it is now possible to satisfy all three factors.

5. Buffers

By definition, ultra-low latency streaming offers little or no place for buffers within a streaming pipeline in order to overcome outages or even temporary interruptions, so it is paramount that media providers carefully plan every stage of streaming, from acquisition through to consumption.

Broadcast habits die hard, and when it comes to reliable and predictable feeds an investment in redundancy is essential to overcome any unpredictability. This applies at every stage, from the origin video signal, across the network infrastructure and the encoded format itself.

Again, there is always a balance, in this case between cost and risk, but leveraging protocols such as SRT and RTP-FEC that offer resilience is key, while the adoption of high quality and robust contribution encoders for initial contribution stream generation from a baseband handoff gets everything off to the best start.

6. Conversion

While we’re talking about contributions, content has to come from somewhere, and successfully integrating the origin sources of all these signals is another thing that has to be carefully planned.

Cloud broadcast environments will always be hybrid. A format like SDI doesn't exist in the cloud, so conversion or encode to an IP equivalent video format is a necessary step to enable these signals to flow into the cloud for downstream production, distribution and consumption. And when low or ultra-low latency is a design objective, that initial encode can massively influence the downstream pipeline.

Along the way, factors such as resiliency and redundancy must also be considered for permanent contribution flows; as we have discussed, IP networks are often victims of outage. Having an alternate backup path into your production environment ensures that content delivery can still continue with the end viewer being none the wiser.

In scope

CDNs in the cloud can and do deliver, but they are very different to traditional delivery networks and not to be taken lightly. Planning is paramount and companies need to plan for every stage across every department.

But once a clearly defined set of objectives has been scoped and agreed upon, these techniques can rapidly accelerate the timeline of migration, as well as reduce costs along the way.

This article was prepared with the kind assistance of system integrators Tyrell.