What are Azure Media Services?

Announced to GA back in 2013 following some successful trials, Azure Media Services was everything you needed to serve live and on-demand content in the cloud. From ingest of VOD and linear feeds, through transcoding and encoding, encapsulation (packing) and delivery via the Azure CDN. AMS presented a set of relatively simple set of management API’s to do the job, before finally coming to rest on V3 API’s that instrumented changes to video and audio processing, FairPlay DRM handling and changes to high availability in VOD.

Why end of life?

Microsoft claims that it’s to focus on other areas of the business. Additionally, they are accelerating solutions from the Microsoft Partner Ecosystem. It makes a lot of financial sense to allow partners to deploy their own kit in the cloud and then make it available through marketplaces or services rather than trying to keep up with the complex development needed to support the media and broadcast landscape.

Funnily, this isn’t too dissimilar to the deprecation of Akamai Media Services back in 2019 – but with Akamai serving (and maybe ‘pre’ the Fastly growth spurt) a good marginal chunk of global OTT traffic, I really thought their offering had more legs given the cost savings to be had of egress from Public Clouds into Akamai’s origins.

Finally - cast your mind back to before AWS made the investment in Elemental, its own media stack was a little lacklustre in capabilities with Elastic Transcoder. Another story where it made more sense to buy or partner than to try to keep up and compete. Even with Amazon’s budgets – that’s saying something.

So how long have I got?

June 2024. Sounds like a long time that date is going to fast approach. From scoping projects around this already – I can attest to timelines being pushed to the max when the end to end cycle is factored.

What are my options?

The good news is that you have a lot of options – the bad news is that you have a lot of options

Azure Partner Network Solutions

For those that have content already stored in Azure Blobs, (and I’ll caveat this statement around other costs to change), but this may be one of the more cost-effective approaches versus extracting content out to AWS. If you’re storing mezzanine assets, the costs of egress of TB’s or PB’s of content into something like AWS could become hard to pallet. Vendors such as Bitmovin, Telestream and Dolby operate good solutions in Azure, but you’re going to have to do some work to plug them in as we’ll discuss shortly.

Amazon AWS

Possibly the best ‘cloud-native’ solution set on offer today. AWS have done a fantastic job of integrating the Elemental solution set into the house. Anyone moving from Azure has a bit of a climb ahead however – and this may come with a cost uplift to not only integrate, but facilitate the simple content migration. If you’re feeling super brave – there’s always the option to use your Azure blob under the AWS stack! AWS runs its solution set as separate stacks with some cross over in functionality – but to get the very best scalable and flexible solution, there’ll be some work to do. Built well – this will be an incredibly powerful workflow with good throughput and some great capabilities. Built badly, and you risk losing sight of where your content is, what’s happening to it and why it’s not scaling.

Google GCP

Google has been making headway in media and entertainment, and has a relatively good solution set for basic media preparation and delivery. Google own YouTube – arguably one of the biggest media transformation platforms in the world, not to mention a huge amount of delivery infrastructure – you’d think this would be the smartest bet on paper. For a freemium service generating basic HLS/DASH with some cropping, you’ll have no problems – but for anything more advanced, there are some sizeable solution gaps that require third party elements / tools / services be spliced into your workflow.

Any others?

Yes – lots. Outside of the big public cloud providers, there are a whole host of vendors providing SaaS and PaaS based media transformation and delivery tooling. Some are well integrated to public clouds via marketplace offerings, some would run in cloud really well but you’d have to deploy / configure / manage surrounding elements. I’d start a list but I’d need a shave before I’ve finished. What is important however is HOW you tackle any migration.

What do I need to evaluate before a move?

Feature gap analysis

The good news here is that there are a lot of options out there – and with a move, you’ll likely gain more than you lose in terms of pure features and capabilities – but this needs to be compared and evaluated carefully. The devil is in the detail. If you’re doing some more complicated things like mid-segment splicing, manifest manipulation, audio track re-mapping or processing HDR workflows – I’d advise potentially looking at some POC’s to confirm functionality and approach before proceeding. I can’t stress enough the element of client-side testing and HOW any content behaves in players. Variations of spec versions between suppliers catches most folks out.

Integration and migration costs

It’s naïve to believe that moving workflows and workloads from Azure won’t have a sizable impact on surrounding integrations and systems. Be this upload from MAM solutions, workflow monitoring, storage, delivery to third parties or QC. If you’re using off the shelf solutions, some of this may need change requests to third parties. The costs and time effort will also need to be factored into your overall timeline. Take care to add contingency to allow for slippage.

Cost analysis

For a true and realistic picture, any cost analysis should factor in;

• compute unit pricing,

• storage pricing,

• cost of transfer,

• cost of integrations to surrounding systems within the chain,

• cost of monitoring and operations and support,

• updates to documentation and processes

and several more.

If deploying from marketplace environments such as those provided by AWS, take note to compare compute costs AND any additional license costs that may be needed. Some marketplace offerings include software licenses within the compute instance per hour cost, others are BYOL (bring your own license). If you’re smart with your architecture and deployment, you can get your base operating cost down to really low digits if you’re not processing any content. BAU run rate will typically be storage and a few service overheads like network elements etc.

I know this is a high-level article, but hopefully, it’s a useful primer on where to point your focus if you’re in a position of needing to move. Finally, if you’d like a hand or some pointers to get moving in the right direction – I’d love to hear from you.