How Akamai’s CIO saved $100 million annually by bringing public workloads back on prem

When Akamai CIO Kate Prouty decided it was time to get her cloud infrastructure bills under control, she knew it would be a tough sell, especially for the engineers. The last thing engineers want to do is look under the hood to find savings. That’s her job. They just want to build stuff, and she understood that.

That’s why she decided to look at the problem from both a technical and change management perspective. By eventually making finding cost savings a kind of game where engineers saw controlling cost as innovative problem solving, she was able to sell an enormous change – no small matter in a large organization – and reel in $100 million in cloud operations cost savings annually along the way. This is the story of how she did it.

The story begins with an acquisition. For most of its life, Akamai has been a content delivery network. That means it moves content close to the user and delivers it faster than if it were on a server somewhere further away. It was so effective, the company decided to get into edge computing in a more general way, moving workloads of all types closer to where they were needed. As part of that approach, it bought Linode in 2022 for $900 million.

With Linode, they got a much loved edge computing service that provided a way to deliver edge computing services without paying the hyperscalers huge bucks, The company soon learned, however, that what worked for customers in terms of reducing latency and cost, could also work in house, and a project they dubbed Cirrus was born. It was time to think about moving some of the workloads sitting in public clouds back in house.

Bringing it all back home

While Linode was a key catalyst for the company’s transformation, Akamai did not buy Linode to improve internal operations. It did so because it saw a commercial opportunity to expand its edge computing approach.

“We did not acquire Linode to cut our costs,” Prouty told FastForward. “We acquired Linode as a pivot for us to move into this adjacent market that just made sense for our business and made sense for our growth, and is a massive opportunity for Akamai.” 

But over time, they began to see Linode as a solution they could put to work internally too, and save some money in the process. “There was a huge opportunity here for us to rethink the way we leverage third-party cloud services,” she said. That led to them start thinking about this as a holistic, strategic idea where they could take advantage of this asset they had purchased in other ways than they originally had envisioned.

That started some internal conversations about how the company was leveraging public cloud services, and where they might have opportunities to potentially move workloads back on prem. Once they realized there was an opportunity to cut costs by decreasing reliance on public cloud vendors, they went to work.

At that point Project Cirrus became like any other internal project with management, structure and governance. They began to assess every workload and if it made sense to leave it in the cloud or move it on prem.

Evaluating the workloads

Prouty began assigning teams of engineers, who would then report into a broader structure, keeping those teams small enough to run efficiently. As each team found areas for savings, it would report to a central program manager. What these engineers did was assess the workloads and if they could save money by repatriating them. 

“And there were places that it became very obvious where we had workloads that hadn't been evaluated or looked at since they were deployed,” she said. And in a lot of cases, they found that it would be more efficient to move then back. 

There was a huge opportunity here for us to rethink the way we leverage third-party cloud services.

But moving out of the public cloud vendor’s environment required a fair amount of evaluation and coding, at least at first. But once that code was written, it got progressively easier to move to the next stage of workloads because engineers began to figure out best practices and share them with each other.

Solving change management problem

While this made sense from a cost and efficiency perspective, Prouty was still dealing with human engineers, who might have seen this as busy work. “Initially, they thought they should be innovating and driving the next product, and not just looking at basically moving workloads around,” Prouty said. But she was really able to flip the script by framing the task a little differently.

“We were able to turn this begrudging process into one that engineers were really excited about and really excited to be a part of,” she said. They got them asking questions like why are we using this hyperscaler and these microservices and what we could build ourselves. “And that really got engineers thinking and got them back to their engineering roots and solving problems that were really interesting to them,” she said.

They also instituted a bunch of training, and got engineers involved in code reviews and looking at strategies and making sure that the architecture that engineers were using was really leveraging cloud computing in the most effective and efficient way possible.

Prouty acknowledges that sometimes it made sense to keep a particular workload in the public cloud, but that has generally been an exception instead of the rule it had been previously. It has really set up her company for whatever comes next, giving her the flexibility to control her own destiny, a position many a CIO would envy.

“We've set ourselves up now for the ability to evolve and pivot as we go, whereas other companies are just still really locked in, and that’s a dangerous place to be," she said.

Featured image courtesy of Akamai