Public travel systems provider Amadeus opened for business way back in 1987 when four airlines wanted to offer a centralized booking system. Today, the company facilitates booking and inventory management for 216 airlines, as well as hotels, trains, airports, online travel companies and even corporations. In short, it covers just about every aspect of travel IT imaginable.
As an IT-heavy company scales, its computing, software and systems requirements change as well. For Amadeus, those pressures came to a head in 2015, when the company realized it needed to move beyond the private cloud it had been running. Embracing a public cloud infrastructure would give the organization greater flexibility to react to customer needs much faster than it could with a homegrown tech stack.
The challenge Amadeus faces isn’t uncommon: Like many companies of its size, it has to pay off massive technical debt, which is the idea that as you make certain technology decisions, you end up pushing modernization down the road, and the “interest” builds up, making it harder to modernize. Eventually, competition and wider market forces pressure you to pay off that debt, and interest, by updating your entire code base, infrastructure and even the way you produce software.
It’s not just giants like Amadeus that face tech debt. Even much smaller companies have to deal with technical debt eventually and come to terms with massive technology changes.
To learn why and how Amadeus has begun such a dramatic change, we spoke with Fredrik Odeen, lead for public cloud transformation and corporate strategy, and Sébastien Pellisé, deputy lead for public cloud transformation. Odeen and Pellisé also explained how the company went about implementing the new processes and tech stack, what the changes mean for the future of the company and what smaller companies can learn from Amadeus.
Why tech debt accumulates
When you launch a company and make key decisions about your technology stack, you probably don’t stop to think about the impact those decisions might have on your company in the future. You want to get the business going, and most early-stage companies don’t have the luxury to consider their tech stack all that carefully.
Over the years, Amadeus slowly began to realize that there was a downside to running its own infrastructure. Its homegrown cloud stack was causing a huge lag in its ability to update its software and systems, react to customer needs and experiment. The company knew that to improve its customer experience, it had to update its systems and how it produced software, or eventually these problems were going to catch up.
The pandemic drove that point home. As the industry suffered from big cut backs in travel, finding ways to operate more efficiently, and reacting to customer needs faster became even more crucial — the company needed to find a way to save money as revenue dried up.
A company the size of Amadeus can’t make a change of this magnitude easily, or even all at once. It takes years to implement and gradually move from old systems to new ones. Given how much tech has advanced since the company launched in 1987, it’s not surprising to learn that this was not the first time the company made a major technology adjustment.
In fact, until 2006, Amadeus operated strictly on mainframes before it began moving to a more open and distributed system that would eventually become its private cloud, even though they didn’t call it that back then. It would be another decade before the last mainframe would be decommissioned.
As with the mainframes, over time, the tech debt accumulated, and the private cloud could no longer meet the company’s needs. Put simply, it lacked the flexibility the Amadeus desperately needed.
“We began to realize that from a traffic management perspective, using cloud technology would be more efficient,” Pellisé said.
It wasn’t purely about being able to scale infrastructure, he said. Amadeus also wanted the ability to adjust quickly to a fast-changing market, and experiment more without worrying about the resources they might need to complete any tasks.
The shift to the cloud is a project paced similarly to the shift from mainframes, and Amadeus is not trying to bite off more than it can chew.
The company made its initial foray into the public cloud in 2015, when, along with Lufthansa, it launched a beta of its first customer-facing cloud application, Amadeus Airline Cloud. Over the coming years, the company flirted more with public clouds but did not fully embrace the idea of moving its operations entirely.
By 2020, it was ready to get serious and began evaluating the big three cloud infrastructure vendors — Amazon, Microsoft and Google. After a nine-month review and lots of internal discussion, Amadeus chose Microsoft early last year.
Odeen said from a business perspective, Microsoft offered the most complete business and technology package along with a group of engineers and consultants that could work with Amadeus and its customers. “I’ve only heard good things from our engineers about the Microsoft platform, but on top of that, they have a whole layer of co-innovation with us as a customer and with our customers. That really makes the difference and adds a lot of value,” he said.
What’s more, Amadeus and Microsoft had a lot of shared customers, so the updates they were making could have a direct impact on customers using their services. To be clear, the company still works with Google and Amazon on other areas of its IT stack, but for this significant move, it decided to hitch its wagon to Microsoft.
Paying it off
Amadeus wanted to begin transitioning to the public cloud with Azure as the main vendor. The plan was never to do a lift-and-shift and move everything all at once, and as with the move from mainframes to the private cloud, it was to be a years-long project starting with some simple workloads to get the engineering team comfortable.
“The plans are to put it in motion to do the technical work on the applications, so that they can be containerized and moved to the cloud, and also to build the platform, working closely with the Azure engineers, to have a landing zone online so that we can learn and operate in a comfortable and secure manner,” Pellisé said.
He pointed out that there is another key part of the transition that has nothing to do with technology — training, change management, digital transformation and getting the engineering team ready for the transition. Of the company’s 17,000-plus employees, more than half are in technical roles.
The goal is to have a base in place that they can build newer, more modern applications on top of. At the end of March, the company announced the first step in that journey — a new Cytric Travel & Expense tool built on top of Office 365, which many Amadeus customers already use.
As they do this, the engineering team is shifting to working on the cloud. “Our engineers are excited about this move, about using new technology and the benefits that we can get by adding quicker capacity, by adding access to new and modern technology, like big data and artificial intelligence,” Pellisé said.
That involves a transition to a DevOps model, where developers are responsible for the code and operations are responsible for getting it into production when it’s ready. Amadeus is training its engineers in this approach with help from Microsoft and people in the company who are already familiar with it.
It’s worth noting that Microsoft is working with the Amadeus engineering team as they make this transition, and the Redmond-based company’s team is actually embedded in Amadeus’ teams, helping them with the day-to-day operational requirements of such a massive change. Smaller companies probably wouldn’t get such good treatment.
Amadeus is also being open with its customers about the change and the benefits the new cloud strategy should bring. Currently, the company is running systems in parallel — both in its own data center and the cloud until it can work out all the kinks.
Smaller companies born in the cloud may wonder how Amadeus’ transition is relevant to them, but as your technology becomes more dated, you too will have to make similar decisions. While a startup’s tech stack may not be quite as complex as a company the size of Amadeus, studying its process can be a useful model for when the time comes to pay off the debt.
Amadeus’ migration is just beginning. The entire process is expected to take three years if all goes according to plan, so while it won’t be as slow as the mainframe migration, it’s a substantial project nonetheless.
Comment