Amadeus’s Journey to Event Mesh: Scaling EDA to Support the World’s Travels

Amadeus is the world’s leading provider of travel technology, processing an astounding 400 million bookings annually, with their systems handling up to 35,000 transactions per second during peak times. With over 19,000 employees worldwide and petabytes of data being collected, distributed and processed, Amadeus’s journey to modernize their system offers valuable insights for enterprises facing similar challenges.

This post summarizes a talk that Senior Architect Serge Beuzit and Senior Software Engineer Nuno Gaspar gave at EDA Summit 2024.

Here’s a one-minute highlight reel, and you can watch the whole thing at the bottom of this post or over at EDA Summit.com.

Starting the Journey to Event-Driven

Back in 2017, Amadeus kicked off what they called their “Dynamic Intelligence” project. The goal was pretty straightforward: help airlines make better operational decisions, like letting passengers know about flight delays more efficiently. This was their first step into the world of cloud-native streaming and event-driven architecture, and it turned out to be just the beginning of something much bigger.

As more teams across Amadeus started seeing the benefits of this approach, different divisions jumped on board. The airlines team, the airports group, and even the hospitality applications folks all started developing their own cloud-native apps. But here’s where it gets interesting – they didn’t all use the same technology. Some teams went with Kafka, others chose NATS for their high-performance needs, and some opted for Azure Event Hub at their edge locations.

Growing Pains and the Path to Event Mesh

If you’ve ever tried to manage a growing system, you know how things can get complicated pretty quickly. Amadeus found themselves juggling over 300 Kafka clusters, and started running into challenges: they couldn’t easily see who was publishing what events, monitoring everything became a real headache, and costs were climbing faster than they’d like.

It became clear they needed a smarter solution. They wanted something that could route events across their global infrastructure without creating unnecessary copies, keep track of everything without getting overwhelmed, and make it easy for teams to work independently while still playing nice with each other.

After looking at their options, Amadeus decided to go with Solace PubSub+ Platform. Think of it as building a smart highway system for their data. Instead of having point-to-point connections everywhere, they now had an organized network that could efficiently move data where it needed to go. The cool thing was, they didn’t have to rip out their existing systems – they kept Kafka for the stuff it was good at and used Solace to connect everything.

They used PubSub+ Event Portal to design how everything would flow, kind of like having a master blueprint. Now they’re focused on making everything as automated as possible, so teams can self-serve their needs without creating chaos.

Benefits

When you look at what Amadeus has achieved with their global event mesh implementation, the benefits are pretty impressive:

• First, they’ve optimized the flow of information across their global infrastructure – imagine going from a tangled web of connections to a well-organized highway system for their data. This means they’re not sending the same data multiple times to different regions, which has helped them cut down on unnecessary data transfer costs.
• They’ve simplified the process of discovering and reuseing events. Their teams can now easily find and use the events they need, which has sped up development time and reduced duplicate work. This self-service capability means development teams can move faster without getting stuck waiting for other teams to help them out.
• Another big win has been in the area of autonomy and ability to use best of breed technologies. Different teams can now work independently while still staying connected to the broader ecosystem. They’re able to use the technologies that work best for their specific needs – whether that’s Kafka, NATS, or Azure Event Hub – while maintaining communication across the organization.
• The streamlined operations have also simplified their system. By replacing hundreds of point-to-point connections with an event mesh, they now have a more manageable and efficient architecture, which has reduced their operational overhead and made it easier to monitor and maintain their systems.

Conclusion

Amadeus managed to solve a complex problem without throwing everything out and starting from scratch. They found a way to keep what was working, fix what wasn’t, and set themselves up for future growth. It’s a great example of how sometimes the best solution isn’t the most dramatic one – it’s the one that carefully balances practical needs with forward-thinking design.

Their approach shows that when you’re dealing with big, complex systems, success isn’t just about picking the newest technology. It’s about finding smart ways to solve real problems while keeping an eye on what’s coming next. And that’s exactly what Amadeus managed to do with their event mesh implementation.