Accelerating the future of travel distribution

THE CHALLENGE

Balancing travel seller needs with airline limits

As a provider of vital IT solutions for airlines, airports, hotels, and more, Amadeus IT Group is a major global operator in the travel industry. It’s also a business that works on a massive scale,  supporting over 400 airlines and handling over 2 billion boarded passengers in one year alone. 

“This generates a huge amount of traffic,” explained Ali Makke, NoSQL Expert at Amadeus. “We recently passed the threshold of 150,000 transactions per second.”

A recent shift in the travel industry is the introduction of the New Distribution Capability (NDC), a standardized data format that enables airlines to distribute rich content, personalized offers, fare lookups, and ancillary services directly to travel agents and consumers in real time. It has revolutionized retail air travel by replacing legacy systems, enabling faster product launches, and providing significantly faster, more tailored booking experiences for consumers.

While NDC has transformed the buying experience, it also allows travel sellers to bypass traditional Global Distribution System (GDS) channels and connect directly with airlines to access personalized, tailored offers, which come with their own set of challenges. 

“NDC makes the travel booking process a lot less generic and a lot more personalized,” said Makke. “But it also involves many technical challenges, especially for airlines’ system loads and traffic.”

Historically, GDS platforms such as Amadeus acted as a shield, efficiently caching and filtering traffic so only 10% of the full traffic volume reached the airlines themselves. With NDC, direct real-time connections remove this critical buffer, putting enormous traffic loads on airline IT systems. As a result, many airlines now impose ‘look-to-book’ ratio limits. 

“Airlines need to manage search traffic from aggregators and travel sellers to keep costs under control, so they limit the number of searches and transactions before one becomes an actual booking,” explained Makke. “On the other hand, travel sellers want fast and reliable access to NDC content because they want to support use cases such as discovery and price comparison.”

The challenge for Amadeus was clear: strike a balance between protecting airline systems from overload and providing travel aggregators with fast, flexible access to shop for complex NDC content.

OUR SOLUTION

Building an intelligent caching system

Amadeus resolved this modern distribution challenge by building an “intelligent caching system” with MongoDB at its core.

Amadeus realized that a basic cache would be insufficient for dynamic flight availability and pricing, and thus implemented an architecture that combines Kafka, microservices, big data, and machine learning models to predict the exact validity period of flight offers. This information is then stored as metadata within MongoDB.

“The idea is pretty straightforward. We want to provide alternative sources for NDC content but without compromising its quality,” said Makke. “It’s smarter than a basic cache and can provide content at a quality that’s very close to an NDC call. We want the traveller to be able to see an offer and book it later.”

Specifically, Amadeus uses MongoDB's flexible document model to support varied access patterns, utilizing an ‘upsert’ pattern with unordered bulk operations to refresh content rapidly. To maintain efficiency, Amadeus uses native Time-To-Live (TTL) indexes to automatically clean out older data.

“We rely on one of the native solutions that MongoDB offers, which is the TTL index,” said Makke. “We have a TTL index on update time. Any document that has not been updated in more than two days is automatically removed from the cache.”

To distribute the massive transactional load, Amadeus intelligently employs hashed sharding for specific key-value trip searches and range-based sharding for complex open queries. This allows the system to avoid the previous “scatter-and-gather” performance bottlenecks.

Then, by applying  high-speed data compression on the application side rather than within the database, Amadeus drastically reduces document sizes by around 90%, shrinking individual payloads from 100 kilobytes down to as little as 3 kilobytes. 

“This way we can uniformly distribute our data over all the available VMs inside our cluster,” added Makke.  “The compression then reduces network traffic, improves the memory usage on the cluster, and enables a smaller operation log (oplog) size.”

OUTCOME

High-speed, scalable search processing

The MongoDB-powered intelligent cache allows Amadeus to provide high-quality search results to travel sellers without overloading airlines’ systems. This gives consumers the search speed and performance they want, but without breaching airlines’ online transaction limits.

“We are achieving great levels of performance with MongoDB,” said Makke. “We’re processing more than 20,000 searches per second with response times in the order of a few milliseconds.”

And for targeted point-to-point travel searches, where the user already knows their trip dates and destinations, the system easily handles 12,000 searches per second, with an outstanding response time of 3 milliseconds.

Furthermore, MongoDB's distributed architecture and sharding capabilities enable Amadeus to scale horizontally, effortlessly supporting the broader travel platform’s global threshold of 150,000 transactions per second. 

“With over 60,000 enterprise customers and 30 million monthly MongoDB downloads worldwide, the platform provided a database solution that’s really gaining continued traction within the market.” Makke added.

Amadeus’s deep collaboration with MongoDB now provides the schema flexibility and foundational scalability to ensure the platform can evolve and stay ahead of the travel industry’s relentlessly growing demands and transaction volumes. 

“We’re giving travel sellers—and their consumers—the access and flexibility they need while respecting the airlines’ traffic limits,“ concluded Makke. “With MongoDB, we can meet every use case, from inspiration to price comparison, even to building full trips. It’s a very powerful partnership.”