combigo_new_cover
folder

400-500 million documents
stored on OVHcloud Public Cloud Databases For MongoDB

magnifying glass

3,000 to 4,000 requests
processed per second

Clock

0% unavailability
since MongoDB was put into production

Executive summary

There are already many websites where tickets for flights, trains and buses are booked, but the startup Combigo has an original proposition: it combines means of transport to offer the lowest prices, and the lowest possible carbon footprint. “The added value of Combigo is knowing how to combine different means of travel to offer the best routes to our customers,” explains Vincent El Khatib, Co-founder & CTO at combigo.com.

Their algorithms will find the cheapest combinations by taking steps to select the best pricing options. The results of this approach are impressive — at €83 and 216kg of CO2 for a Paris-Lisbon journey, €91 for a Paris-Athens journey, combining train, plane and coach, Combigo is unbeatable.

The startup announces a 70% price drop and a CO2 emission drop of up to 96% for a train + train journey from Paris to Berlin, for example. In short, Combigo optimises all journeys for travellers who are willing to make one or more connections.

Currently, France and Europe account for around 90% of their ticket sales — but the startup already has customers in the US and the rest of the world, and this rate will only increase when travel restrictions are lifted.

The challenge

To determine the optimal path in terms of price, carbon footprint or even the quickest route, the startup has been working for more than 2 years now on highly complex optimisation algorithms. In fact, if price is the main priority rather than journey time, the algorithm can offer the very cheapest prices by offering journeys with more connections.

As with any algorithm-based company, data is Combigo’s raw material. Large volumes of data: The service already compiles data on all routes and prices for 4,000 airlines, rail and coach operators. This activity involves the startup manipulating a lot of data. We are mainly connected to Kayak and Liligo, as well as several other comparison sites. Our data may be updated 30 to 40 times per second for some sources.

The startup, which is still only in its launch phase, must already store between 400 million and 500 million documents in its main database. If the audience grows rapidly as the service succeeds, the database already needs to process between 3,000 and 4,000 requests per second.

The solution

Since the project was launched, the CTO opted with the MongoDB database to handle this flood of data. Vincent El Khatib stayed true to this choice of MongoDB — the database is simply supplemented on the production platform by the Redis solution, which provides a caching function on the results delivered by MongoDB.

Combigo diagram

 

The data handled by Combigo is structured, but it is accessible in the form of documents specified by the Technical Director. Moreover, he believes that a standard relational database is much more difficult to scale. In his opinion, a document-orientated database is best suited for simply loading all of the files, as well as modelling all schemas and creating the new index.

As a longtime user of OVHcloud services, Vincent El Khatib naturally chose Europe’s trusted cloud provider to switch to his MongoDB cloud solution. In February 2021, as commercial activity started to take off, Combigo decided to join the Startup Program — which would provide them with usage credits for their services. “With these credits, we were able to get our business off the ground smoothly — and as our activity gained momentum, OVHcloud offered for us to join its Scale Program. We were able to get more credit, and we also had the chance to participate in the beta-testing phase of the Public Cloud Databases for MongoDB solution.

The CTO chose OVHcloud over other MongoDB cloud solutions, because he could ensure that the database would be as close as possible to his computing infrastructure. This was an important factor for him to consider, as the algorithms they use achieve a high degree of data access. “Since the network between our computing instances and the database is so close, we can get the best possible performance. On the other hand, we are guaranteed that all backups are stored in another datacentre for security reasons,” he explains.

“Data is absolutely critical to our business because without it, we can no longer deliver our service to customers. The Automatic Backup option — delivered by OVHcloud in its Public Cloud Databases For MongoDB solution — gives us confidence and saves us valuable time, as we do not need to configure backups manually.”
Vincent El Khatib, Co-founder & CTO at combigo.com

The result

With the OVHcloud Public Cloud Databases For MongoDB solution, Combigo developers can create a MongoDB database in just one click with a ready-to-use cluster — and above all, they get automated data backup.  “Data is absolutely critical to our business because without it, we can no longer deliver our service to customers. The Automatic Backup option gives us confidence and saves us valuable time, as we do not need to configure backups manually.”

Also, there is no need to worry about MongoDB version upgrades any more, as OVHcloud is responsible for the migration: Combigo operators no longer have to focus on time-consuming version upgrade processes.

The Combigo service is still scaling up at high speed, and relies on its infrastructures and the PaaS MongoDB solution to facilitate its iterations. “With the Public Cloud Databases for MongoDB solution, we were able to build new databases quickly in order to carry out tests, and check how the platform reacts to both new configurations and new algorithm versions. This enables us to compare the performance of the two configurations, and advance by iterations as a result. With our credits and the ability to create MongoDB databases quickly, we can develop rapidly.”

In terms of availability, the CTO explains that he has never encountered any downtime since he launched his Public Cloud Databases For MongoDB solution — and this includes the beta-testing phase, which ended in October 2021.

The next changes to appear on the startup's roadmap will include the incorporation of data from a large GDS (Global Distribution System) from the airline industry. This will represent another leap forward in terms of volume. The startup is also constantly developing its algorithms. It uses several models in parallel, in order to evaluate the pricing of journeys.

Soon, Combigo will use AI to further refine its price forecast based on past pricing. “We know that we can count on OVHcloud to scale the infrastructure when we launch these new algorithms.

From 15 to 20 million documents stored when it was launched, the startup had more than 400 million documents in late 2021. Commercial activity now reaches 1 million GMV (Gross Merchandise Volume), and the founders expect activity to skyrocket in the coming months. However, they remain calm about their infrastructure’s ability to handle the growth in demand. This is one of the virtues of OVHcloud’s Managed Database as a Service solutions.