Decathlon boosts scalability and speed With MongoDB

THE CHALLENGE

Need for high availability and dynamic scaling

When sporting goods retailer and manufacturer Decathlon embarked on redesigning its platform, the team faced a significant hurdle. Its SQL-based system couldn’t handle the massive influx of requests—ranging from hundreds to thousands per second—during peak times. Attempts to integrate Elasticsearch highlighted further challenges, particularly in managing the synchronization between SQL and Elasticsearch, which proved cumbersome and inefficient. Decathlon’s global footprint spanning more than 60 countries brought additional complexities. Customer reviews, essential for guiding purchases and improving products, are published globally, but each country prioritizes its own local reviews. This demanded a solution that could guarantee responsiveness across regions.

“We needed a system capable of supporting high availability and dynamic scaling without the complexities of synchronization,” said Cyril Gambis, Lead Architect at Decathlon.

OUR SOLUTION

Adopting NoSQL architecture for flexibility and cross-region functionality

The team turned to MongoDB Atlas—a NoSQL architecture offering distinct advantages. “We needed horizontal scalability for growing volumes,” said Gambis. “MongoDB offered this, along with features like auto-scaling and shard capabilities to distribute load globally.” MongoDB’s document-based model could also manage diverse data formats like text, images, and videos, essential for modern user-generated content. Moreover, MongoDB’s cross-region functionality ensures that reviews are hosted close to their origin, improving user experience through faster load times while efficiently supporting Decathlon’s growing demands. MongoDB’s architecture employs a cluster of nodes to coordinate operations: The primary node handles writes, and multiple secondary nodes manage reads. This setup enables Decathlon to fine-tune the balance between data consistency and availability.

To achieve scalability, MongoDB uses sharding to distribute data across multiple servers. A dedicated analytics node aggregates data from all shards, enabling efficient cross-region queries without slowing performance. “Sharding is an important capability of MongoDB Atlas,” said Gambis. “When queries come in, MongoDB directs them to the appropriate shard based on the shard key.”

OUTCOME

Streamlined analytics with a resilient, scalable, and future-ready platform

The transition to MongoDB transformed Decathlon’s review management system. The new platform seamlessly handles surges in traffic, scales horizontally, and provides local responsiveness through zone-based sharding. Its ability to handle different data models has eliminated inefficiencies seen in PostgreSQL’s JSON storage, which struggled with mixing structured and unstructured data in the same schema.

The improved architecture has also streamlined analytics. While data from Europe and Asia might be stored separately, the analytics node facilitates global queries without significant slowdowns. If a node fails, requests are automatically rerouted to another node, a process typically taking less than two seconds, ensuring minimal disruption.

Best practices, like using odd-numbered nodes for quorum and strategic distribution across zones, further enhance reliability. “With MongoDB, we’ve built a system where even during node failures, operations continue almost unnoticed,” said Gambis. Thanks to the evolving consistency features of MongoDB, Decathlon plans to extend its use from social networks to enterprise systems.

To learn more, visit MongoDB Atlas.