Rolling Upgrades & Low-Latency Processing: How SNCF Optimized Train Operations and Geolocation
Challenge
- SNCF required low-latency performance to support and maintain session and communication data across an international train network.
- Always-on, real-time data transfer between trains and ground systems is required to ensure smooth operations.
- Needed precise geolocation tracking and instant operational decision-making to ensure on-time performance and passenger safety.
Solution
- SNCF implemented Hazelcast Platform Enterprise Edition for low-latency compute to support real-time data needs.
- Enabled rolling upgrades and high-performance, scalable data handling for uninterrupted service across the network.
- Integrated with existing Java and Kubernetes-based environments, supporting SNCF's complex IT landscape.
Results
- Zero downtime, providing continuous real-time operations
- Faster response times and improved train scheduling.
- Improved service delivery through accurate geolocation and train operational data.
- Increased reliability and safety in train services, boosting passenger confidence.
- Reduced costs from fewer delays and unscheduled maintenance.
During trials, Hazelcast Platform proved the most suitable caching solution due to its proven performance and scale, intuitive development environment and richness of functionality which ultimately amounted to a far superior solution than the alternatives for developing real-time, mission-critical applications.
— Denis Jouvin, Architect SI
How can your business benefit from real-time insights?
Industry
Transportation and Logistics
Year Founded
1938
Project Name
Train Connect
Business Need
Efficient management of real-time operational data across extensive train networks
Technological Challenge
High demand for real-time, always-on data processing and scalability in train operations
Solution
Hazelcast Platform Enterprise Edition
Founded in 1938, Société Nationale des Chemins de fer Français (SNCF) is France’s national, state-owned railway company. A Fortune 500 company with more than €30 billion (~$33 billion USD) in revenue, SNCF has transport and logistical operations across 120 countries.
On a given day, SNCF operates 10,000 commercial trains to transport 5 million passengers and over 250,000 tons of cargo. Across Europe, SNCF’s network consists of 35,000 kilometers, which includes 2,600 of high-speed railway lines and nearly 15,000 km of electrified lines.
Most recently, SNCF transported more than 2.4 million passengers for the 2024 Olympics in Paris.
Background
A central hub, known as Train Connect, is at the core of SNCF’s daily operations. This system serves as a communication service to/from trains and is relied upon by multiple applications, including the company’s customer mobile application, SNCF Connect.
At the core of Train Connect are two subsystems:
- A geolocation service: Sending real-time location updates to, from, and between trains.
- A data acquisition service: A real-time operational status of each train, including passenger information, the exact vector path route taken, the quality of the communications between the train and the ground systems, the passcom status, temperature, and many others.
Both subsystems are critical to multiple SNCF applications, including train mapping and status for internal teams, updating passenger reservations, and managing traffic to ensure on-time arrivals and — most importantly — overall safety across SNCF’s network. Given the variables across SNCF’s rail network and trains, the company’s Train Production team employs multiple geolocation methods to provide the most accurate information to internal agents and passengers, including real-time mobile communication and path vectorization.
Hazelcast Platform enables us to provide critical applications with an exact, up-to-the-second position vector of all 10,000 trains every day to ensure we run our business as safely and efficiently as possible.
— Denis Jouvin, Architect SI
Challenges
The primary challenge faced by SNCF was the integration and processing of large volumes of real-time operational data generated daily by hundreds of trains across the network. Denis Jouvin, the IT architect at SNCF, made clear that his team required a solution that could guarantee:
- Real-time, highly available data transfer and processing to ensure a continuous data flow between trains and ground control with zero downtime and zero data loss.
- Data accuracy for precise geolocation tracking and instant operational decision-making.
Only an in-memory solution could enable instant action on event-driven data, and provide zero downtime, zero data loss, and integrate with SNCF’s existing infrastructure.
Solution
Alternative solutions were used for isolated, low-impact applications, but Hazelcast Platform proved the most capable due to its intuitive development environment and richness of functionality, such as independently scalable data structures, secondary indexes, queries, and entry processors. Ultimately, Hazelcast Platform proved to be a superior solution.
A critical component of the Train Connect project, Hazelcast Platform exceeds the performance and scalability requirements and is key to several SNCF implementations, including:
- Operational Reliability: Hazelcast Platform’s Rolling Upgrade capability - built on its distributed architecture - allows SNCF to seamlessly update individual server nodes, one at a time, while the remaining nodes continue processing the full volume of event-driven data. This rolling approach ensures uninterrupted data availability and accuracy across all points of operation, maintaining seamless performance even during system updates.
- Geolocation Services: Hazelcast Platform’s data cache manages geolocation data to track and manage train movements accurately.
- Data Acquisition: Hazelcast Platform processes high volumes of data collected from trains in real-time to ensure quick decision-making.
The deployment focused on integrating Hazelcast Platform with existing Java-based applications using Spring Boot, which is prevalent in SNCF’s IT environment. Another critical factor was the tight integration with Kubernetes, which is core to SNCF’s containerization strategy.
Hazelcast Platform is now approved as the default caching solution in our software stack across our whole development ecosystem.
— Denis Jouvin, Architect SI
Impact
The Enterprise Edition of Hazelcast Platform significantly enhanced SNCF's ability to manage train operations by providing:
- Always-on availability: Zero downtime assures a continuous, real-time experience around the clock.
- Enhanced Operational Efficiency: Real-time data processing capabilities facilitate quicker response times.
- Increased Data Accuracy: Improved data accuracy in train positioning and operational status, leading to scheduling and service delivery improvements.
Business Impact
- Customer Satisfaction: Enhanced safety, precision, and reliability of train services contribute to greater customer satisfaction.
- Operational Cost Savings: Higher data processing efficiency and reduced operational costs associated with delays and unscheduled maintenance.
- Innovation Facilitation: The flexibility of Hazelcast Platform — such as support for Entry Processors, Compact Serialization, Atomic Longs, and more — enables SNCF to experiment with and implement new innovative approaches to train management.
The support team helps us understand how Hazelcast Platform works internally. The overall experience was satisfying because we felt we were not left alone.
— Denis Jouvin, Architect SI
Conclusion
SNCF’s deployment of Hazelcast Platform demonstrates its effective handling of large-scale, real-time data across complex transportation networks. As SNCF looks to the future, Hazelcast Platform will continue to play a vital role in enabling the railway company to achieve its efficiency, reliability, and customer value goals, ensuring that it remains at the forefront of transportation and logistics technologies.