Founded in 1889, Michelin began as a small rubber factory in France and invented the radial tire. Today, the French company is one of the world’s largest tire manufacturers. With nearly €650 million spent yearly on R&D, Michelin produces tires for almost every type of vehicle, such as automobiles, bicycles, airplanes, farm equipment, heavy-duty trucks, motorcycles, and even NASA’s space shuttles. But Michelin is more than just tires. With a rich history of over 100 years of innovation under its belt, the company has evolved from a manufacturer that makes and sells tires into a data-driven services powerhouse. Michelin now harnesses vast amounts of data to deliver sustainable mobility solutions to its diverse customer base—such as predictive insights on when tires need to be replaced and the best routes to optimize fuel. Outside its transportation business, the multi-faceted company also publishes the prestigious Michelin Guide to restaurants worldwide.
Using Apache Kafka for Real-Time Data Pipelines and Integration
Michelin relies heavily on data for all its manufacturing, services, and publishing lines of business. As the company’s digital transformation journey progressed, they chose Apache Kafka®, an open source distributed data streaming technology used for real-time data pipelines, data integration, and stream processing. Open source Kafka gave Michelin the footing it needed to leverage real-time data for compelling business use cases. However, it also required full-time maintenance and infrastructure management by in-house Kafka experts. Michelin employed three full-time employees to manage Kafka clusters, but soon found it difficult to scale—particularly for their inventory management system.
The challenge: Real-Time Inventory Requires Cloud-Native Agility
The manufacturer had long dealt with unreliable and outdated reporting on inventory especially for raw and semi-finished materials. It affects their global supply chain and logistic operations thus their customers at the end. They needed a way to access real-time data and gain accurate views of inventory across their ecosystem. Through a proof of concept, they were able to monitor and analyze logistic flows between two sites using an event-driven architecture and a streaming solution—fueled by the power of the cloud. It worked, but recent worldwide crisis situations (COVID-19, Ukraine, etc.) put the initiative on pause. But they are soon to re-start and get the benefits demonstrated.
The Challenges of Using Kafka
“One of the challenges with Kafka was its operational complexity, especially as the footprint expanded across our organization. It’s a complex, distributed system, so we had to allocate a lot of our valuable technical resources and expertise to babysit it and keep it running. This also required us to incur additional in-house maintenance costs and risks. The other issue © 2022 Confluent, Inc. | Confluent.io 2 Case Study | Michelin was that it did not support our path to cloud, which is one of our company mandates. It was built for private infrastructures, not as a cloud-native data system, so it could not scale the way we needed it to and handle the volumes of data that we needed to process,” said Olivier Jauze, IT Architect and CTO of Mastero Marketplace, a service line in Michelin. Michelin knew that as an enterprise, it needed to improve the scalability and resilience of its Kafka deployments to gain reliable, real-time inventory management visibility and meet growing customer expectations for a seamless, online ordering experience.
Saving 35% in operational costs by using Confluent over open source Kafka
To address the challenges of operationalizing open source Kafka and accelerate their journey to the cloud, Michelin chose Confluent Cloud, a fully managed, cloud-native Kafka service. By centralizing data streaming and subsequently leveraging Confluent Cloud in a Microsoft Azure environment, Michelin expects to significantly reduce operational issues and free up cash flow. They estimate 35% savings compared to on-premises operations. Confluent Cloud has enabled Michelin to reduce its total cost of ownership (TCO), improve uptime with Confluent’s 99.99% SLA, drive faster time to market due to elastic scalability that supports 10TB/day of throughput, and refocus its Kafkadedicated teams on higher value tasks, rather than maintaining infrastructure. “We decided Confluent was necessary when we realized the limitations of self-managing open source Kafka on our own— inability to scale cost-effectively, monitoring and security limitations, and no well-defined path to the cloud. Confluent offered a complete platform with all the enterprise-grade capabilities we needed to run mission-critical use cases, end to end. We needed to work with a partner with deep Kafka expertise, and Confluent’s millions of hours of experience working with customers running Kafka in production helped make our decision easy,” Jauze said. “We estimate that for the last two years, Confluent has helped us to gain eight or nine months in terms of time to market to deploy the technology.”
Michelin began implementing Kafka in an on-premises model in its data centers to gain a real-time view of the business and to start collecting, processing, and storing data as continuous streams— instead of data at rest in databases or old legacy applications. By the end of 2019, the manufacturer decided to pursue a move to the cloud to mitigate some of the operational challenges of managing Kafka and realize the benefits of a cloud-centric digital transformation strategy. Michelin selected Microsoft Azure as its cloud partner and migrated to Confluent Cloud for Azure in 2021. “Azure is our top preferred cloud partner, and we wanted to select a cloud Kafka infrastructure provider that would fit our own choices. The fact that Confluent allowed us to adhere to our own multicloud decisions was really important for us,” said Valérie Servaire, IT Integration Architect at Michelin. With Confluent’s support, the Michelin team spent nine months assessing their migration needs and launching exploratory use case projects. Their most mission-critical project—online order management—went live in spring 2022. Online order management is the orchestration and choreography layer at the heart of Michelin’s supply chain. The company needed to migrate off its existing on-premises orchestrator, Oracle BPM, to a cloud version. However, after several successful projects and a conclusive proof of concept, Michelin decided to completely rewrite its own orchestrator based on Kafka and Kafka Streams. “With Confluent, we are transforming this very batch, very monolithic information system into one in which data is constantly in motion. It really helps us to decouple our monolith into autonomous systems, to help them evolve independently of each other, and therefore, to be more agile for our business to drive real-time, data-driven decisions and operations. It is data in motion serving business agility,” explained Jauze. Several departments at Michelin now use Confluent Cloud to free data silos across the business, including supply chain, customer services, manufacturing, and R&D.
Business Results - From the CIO’s perspective
“Moving to the cloud ultimately means saving time and money for us. It’s great from a cost-optimization perspective, and for tapping into the benefits of the cloud like ease of use, elastic scaling, resiliency, and enhanced security. This means that we can focus more on innovating and building products and solutions for our end customers instead of worrying about the day-to-day management of infrastructure,” said Caseau. Data is increasingly central to corporate strategy, and that is particularly true at Michelin. “Kafka enables us to unlock realtime data throughout our organization. But Confluent goes way beyond Kafka to offer a platform for data in motion that’s truly cloud-native and re-imagined for the cloud, while offloading Kafka management and removing operational burden. It’s also complete in that it has all the enterprise-grade capabilities we need for our mission-critical use cases, and it’s available everywhere our data resides,” explained Caseau. “Confluent plays an integral role in accelerating our journey to becoming a data-first and digital business. Today’s customers demand rich, personalized experiences, and business operations must be optimized to stay ahead of the competition. We use Confluent Cloud as an essential piece of our data infrastructure to unlock data and stream it in real time. This enables us to quickly and cost-effectively scale data in motion to additional use cases across the organization, such as customer 360, e-commerce, microservices, and more,” said Caseau. “As the CIO, my job is to help Michelin satisfy our customers every day with the best level of performance—moving to an event-driven architecture together with Confluent helps me get there faster and more efficiently.”
“Reliability was an essential part of our decision to move to Confluent. Kafka outages for mission-critical applications and downtime are a huge risk factor for our business. Building a resilient architecture that can withstand scaling up and down effortlessly, can manage spikes in demand, and can handle the volumes of data that we need processed were big parts of our decision to choose Confluent as well. Confluent Cloud’s cloudnative elasticity and 99.99% SLA means I can have the peace of mind to allow my teams to focus on innovation and building applications rather than Kafka infrastructure management and maintenance,” said Servaire.
The only true cloud-native Kafka experience
“Confluent offered the only truly cloud-native Kafka experience— Kafka reimagined for the cloud. They also offered unparalleled Kafka expertise, which is to be expected because Confluent was founded by the original creators of Kafka,” commented Jauze.
“One of the biggest assets Confluent has is knowledge of Kafka and streaming technology. It’s quite unique, in my 15 years of experience. They guide us very well, and through professional services, they both teach us and listen to us,” said Jauze. “When we first talked to Confluent, we got in touch with people who had developed the Apache Kafka project. And since then, whether it’s in London or in the French ecosystem, the people we’ve dealt with have always been incredibly experienced, very technically advanced, and have possessed a lot of instructional know-how. It’s a real differentiator for Confluent compared to other experiences we’ve had. Confluent knows the technology, knows how to deploy it, and knows how to explain it. When you are supported by people like those at Confluent, adoption happens on its own without barriers,” concluded Jauze.
Michelin continues to implement Confluent Cloud across the enterprise as they pursue their digital transformation journey. They foresee widespread adoption of data in motion across a number of new use cases as the business continues to experience a high ROI on Confluent projects.