Monolithic applications present challenges for organizations like Saxo Bank, including difficulties when it comes to transitioning to cloud, data efficiency, and performing data management in a regulated environment. Graham Stirling, the head of data platforms at Saxo Bank and also a self-proclaimed recovering architect on the pathway to delivery, shares his experience over the last 2.5 years as Saxo Bank placed Apache Kafka® at the heart of their company—something they call a data revolution. 

Before adopting Kafka, Saxo Bank encountered scalability problems. They previously relied on a centralized data engineering team, using the database as an integration point and looking to their data warehouse as the center of the analytical universe. However, this needed to evolve. For a better data strategy, Graham turned his attention towards embracing a data mesh architecture: 

1. Create a self-serve platform that enables domain teams to publish and consume data assets
2. Federate ownership of domain data models and centralize oversights to allow a standard language to emerge while ensuring information efficiency 
3. Believe in the principle of data as a product to improve business decisions and processes 

Data mesh was first defined by Zhamak Dehghani in 2019, as a type of data platform architecture paradigm and has now become an integral part of Saxo Bank’s approach to data in motion. 

Using a combination of Kafka GitOps, pipelines, and metadata, Graham intended to free domain teams from having to think about the mechanics, such as connector deployment, language binding, style guide adherence, and data handling of personally identifiable information (PII). 

To reduce operational complexity, Graham recognized the importance of using Confluent Schema Registry as a serving layer for metadata. Saxo Bank authored schemes with Avro IDL for composability and standardization and later made a switch over to Uber’s Buf for strongly typed metadata. A further layer of metadata allows Saxo Bank to define FpML-like coding schemes to specify information classification, reference external standards, and link semantically related concepts. 

By embarking on the data mesh operating model, Saxo Bank scales data processing in a way that was previously unimaginable, allowing them to generate value sustainably and to be more efficient with data usage. 

Tune in to this episode to learn more about the following:

• Data mesh
• Topic/schema as an API
• Data as a product
• Kafka as a fundamental building block of data strategy

EPISODE LINKS

• Zhamak Dehghani Kafka Summit 2021 Keynote
• Data Mesh 101 Course
• Data Mesh Principles and Logical Architecture
• Saxo Bank’s Best Practices for a Distributed Domain-Driven Architecture

SEASON 2
Hosted by Tim Berglund, Adi Polak and Viktor Gamov
Produced and Edited by Noelle Gallagher, Peter Furia and Nurie Mohamed
Music by Coastal Kites
Artwork by Phil Vo

• 🎧 Subscribe to Confluent Developer wherever you listen to podcasts.
• ▶️ Subscribe on YouTube, and hit the 🔔 to catch new episodes.
• 👍 If you enjoyed this, please leave us a rating.
• 🎧 Confluent also has a podcast for tech leaders: "Life Is But A Stream" hosted by our friend, Joseph Morais.