Data Mesh in Practice

Chapter 5. Sustaining the Mesh: Federated Computational Data Governance

Building a data mesh at company scale addresses several different angles of working with data. We have already covered what it takes locally to start building a data product. We also introduced how it is possible to support those local data product builders through infrastructure platform capabilities to ease their journey toward high-quality data products. What we have not addressed yet is how can we make sure various data products are not starting to drift apart? How can we prevent different domains from becoming isolated silos of information?

Ensure Interoperability Through Semantic Cross-Domain Modeling

We wanted to build an integrated viewpoint between the sales data of our company and the behavioral data of our customers that was collected along their journey on our platform. Both areas had high-quality data products that were well described, easy to find, provided strong guarantees, and had contact product people to work with and discuss those products’ usages. Unfortunately, both data products were residing in different source systems, and we heavily underestimated the integration effort between them. After one month of integration effort for each of those two systems into our usual analytics platform, we had to realize that those well-defined data products were not compatible at all. The identifiers used in each product not only had different data formats but also followed different semantics. ...