An approach to data architecture that gives business domains ownership of their own data, shared as products across the organization.
Data mesh was proposed by Zhamak Dehghani in 2019 as an alternative to centralizing all data in one platform run by a single data team. It rests on four principles: domain ownership of data, treating data as a product, a self-serve data platform, and federated governance with shared standards.
For example, in a large retailer, the logistics team would own and publish reliable delivery data, the marketing team would own campaign data, and each would document and maintain its datasets so others can use them. A central platform team provides the shared tools. Data mesh can reduce bottlenecks in large organizations, but it requires mature teams, strong standards, and significant cultural change. Smaller companies often find a central data team simpler and more effective.
The policies, roles, and processes that determine how an organization's data is managed, protected, accessed, and kept accurate.
Data LakeA central repository that stores large volumes of raw data in its original format until it is needed for analysis.
Data WarehouseA system that stores cleaned, structured data from many sources, optimized for reporting and business analysis.
What a data catalog is, the 3 kinds of metadata it holds, the features that matter, how it differs from a data dictionary and how to roll one out without it going stale.
Data mesh changes who owns data. Data fabric changes how it is connected. A side-by-side comparison, who each approach suits and how to tell whether you need either one.