Where data mesh changes who owns data, data fabric changes how data is found and connected. Analyst firms such as Forrester and Gartner promoted the term for an integration layer that spans every store a company already has, on premises and across clouds, without first copying everything into one place. Its engine is active metadata: the fabric continuously collects schemas, lineage, query logs, glossary terms and owners, links them in a knowledge graph, and uses the result to recommend, automate or enforce integration. IBM, one vendor that sells it, also stresses data virtualization, which makes data "accessible without physically moving it".
In practice a fabric is assembled from parts you meet later in this book: a catalog with lineage (Governance), a federated query engine such as Trino 403,499 that joins tables across databases and lakes (Query Engines on the Lakehouse), and policy-based access control and masking (Securing the Warehouse and Governance). No single open-source project is "the fabric". Mesh and fabric are compatible: a mesh's self-serve platform can use fabric tooling to publish and find data products.