Data Mesh in Practice

Chapter 2. The Pillars of Data Mesh

Data mesh is more than another iteration of an analytics data architecture that sticks with the centralization paradigm. Instead, it presents entirely different ways of interaction between data producers and data consumers and therefore should rather be considered a paradigm shift.

Data mesh builds mostly upon concepts and practices that have already successfully been applied in general operational systems architecture but have yet to be widely applied in the analytics data space. Figure 2-1 shows the four main pillars of data mesh, as presented by Zhamak Dehghani,¹ each pillar representing an application of an already established architectural principle to the analytics data domain. The following sections will introduce them one by one.

Figure 2-1. The pillars of the data mesh paradigm

Decentralized Domain Ownership of Data

One key departure from the established data architectures is that data mesh promotes decentralized domain ownership of data. The keyword is “ownership” here. With ownership, we mean full end-to-end responsibility. In order for this decentralized data ownership to succeed, data mesh applies domain-driven design.² A domain in this context can comprise different things. Typically, there are business entity domains such as customer, sales, and the like, but there can also be more technical domains (or subdomains) such as ...