DATA ARCHITECTURES ARE CHOICES, NOT IDEALS

In Deciphering Data Architectures, [1] James Serra delivers a needed and rare perspective on data and technology that is intellectually far beyond the flashy ideas of the moment. Historians have a dedicated term for the study of such a stretched timespan, exceeding more than the important events in a given era. It’s the french notion of la longue durée, the long term in English — the study of long term structural change, as transformations occur and alter the reality, slowly, as actions and reactions push each other again and again.

Serra writes with ease and clarity that testifies to many years of practical experience, as a seldom found description of a longue durée of data architectures and technology: Complicated topics are described so that everyone easily understands them and manages to follow an evolution over the course of many decades.

Serra defines a data architecture as the design and organization of data in an information system, and on that basis unfolds the long term development of data architectures. Starting with the Relational Data Warehouse, Serra moves on to Data Lakes, Modern Data Warehouses, Data Fabric, Data Lakehouse and Data Mesh, contextualizing each architecture as action and reaction, evolution towards ever more powerful architectures to unleash all the potentials that data holds.

A hidden treasure in Deciphering Data Architectures is the Architecture Design Session, that through a series of carefully crafted questions allow the participants to decide what data architecture the should choose based on their need to store, transport and analyze data. This is important, as there is, behind the promises of each data architecture, a set of constraints to also take into account. Accordingly, data architectures are choices — with consequences — and not ideals to strive for.

With the substantial evolution in the data architecture domain the last 3–4 decades, companies find themselves bewildered about the choices they have to take — and thanks to Serra’s clear cut text, these companies are now Deciphering Data Architectures.

[1] J. Serra, Deciphering Data Architectures (2024), O’Reilly