Book Review: Data Mesh — Delivering Data-Driven Value at Scale by Zhamak Dehghani

Venkataraman Balasubramanian
13 min readOct 27, 2023

--

Background

During the last few months of my recent 4+ years of experience in leading a team to setup an enterprise-wide, cloud-based data lake platform, the concept of a data mesh started to gain traction in discussions within some of the use cases utilising the platform. I found this concept intriguing, as my initial understanding suggested that the data mesh approach contradicted the conventional centralised data lake model typical of enterprise-wide setups. However, during that period, I lacked the time and resources required to delve deeper into this novel and seemingly revolutionary concept.

I decided to devote time to the topic during the last few weeks.

Introduction

My motivations to read the book were as follows:

- To understand in detail, How the data mesh concept and approach was different that of a centralised data lake?

- Compare and contrast the 2 approaches from data & analytics platform point of view.

- What are the learnings (if any) we can re-use from setting up a centralised data lake platform for the data mesh platform.

I read a technical book after a long time. I would unequivocally say that it was time well spent. I felt a sense of elation after reading the nearly 400-page book, because:

- I am now convinced of the data mesh framework, which is a very logical next evolution to improve the downsides of a centralised data lake architecture.

- I was glad to see “Principle of Self-serve Platform” as one of the 4 core principles of Data mesh. Dehghani’s idea of the Self-Serve Data Platform revolves around autonomy but interoperability, As someone who was deeply engaged in building a data & analytics platform, It provided comfort to me, that the Platform is indeed a core enabler of a good data mesh architecture. (or rather any data & analytics architecture)

Zhamak Deghani’s writing style is detail oriented, compelling, sometimes brutally honest, and yet humble. I could see she has poured out her decades of experience, heart and soul into the book and is well worth the effort.

She addresses every aspect of the topic. In fact, as I was reading the book, some question would crop up in my mind and would haunt me — how can this work? or what should be done about this topic in data mesh architecture etc… I was delighted that the author would address the same in the subsequent chapters or even next pages. It was as though she precisely knew the concerns of the readers.

I could also see Dehghani’s philosophical bent of mind while addressing this transformational topic, as she relevantly quotes many philosophers & business / tech leaders throughout the book. It is up to the reader to make the connection between those quotes and the material presented in the relevant sections of the book.

I have summarised those quotes at the end of this article, as I have a shared interest on this topic.

Before we go further, I have a small piece of advice though.

This is an “intense” book. Flipping few pages and reading selectively some chapters can help one grasp the high level concepts (and help in the coffee table chat on the topic of data mesh), but If you are a data and analytics leader, seriously thinking to implement or transform your analytics architecture, it would require your meticulous reading — every chapter, page of the book to assimilate and work out a transformation plan for your specific organisational context.

Review

The book is logically well structured into 5 parts. What, Why and How of data mesh concept, architecture, and implementation. The How part is more detailed understandably focusing on how to design the core data mesh architecture, how to design the Data product architecture and how to get started to implement it.

The diagrams and tables are cool and easy to understand.

In the Prologue section, Dehghani creates a story about a music company and its data and analytics architecture, and how it works today with Data mesh implemented and how it was in the past with a traditional architecture.

It is interesting that the author always presents the TO-BE state first in detail, before talking about the drawbacks of the AS-IS state. The same flow is also seen in the Part 2 where she delves more deeply into the “Why data mesh” question.

Part 1 (What is data mesh?)

In the 1st Part, Dehghani builds the narrative on Data mesh concept; what it is, in a coherent manner.

The Figure 1–1 in the book summarises it all: Data mesh results in 6 dimensions of change.

Of all the above dimensions of change, some are obvious, while the last is the most disruptive to currently prevalent approaches. The rationale for all these dimensions of change is well explained in the book.

Readers may have one basic question: Is data mesh an architecture? Is it a list of principles? Is it an operating model?

The author sums it up nicely as “It is a socio-technical paradigm: an approach that recognises the interactions between people and the technical architecture and solutions in complex organisations. This is an approach to data management that not only optimises for the technical excellence of analytical data sharing solutions but also improves the experience of all people involved: data providers, users, and owners.”

There are 4 key principles the define the data mesh approach and architecture, and Dehghani dedicates one chapter for each of them, going into great depth to explain the rationale for each of them.

- Principle of Domain Ownership

- Principle of Data as a Product

- Principle of the Self-serve Data Platform

- Principle of Federated Compute Governance

Overall, I found that “Data as a Product” as a radical principle compared to current widely prevalent thoughts in the industry. It is one final end-product that is feasible, usable, and valuable. The chapter 3 deals with this topic in detail

Dehghani goes even to the extent to say that ETLs, CDC and Application Virtualisation are considered as anti-patterns, because these methods simply throw “raw data” from the operational systems to the outside world, instead of treating data as a final consumable product.

If you have to derive the maximum value out of the book, I would recommend to read thoroughly and be well convinced about the “data product” concept.

Dehghani does well to provide 5 key transition statements on how to go towards the approach of Data as a Product in the 3rd Chapter.

Personally, it was heartening to read the Chapter 4, on the Principle of Self-serve data platform that is close to my heart. Figure 4–1 confirmed my belief and experience of last 4+ years that a central platform team is essential to the success of any data & analytics infrastructure (including data mesh).

Figure 4–2 summarises how a data mesh platform differs from the other data platforms.

In the 5th Chapter, which is dedicated to the “Principle of Federated Computational Governance” Dehghani shares her frank opinion on the classical thinking of Data governance.

In her own words:I must admit, governance is one of those words that makes me, and perhaps many, feel uneasy. It evokes memories of central, rigid, authoritative decision-making systems and control processes. In the case of data governance, it evokes memories of central teams and processes that become bottlenecks in serving data, using data, and ultimately getting value from data.”

Soon after she proposes a way forward with this statement In short, in this chapter I aim to address the uneasy feeling of loss of control and indeterminism that a decentralised data ownership model like data mesh can arise in many of my tenured data steward and governance colleagues.”

In concluding the Part 1 (What is Data Mesh), there is an interesting statement that Dehghani makes “I have been in too many conversations where the number of petabytes or thousands of data tables have been mentioned as a sign of pride and success. While it’s understandable what has led to such metrics, there is no direct and reliable link from the volume to the value.”

This is an important topic for Data & Analytics leaders to think on how they can demonstrate the value of their investments (whether data mesh or not) Later in Chapter 15, Dehghani talks about “fitness functions” as a way to measure the value of Data mesh implementations.

Part 2 (Why data mesh)

This part consists of 3 chapters.

The normal order of the justification is reversed.

It starts with about the current inflection point in the data and analytics space (chapter 6), what can / will change by moving to the data mesh approach (chapter 7) and interestingly what is the history and evolution of the data and analytics space that has resulted in the current inflection point, in the final chapter of this part.

Again, Dehghani goes to great details explaining through nice diagrams, tables, and examples all the concepts.

Part 3 (How to design the Data Mesh architecture)

There are 2 chapters in this part, the first one (Chapter 9) dedicated to the logical architecture, and Chapter 10 dedicated to the physical architecture.

In the logical architecture Dehghani defines the main logical components of data mesh architecture such as domains, data product as an architecture quantum, and multiple planes of data platform and how they interact with each other.

Towards the end of Chapter 9, she lists in a table, the various architectural components, their description, and the current state of maturity of those components in the industry implementations (whether there is an out of the box product or a custom implementation for this type of component)

In the Chapter 10, she defines the capabilities that an infrastructure platform must offer to build, run, and operate an implementation of data mesh. It offers an overall approach to designing the platform, independent of its technologies.

The Multiplane Data Platform architecture presented in this chapter, illustrates the 3 components (Data infrastructure Plane, Data product experience plane and the Mesh experience Plane) and shows how different personas interact through the different planes.

In Dehghani’s words “The ultimate purpose of the platform is to serve the cross-functional domain teams so they can deliver or consume data products”.

She lists the key user personas (Data product developers, Data product consumers, Data product owners, Data governance members, Data platform product owner — Yes! That’s I, Data platform developer) their roles, and their journey touching various parts of the Platform.

Dehghani takes 2 of the key user personas (data product developers, data product consumer) and goes through their user journey in detail.

Part 4 (How to design the Data Product Architecture)

In this part Dehghani relies heavily on fundamental software design & engineering principles to recommend the way that data products should be designed, so that they are self-sufficient in all aspects.

She uses the term “affordances” to describe how a data product could be used.

In her own words “An affordance is a relationship between the properties of an object and the capabilities of the agent that determine just how the object could possibly be used.”

To sum up this part, I would quote 2 sentences from the book.

Dehghani states “If I could leave you with one takeaway from this chapter (14), it would be to invert your perspective on whose responsibility it is to manage, govern, and observe data; shift the responsibility from an external party getting engaged after the fact to the data product itself.”

That is in essence the motivation for how a data product should be designed.

When this happens naturally Data products contribute to the overall success of a Data mesh implementation, Dehghani confidently states that “Data mesh, like complex adaptive systems, doesn’t need a central controlling architectural element. The flock of birds doesn’t have an orchestrator, but what it does have is a set of biological standards defining how it detects speed, distance, and the leader. Similarly, data mesh introduces a set of standards that each data product follows for interoperability and cohesion of behavior.”

Part 5 (How to get started)

It is one thing to recommend a new concept & architecture, justify it, but it is another level of difficulty to address the non-technical elements of the transformation. Dehghani delves deep into this topic in Part 5

This part contains 2 chapters.

In the first chapter (Chapter 15), Dehghani starts with a frank statement that the data mesh architecture may not be suitable for every organization. She proposes a ‘spider map’ of criteria for self-assessment and goes on to explain each criterion.

In her view “Organizations with a centralized IT unit that shares people (resources) based on projects across all business units, without continuous and long-term ownership of technical assets for each business domain, are not a good fit for data mesh.”

Once an organization commits itself to the data mesh approach, Dehghani lays out a blueprint on how to derive strategic business initiatives and use cases based on the data strategy, plan out intelligent business applications and touchpoints, define data products for analytics for final execution on the multiplane platform.

All of that should be supported by organisational alignment structure & business driven iterative end-to-end evolutionary execution.

Finally in the last chapter (Chapter 16) Dehghani elaborates on the organisational aspect of the change, and how culture, people, process, organisational structure, and reward aspects to be defined to ensure the successful transformation.

My Personal Reflections

According to me, Everything is a circle and a cycle in this creation.

Be it enterprises’ organization structures, IT organization structures, IT systems (including analytics systems) all of them have gone through changes from being de-centralised > loosely coupled > highly centralised models, with relevant optimisations tuning the deficiencies in the past models.

Dehghani’s approach on data mesh focuses to eliminate downsides of a highly centralised, rigid data & analytics architecture by proposing a mesh of data products to offer self-contained, flexible, and valuable information for consumption, to ultimately create monetary value for the organization.

It will work very well when the right ingredients are there: a clear understanding of the benefits & mandate from the top management, competent data & analytics leaders willing to do the right thing and a solid execution plan.

Without those ingredients of success, the data mesh initiatives could become “data mesh”-washed, resulting in data & analytics silos resulting in personal fiefdoms of data

Quotable Quotes from the Book

The quotes mentioned by Dehghani are interesting to be quoted again.

Imagination will often carry us to worlds that never were. But without it we go nowhere.— Carl Sagan

….the only simplicity to be trusted is the simplicity to be found on the far side of complexity.— Alfred North Whitehead

“Think in simples” as my old master used to say — meaning to reduce the whole to its parts in simplest terms, getting back to first principles.— Frank Lloyd Wright

To reject one paradigm without simultaneously substituting another is to reject science itself.— Thomas S. Kuhn, The Structure of Scientific Revolutions

All models are wrong, but some are useful.— George Box

Two of the most important characteristics of good design are discoverability and understanding. Discoverability: Is it possible to even figure out what actions are possible and where and how to perform them? Understanding: What does it all mean? How is it supposed to be used? What do all the different controls and settings mean?— Don Norman

[Trust is] a confident relationship with the unknown.— Rachel Botsman

Simplicity is about subtracting the obvious and adding the meaningful. — John Maeda

Platform: raised level surface on which people or things can stand.— Oxford Languages

“The word platform is one of the most commonly used phrases in our everyday technical jargon and is sprinkled all over organisation's’ technical strategies. It’s commonly used, yet hard to define and subject to interpretation.” (I like this one, and Dehghani quotes several definitions of the platform in the book)

This is the Unix philosophy: Write programs that do one thing and do it well. Write programs to work together… — Doug McIlroy

For peace to reign on Earth, humans must evolve into new beings who have learned to see the whole first.— Immanuel Kant

Placing a system in a straitjacket of constancy can cause fragility to evolve.— C. S. Holling

By doubting we are led to question, by questioning we arrive at the truth.— Pierre Abelard

A strategic inflection point is a time in the life of a business when its fundamentals are about to change. That change can mean an opportunity to rise to new heights. But it may just as likely signal the beginning of the end.1— Andrew S. Grove

The only way to make sense out of change is to plunge into it, move with it, and join the dance.— Alan Watts

Today’s problems come from yesterday’s “solutions.”— Peter M. Senge, The Fifth Discipline

Any organization that designs a system (defined broadly) will produce a design whose structure is a copy of the organization’s communication structure.— Melvin Conway, 1968

The definition of insanity is doing the same thing over and over again and expecting different results.— Albert Einstein

The supreme goal of all theory is to make the irreducible basic elements as simple and as few as possible without having to surrender the adequate representation of a simple datum of experience.— Albert Einstein

Form ever follows function.— Louis Sullivan

The real truth of existence is sealed, until after many twists and turns of the road.— Rumi

The misconception which has haunted philosophic literature throughout the centuries is the notion of “independent existence.” There is no such mode of existence; every entity is to be understood in terms of the way it is interwoven with the rest of the universe.— Alfred North Whitehead

No man can cross the same river twice.— Heraclitus

It seems that change and time are inseparable: changes take time; are located and ordered in time; and they are separated by time. The inseparability of time and change is a kind of logical truth.— Raymond Tallis

Never tell people how to do things. Tell them what to do and they will surprise you with their ingenuity.— George S. Patton

Don’t control but observe.— Gregor Hohpe

Man muss immer umkehren. (One must invert, always invert.)— Carl Jacobi

A journey of a thousand miles begins with a single step.— Lao Tzu

The essence of strategy is choosing to perform activities differently than rivals do.— Michael E. Porter

Brilliant strategy puts you on the competitive map, but only a solid execution keeps you there.— Gary L. Neilson, Karla L. Martin, and Elizabeth Powers, “The Secrets to Successful Strategy Execution” (Harvard Business Review)

The only thing a Big Bang re-architecture guarantees is a Big Bang!— Martin Fowler

Culture eats strategy for breakfast.— Peter Drucker

--

--