Data Mesh in Action

by Jacek Majchrzak, Sven Balnojan, Marian Siwiak

Revolutionize the way your organization approaches data with a data mesh! This new decentralized architecture outpaces monolithic lakes and warehouses and can work for a company of any size.

In Data Mesh in Action you will learn how to:

* Implement a data mesh in your organization
* Turn data into a data product
* Move from your current data architecture to a data mesh
* Identify data domains, and decompose an organization into smaller, manageable domains
* Set up the central governance and local governance levels over data
* Balance responsibilities between the two levels of governance
* Establish a platform that allows efficient connection of distributed data products and automated governance

Data Mesh in Action reveals how this groundbreaking architecture looks for both small startups and large enterprises. You’ll see a data mesh in action as you explore both an extended case study and multiple real-world examples. As you go, you’ll be expertly guided through discussions around Socio-Technical Architecture and Domain-Driven Design with the goal of building a sleek data-as-a-product system.

Genres: Engineering

300 pages, Paperback

Published July 1, 2022

Reviews

[Sebastian Gebski · Rating 3 out of 5]

The 2nd book on the topic of data mesh I've read (after "Data Mesh: Delivering Data-Driven Value at Scale"). And probably the last one.

1. It IS better than the other one, but it doesn't mean it's VERY GOOD.
2. It is definitely more practical - apart from the whole concept (which, of course, takes a lot of space & focus) of decentralization & data-as-a-product, the authors put some effort into diving deeper into technical details (more on that later).
3. This book also has far better examples, which are heavily under-utilized. I hoped the authors will use the scenario introduced in the beginning to illustrate all the challenges covered, but they are not consistent about that.
4. Unfortunately, the practical part disappoints as well. The authors have focused on mentioning the typical "big data" implementation technologies instead of focusing on what's really important:
- dependency management
- contract management
- data lineage
- public versus private sets & good practices
- operational aspects of self-service for data

Frankly, if the book was supposed to "sell" me the idea of "data mesh", it has failed. I strongly believe that the same (or better) results can be used in a properly governed data lake with:
- standardized access formats
- shared contract publication conventions + open standard data catalog
- clear split of DL into "stages" (where more mature stages can depend only on the less mature ones)

That's why data mesh sounds (to me) like yet another consulting concept to sell old practices in some new clothes to naive/clueless customers.

[Ruslan Bes · Rating 2 out of 5]

Data mesh is a very cool concept. It's like Web 2.0 but for internal company data.

In Web 1.0, there were just a few sites with content, and most of us consumed that content in the available form (HTML). In Web 2.0, everyone can generate and share content and remix content generated by others. That's the power we have now — the power of distributed content generators.

Data mesh is similar. The traditional approach to working with large heaps of data is to have one data silo (Data Warehouse, Data Lake, or Data Lakehouse) and let everyone consume it. The data mesh approach instead puts the responsibility to provide data on those who produce it and forces them to define API, structure, permissions, sanitize the data, and so on. The data-producers become "data-sellers" in a sense.

And that's the whole theory behind data mesh.

The theory is good, but how do you implement that in practice? I expected that this book would tell the answer, but it didn't. Despite being part of the "In Action" series, it has very few examples, all of which are more theoretical concepts than something you can try. A big portion of the book is filled with very generic statements like:

"As you can see, data governance describes an environment in which data can be transformed into information, information to knowledge, and so on, up to the business value."

or

"We also need to understand the reasoning behind decisions that were made so we can make conscious choices about changes"

or

"In this context, the governance structure is the design framework and implementation of rules and policies aimed at ensuring proper handling of the company's data. On the one hand, it should be as transparent as possible, not imposing any unnecessary work related to maintenance of the framework itself. On the other hand, the governance structure needs to be all-encompassing; there should be no way to handle company's data outside of this framework. Ensuring that the following elements are deeply embedded in the governance structure will increase its resilience and purposefulness"

You get the idea.

Additionally, the book suffers from typical filler, where every chapter begins with something like "In this chapter, we will discuss XYZ" and ends with "Summary: in this chapter, you've learned XYZ." Remove that, and the book would be 5% smaller without any disadvantages.

I'd recommend only the first 2-3 chapters to understand the theory behind data mesh. Reading the rest is optional.

[Matt · Rating 5 out of 5]

The authors do a solid job breaking down why decentralizing data makes sense and how to actually do it without making a mess. Helpful.