Data Mesh is a socio-technical paradigm shift towards the decentralisation of data efforts within organisations that has seen growing adoption since its introduction in 2018. Fast forward to 2023, Microsoft announced Fabric, their new and fully integrated data platform. Fabric has received mixed feedback from the data community so far, ranging from positive to negative. This begs the question, does Fabric play nice with the Data Mesh concepts? In this post I wanted to share what I learned from my experience of implementing Data Mesh on Microsoft Fabric. We’ll start with a brief overview of both concepts before diving into how they intersect in practice.
Data Mesh
Often referred to as the micro-services approach to data, Data Mesh is seen as response to the centralisation — monolithic approach — of data efforts organisations experienced over past few years. The concept is motivated by the idea that centralisation causes additional dependencies throughout the development lifecycle of data products, causing unnecessary friction and delay. If you think about it, if every change request of a data product has to be resolved by a centralised team, this can become a bottleneck in which the priorities of the requester and the resolver are not necessarily aligned.
To solve this, Data Mesh provides a set of guiding principles on both the organisational as well as on the technical front:
-
Domain Ownership: Domains within an organisation should own their data (products), end to end.
-
Data as a Product: Data Mesh proposes to see and treat data as a product, designed and developed for downstream use.
-
Self-service Data Platform: Organisations needs a self-service data platform with all tools and support to empower domains and allow them to get to work and develop and expose their own data products.
-
Federated Governance: To manage the independent data product creation within an organisation, it should think about the standards and (automated) enforcement of data governance.
Being a mere set of principles, there is no off-the-shelf solution for a data mesh implementation organisations can buy. The closest you can get, are integrated data platforms such as Databricks and Microsoft Fabric, or adopt a more fragmented custom solution. In this post, we investigate if Fabric is an actual candidate, but before we do I want to dedicate a few words to what Fabric is.
Microsoft Fabric
Microsoft has been investing a lot of resources into supporting data professionals, which led to a wide range of products including the famous PowerBI and tools like Azure Machine Learning and Azure Data Factory. Microsoft Fabric is their latest investment, and aims to bring these tools together and offer unified experiences to their users. These experiences are tailored to Data Science, Data Engineering and Data Analyst that each contain all the relevant tooling within a shared interface.
Microsoft has opted for the Software as a Service strategy for Fabric, meaning organisations can expect transparent and predictable pricing, simple onboarding, and a generous free trials.
The foundation of Fabric is OneLake, marketed by Microsoft as the OneDrive for data. OneLake forms the shared data lake for all data that lives within an organisations’ Fabric environment, and has adopted the Delta Format standard which allows for good accessibility and interoperability with tools like Spark. On top of this, Microsoft is working on making it as easy as possible to get data into your OneLake with features as database mirroring.
This all sounds great, and the concept is beautiful in its simplicity: a one stop shop for all things data at a transparant and predictable price. However, Fabric has been met with mixed feelings by the community so far. A lot of users are commenting Fabric feels like an unfinished product, and have a hard time migrating existing solutions towards it. This is something I’ve written about extensively in my previous post; Microsoft Fabric: The Good, The Bad & The Ugly. On the flip side, the SaaS foundation of Fabric seems to resonate with leadership teams, pushing for the transparency and predictability. The consensus seems to be that Fabric in its current state can be well suited for organisations that are just starting out with data, but probably is not the best fit if you’re looking to move away from an existing solution.
Implementing Data Mesh with Microsoft Fabric
The big question that remains, is: how well does Microsoft Fabric align with the principles of Data Mesh? Can it live up to the expectations set by the paradigm, and is it truly possible to implement a Data Mesh architecture using Fabric? To answer that, we need to look at it from both the technical and organisational perspective.
Technical Perspective
From a technical standpoint, Microsoft Fabric offers many of the core capabilities required to support a Data Mesh architecture:
-
OneLake acts as a unified data lake, enabling seamless data storage and sharing across domains. In our experience, it’s been a delight to work with — especially in terms of improving dataset discoverability.
-
A wide array of integrated tools— such as DataFlows, Data Pipelines (essentially rebranded Azure Data Factory), Notebooks, Azure Machine****Learning, Power BI, and orchestration tools like Airflow — offer everything needed to process, analyse, and deliver data products. If you’re already familiar with these tools, you’ll find them intuitive and well-integrated, although we did encounter a few situations that required workarounds.
-
Workspaces provide a way to isolate domains and data products, helping to establish clear boundaries and promote domain-oriented ownership. Before diving in, it’s important to develop a strategy for how you want to structure workspaces to suit your organisation’s isolation and governance needs.
-
Microsoft Purview brings in federated governance and data cataloging, both critical for discoverability and compliance in a decentralised environment. However, since Purview is part of Microsoft 365 and not natively integrated with Fabric, we faced some challenges around permissions during implementation.
-
Lastly, self-service tooling, template workspaces, and rich documentation of Microsoft help reduce the barrier to entry, allowing teams to independently manage their data products. Template workspaces, in particular, have been incredibly effective in supporting a self-service model.
Taken together, these components make Microsoft Fabric technically capable of supporting the Data Mesh paradigm.
Organisational Perspective
However, technology is only half the story. Data Mesh is not just about tooling and infrastructure — it’s a fundamentally different way of organising teams and responsibilities around data, a socio-technical paradigm shift.
Here’s where things get more nuanced. As mentioned earlier, Microsoft Fabric is often considered a great starting point for organisations that are early in their data journey. It simplifies many aspects of the data stack and provides a unified platform that’s easy to onboard.
This poses a challenge: the principles of Data Mesh demand a relatively high level of data maturity. Decentralised ownership, cross-functional teams, product thinking for data, and federated governance — all of these require a strong foundation in data culture, strategy, and operational discipline.
For organisations still laying that groundwork, Fabric can be a great entry point — but diving straight into a full-scale Data Mesh may be premature.
Conversely, more mature organisations looking to move beyond monolithic architectures might find Fabric a suitable platform for adopting Data Mesh principles, depending on their existing solutions. However, it’s worth noting that Fabric is still a relatively new product and may come with the occasional instability or bug. Testing through Proof of Concepts is strongly recommended before making any major commitments.
Final Thoughts
In conclusion, implementing a Data Mesh with Microsoft Fabric is technically possible and increasingly practical, especially as the platform continues to evolve. But success hinges less on the tools and more on the organization’s readiness — its structure, culture, and maturity with data.
Ultimately, adopting Data Mesh with Fabric is not just a technology initiative; it’s a strategic transformation. For organisations prepared to make that leap, Fabric can serve as a strong foundation to bring the vision of Data Mesh to life.
Hi, I’m Bastiaan 👋🏼 Founder of datalyft, a small data agency helping companies transform their raw data into real value. I write about the Modern Data Workflow, where I explore tools & processes to supercharge your data capabilities. Follow me for more!