Data Mesh — a not so complicated explanation

Biju Krishnan @ DataSiens
4 min readJul 17, 2023


Big data is transforming businesses — but only if you can find, understand, and effectively leverage your organization’s data assets. The rise of data lakes, warehouses, and other repositories has led to organizations amassing vast troves of data. However, data often ends up fragmented in silos, with different teams owning different datasets. Accessing and making sense of relevant data becomes challenging.

Enter the “data mesh” — an emerging architectural paradigm for decentralized data management. The data mesh aims to enable self-service access to reliable, trustworthy data products. Rather than having a centralized data team control and gatekeep everything, ownership is distributed across domain teams closer to the source.

Chaos vs. Harmony

Data Governance in the past — everything centralised via the Central Data Team

Without a data mesh, data workflows often resemble a game of telephone. The analytics team might request data from the central data team, who then has to track down the app team who generated the operational data, and so on. Bottlenecks abound, innovation lags.

The data mesh flips the script — empowering teams to own, manage, and serve up their own data products. This puts data producers and consumers in direct contact for faster, higher-quality collaboration. No more waiting around for approvals or documentation.

Four Pillars of the Data Mesh

The data mesh stands on four foundational principles:

Foundational principles of Data Mesh

Domain-oriented decentralization

Domain ownership

Rather than a single centralized data team, ownership is distributed across product domains and consumer domains. This breaks down data silos and bottlenecks.

Data products mindset

Data as a product

Data assets are developed and managed like products — with a product owner responsible for maintaining their value and accessibility. This promotes discoverability and accountability.

Self-serve data infrastructure

Self-serve platform

The central data platform team provides the tools and building blocks for domain teams to assemble, manage, and serve their own data products. Access is frictionless.

Federated Computational governance

Policies and standards are enforced automatically via the platform. No human gatekeeping required. This maintains integrity while keeping innovation humming.

Data Catalog Creates Harmony

Data Catalog

A data catalog sits at the center of the data mesh, acting like a card catalog for data. Users can easily search for, understand, and request access to available data products. Descriptions, previews, and metadata offer insights into the data.

The catalog also enforces critical data governance, automatically applying tags and usage guidelines. With all data assets inventoried and governed in one place, data harmony flourishes across the enterprise.

Hands on lab for experiencing the Data Catalog

If you are willing to spend 2–3 US$, I highly recommend following along the Qwicklabs on Google Data Catalog. Click on this link to enroll.

Conducting the Data Orchestra

Implementing a data mesh requires cultural change, as roles evolve from centralized gatekeepers to empowered self-service owners. But the payoff is data that flows freely across domains, underlying analytics-driven innovation and decision making. Data transforms from mess to mesh.

If you’re interested in learning more about data architecture, I encourage you to enroll in my course on Udemy. The course covers all aspects of data architecture, including a deeper understanding of the data mesh.

Data Architecture for Data Scientists on Udemy

The data mesh promises to resolve longstanding data access bottlenecks. Are you ready to decentralize data and achieve data harmony? Let me know your thoughts in the comments!

Next up we shall look at the role of streaming data in data science and explore some use cases which benefit from streaming or real time data.



Biju Krishnan @ DataSiens

I have over 20 years of experience in helping enterprises manage data, and more than half of this in building scalable platforms for analytics, AI and ML.