Data Mesh Architecture – The Tenets of Federated Governance Model (2024)

Data Mesh architecture principles such as Data Domains, Data Decentralization, Distribution Democratization and Productization, have been around for a few years now. And while they encourage innovation and experimentation, they can also just as easily be a cause of chaos and confusion, if not governed appropriately.

In this article, let us delve into the Data Mesh governance models and how federated governance model can help build a robust Data Mesh platform.

But before we get on with Data Mesh Governance Models, let us quickly recap what a data mesh really is...

Large organizations with a centralized data org setup to serve their ever-increasing and varying data needs quickly come to one conclusion – It’s easier said than done! The data team, even with the best of intentions, often become a bottleneck on the growth path.

Enter Data Mesh.

Data Mesh is a paradigm shift in data architecture where Global Data Platform, Domain & Data Products come together. It advocates treating data as a product and advocates domain-oriented, cross-functional teams to create, own, and share data products. By emphasizing data ownership, autonomy, and standardization, Data Mesh improves data quality, accessibility, and collaboration. It puts the onus for data availability, quality, agility and usability right where it belongs – with the data owners themselves.

A typical Data Mesh architecture may look like this. It encapsulates Data Domains, Productization, Self- Service and Governance as core design principles.

One of the cornerstone of Data Mesh architecture is it’s Governance model. It is easy to imagine a Data Mesh implementation devolving into chaos even with the first 3 principles (Data Domain, Productization and Self-Service) in place, if the Data Mesh governance model is inadequately or ill defined.

3 different governance models, namely Centralized, Decentralized, and Federated can be applied to a Data Mesh. As the names suggest,

• Centralized governance model has a monolithic approach to plan, execute and manage the Data Mesh. Platform, Domain and Products are managed by a single org and generally IT driven. While this model carries certain advantages such as easier to adopt standardization, tighter cost control, and minimal duplication among other things, it may stifle domain and business participation and slowdown the adoption.
• Decentralized model is the pole opposite of the centralized model, where every domain, business define and implement data products without central governance oversight. While this model gives greater autonomy to business and product owner to be the master of their domains, it often results in duplication, inoperable and nonstandard platform as a result of minimal coordination between and across domains.
• Federated model, otherwise known as the “Hub and Spoke” model, where “Hub” focuses on the on certain foundational aspects of the Data Mesh and the “Spoke” deals with building and scaling domains, data products and unlocking business value. This model offers the best of both worlds and particularly suitable for large and complex enterprise implementations.

Let us now delve a level deeper into the Federated Governance model.

Federated Governance Model

The Hub

In a federated governance model, “The Hub” establishes the right foundation in terms of data sourcing, processing and storage through carefully crafted architecture, tooling and processes. It devises the right standards and guidelines for implementation of technical and business metadata, DataOps, DevOps, CloudOps etc. The Hub takes on the accountability for implementation and maturing these platform services to ensure uniform application across domains and products. The Hub also defines and institutionalizes the Data Marketplace to facilitate product discovery, subscription, consumption in a uniform and scalable fashion.

How does this help?

Well, a data consumer may require data across multiple domains, and hence may require access to many different data sets with common standards such as tooling, services, methodology etc. Moreover, the hub governance ensures that that overall data platform is robust, reliable, performant, and interoperable as we scale products and services on it.

So, what are the prime accountabilities that fall under “The Hub” governance -

• Architecture, Tooling decisions and platform capability development
• Defining and designing enforcement mechanisms
• Setting enterprise data sharing policies, standards, and processes
• New domain buildouts decisions
• Liaising and fixing accountability for cross domain data product ownership
• Acting as a central governance body to settle issues, provide guidance and facilitation.
• Managing data platform operations, security, ensuring reliable, performant, and scalable services

Who are the primary constituents of the Hub governance board? It is the -

The Spoke

“The Spoke” dominion primarily consists of data domains, data products there in, and responsible for charting the domain data needs, data integration, Data Product design, implementation, management and consumption of data and data products.

“The Spoke” majorly focuses on data value unlock by scaling domains and rolling out Data Products as demanded by business. Business involvement is deep and they are guided by “The Hub” principles as they bring products and services to life.

Some of the important aspects of “The Spoke” responsibilities include -

• Defining the domain data needs, prioritizing data consolidation, ensuring accessibility of data for domain and cross-domain products
• Managing data contracts between data producers and consumers
• Avoiding duplication of products by ensuring proper discoverability, metadata and product information
• Reviewing and providing oversight of data product lifecycle from design to consumption
• Ensuring that “develop once and deploy multiple times” principle is applied when product is being designed
• Integration of Product Driven Data Reliability aspects at Data product level
• Ensuring product is fully supported during integrated data operations

And important stakeholders for the “The Spoke” are the -

Finally, to summarize…

Federated Governance Model plays an important part in implementing a successful Data Mesh solution for large and complex enterprise. When implemented and managed correctly, the Data Mesh platform can achieve the agility, scalability, interoperability that you envisage and ensure no compromise on security, reliability, robustness aspects of the platform.

Accenture’s experience in partnering with our clients in their Cloud Data journeys and implementing State of the art solutions are pathbreaking and across industries. We not only help our clients on Data Platform implementations but also collaborate on cloud data strategy, design, tooling, technology consulting and much more.

Excited to be a part of this journey? Explore fascinating career opportunities atCloud Careers.