Practical Approaches to Implementing a Data Mesh
Data mesh offers a new paradigm for fulfilling the promised value of data. It rejects long-standing centralized data architectures such as the data lake and the data warehouse and their associated centralized teams. Instead, it decentralizes both data ownership and the data itself, shifting them toward the functional domains that create and use data to run their business.
Data mesh is designed for the highly distributed and diverse world of modern data environments. Its four pillars (domain ownership, data as a product, the self-serve data platform, and federated computational governance) are the foundation for a very different approach to managing data for analytics.
Many organizations find this approach compelling. They recognize the flexibility and delivery advantages of decentralized domain ownership and data as a product. They see how a self-serve data platform removes overhead and friction for both developers and consumers of data. And they want to mature and expand their data governance practices without slowing down development work or stifling innovation.
However, practitioners implementing data mesh must work within the bounds of what their organization’s culture, existing technologies, and resources will allow. This report explains how data mesh is a journey that unfolds iteratively according to a company’s needs and readiness for the changes that it imposes.
Key Takeaways
> Data mesh distributes data ownership from a central data team and central data systems to business domain teams who create and use their data.
> Data mesh is not for every organization. You need sufficient data size and complexity to justify the investment in a data mesh program.
> Domain ownership and data as a product require a change in mindset. Engineering teams must expand their focus to include developing data solutions for their stakeholders, which is likely to be a significant change in how they work.
> Product thinking applied to data is critical to data mesh. Domain teams must discover stakeholder needs, iteratively develop data products, and manage them over time.
> Our survey of implementations shows there’s no one way to approach data mesh. Each initiative is driven by the organization’s needs and cultural readiness for the changes that it imposes.