Michele Goetz, vice president and principal analyst with Forrester, describes data fabric as a technology layer that allows for data decentralization. “Data fabric is specifically the technology platform,” she says. “It’s about the tools and capabilities that you’re running your data on, and it’s the ability to take advantage of your data, even when it’s distributed.”
In his personal commentary on the topic, Sandipan Sarkar, a distinguished engineer of hybrid cloud transformation with IBM Global Business Services, describes data fabric as “a design or architectural pattern” that helps to ensure that data is available wherever it is needed.
“It is distributed and heterogeneous architecture,” Sarkar says. “It consists of nodes that can be transactional, such as an application database or a data stream, or analytical, like a data warehouse or data lake.”
Sarkar, who co-authored a white paper on integrating data fabric into a hybrid multicloud environment, adds that data fabric allows for the delivery of data at two levels: how the data is managed, and how the data is integrated. Data fabric, he says, allows data to be managed through a marketplace of information that is acquired based on needs.
“As a product, data is discoverable through a metadata search and knowledge graph. The consumers — inside or outside the enterprise — may come and shop for the product that they are interested in,” he says. “They would not bother with how the data is prepared. They are simply focused on easily consuming it.”
What Does Data Mesh Offer as a Data Management Strategy?
Data mesh, on the other hand, is an application layer on top of data that distributes relevant information to the desired audience quickly, effectively creating a context around the data’s eventual use case.
“Data mesh is an approach that brings process and technology together more easily and effectively to concentrate on the people, process and technology,” Goetz says.
Imagine, for example, an organization with many different departments that are looking to access data internally. The HR department may not have the same needs as the marketing department. A data mesh approach would allow each department to access data resources based on its business needs, and choose those data packages as “products” rather than having the data access being deeply integrated into the architecture.
“Data mesh really focuses on helping you home in on the domain that matters, which is in the context of your business,” Goetz says. “This then allows you to figure out how you shape that data, define it and apply the right policy.”