Data fabric offers the opportunity to scale data management while sustaining flexibility and agility. It also supports the creation of data products in under a day rather than in three months – but as in so many data management scenarios, the outcomes of implementing data fabric depend on the input quality of the data.
An audience poll during a panel on implementing data fabric at last week’s A-Team Group Data Management Summit London suggested the majority of financial firms do not yet have plans for data fabric, although a substantial number are considering the option, and a handful have implemented solutions or are at the implementation stage.
The Data Management Summit London panel was moderated by Colin Gibson, senior advisor and regional advocate at the EDM Council, and joined by expert speakers Peter Jackson, director at Carruthers and Jackson, and Ben Clinch, principal enterprise architect, information architecture and data governance at BT.
How data fabric differentiates
The speakers noted data fabric’s differentiation in being able to handle huge volumes of data. Putting it in the data management picture, Jackson said data warehouses use a technology approach, while data fabric uses an architectural approach including metadata to manage data at scale, find data easily, provide an understanding of the data, and help users get to value quickly. He added: “Data fabric saves spend on technology as it works well across a federated business and doesn’t require a massive data warehouse to be built. It also supports production of new products in under a day, rather than in three months.”
Describing data fabric as a ‘logical data warehouse’ that uses automation, architectural constructs and the disciplines of data management to unlock the value of data, Clinch said: : “Data fabric provides smart, automated data governance at scale. You can pull data from any organisational store and curate data for specific domain areas such as customers or network traffic. It works best where data is a viewed as product.”
Touching on data mesh, Jackson said it uses a more conceptual approach that drives data governance and ownership to SMEs. Clinch commented: “Ideally, you need to wrap data fabric around data mesh to avoid the risk of a data mess.”
Challenges of implementation
The speakers cited the challenges of implementing data fabric as a tendency for vendors to overplay the capabilities of their solutions, winning management buy-in, and data quality.
To resolve these problems, they suggested checking vendor roadmaps and looking for open standards and interoperability, and getting started on data fabric by first socialising the concept. CDOs can then tell the story of why to implement data fabric using related use cases and potential added value, and collaborate with CTOs, who will be pleased that they are not being asked to build solutions, but to govern and manage data. CFOs will be equally pleased that they do not need to invest in a new data warehouse. Data quality is a persistent problem that must be addressed if data fabric is to be used to best advantage.
Subscribe to our newsletter