Data fabric sits on top of existing disparate data repositories to enable companies to extract valuable insights from them and implement business decisions at a faster pace.
Unlike data lakes and data warehouses, data fabric is not designed to collect and store information, and it does not require replicating data or starting from scratch. Data fabric sits on top of existing disparate data repositories to enable companies to extract valuable insights from them and implement business decisions at a faster pace.
As businesses shift to a data fabric approach, the seven pillars that make up the data fabric architecture can be applied incrementally and phased into a company’s data strategy.
1. Data virtualization
For data to be available on demand, it must be virtually integrated from various systems. Data fabric does not replicate data into a new format; it simply transforms data automatically in such a way as to be available in the format needed, regardless of where the data resides or the platform it uses.
Any coding or file information is kept hidden and out of the way of the user. In practice, it’s as if all the data came from the same database. Virtualizing data is five times faster and far less costly than the Extract, Transform, Load (ETL) approach of integrating multiple data sources.
2. Artificial intelligence auto discovery
Artificial intelligence (AI) and machine learning can discover new data sources and make matches between technical data and business terms.
AI auto discovery accelerates the integration of data and reduces the need for human effort, helping companies to scale the approach, gain benefits sooner and act in an agile fashion. AI builds upon its own learning, so as more sources are mapped, the greater the mapping accuracy becomes.
This philosophical term refers to concepts and relationships, and it’s foundational to data fabric. It gives consistent meaning to data. For instance, a customer ontology might include properties such as a person’s name, address and credit card number. Ontologies are reusable, sharable and amendable.
Ontological organization of data provides a codification of knowledge within a domain, e.g., the way a business views its customers. It connects the real data with the questions, without you having to reinvent everything as the data changes or the question changes. This also allows for greater re-use and hence faster scaling.
Companies should start small, building out parts of their ontology with just a few simple attributes. As they become more data-driven or develop additional needs, they can then simply re-use existing ontologies and build upon them.
4. Semantic consumption
This is what enables all the benefits of data fabric — how the data is consumed. A single view of data across systems is made possible through abstracted knowledge graphs enriched with common business ontologies and taxonomies used across the enterprise. As a result, users have self-service access to consistent and real-time or near-real-time information.
This layer also integrates data with controls, processes and policies which in turn can also be rolled out globally, ensuring that the same governance principles are consistently being applied where the data is being used.
5. Unified security and governance
Security policies are defined, administered and enforced centrally. Role-based access controls or attribute-level controls can be set to allow users to access only the data they need to do their job, ensuring data privacy regulations are adhered to at all times.
This function in the data fabric also supports controls across hybrid environments, along with ensuring and maintaining data consistency, data lineage and data quality.
Data owners have complete control over the data they share. Business teams have very granular control over how they access the data. This access can be facilitated in real time, and sharing can start or stop as needed.
Microservices are embedded against the semantic data layer, not against the underlying data. This significantly simplifies adoption as it accelerates building of respective applications and enables portability (e.g., think of a risk scoring application simply ported across geographies by exchanging the underlying semantic data layer).
Business leaders are under pressure to lead their business forward fast. Data fabric helps to deliver short-term results while also contributing to long-term transformational strategies.
In short, data fabric might very well be the key for business leaders to fully unlock the power of data to make better decisions faster.
Often, companies see the value in becoming more data-driven, however they struggle to make this a reality because every time they start from scratch by searching for data, aggregating it, and building respective applications from the ground up without any synergies or scaling effects. Data fabric is a fundamentally different approach that enables organizations to start orchestrating data in a way that allows for maximum re-use.