Towards Strategic use of Data Virtualization – Unified View of Data

One of the biggest headaches that big corporations face is how to provide a unified view of data across regions. Most of them have grown as a result of acquisitions of regional companies that in many cases operate independently and maintain their local IT infrastructure.

A very frequent situation is that the representation of their key business entities such as customer, product, order, etc., varies in every geography, causing problems when trying to exploit the information whether for their operational or informational needs. There are data model mismatches, different levels of granularity when representing the same entity, and different interfaces and protocols to access the same kind of information, making difficult to achieve a unified view of data.

To solve these issues data virtualization provides an abstraction layer that keeps a unified and common representation of the business entities, in the form of a Canonical Business Model that can be consumed by all the enterprise applications regardless of the underlying data schemas in every location.

A large European insurance company faced this problem and solved it by creating an abstraction layer with the Denodo Data Virtualization Platform. In their case the problem was exacerbated by the fact they were suffering from various definitions of the same entities across their business lines and branches (car insurance, health insurance, home insurance, enterprise insurance, etc.).

They made use of a top-down data modeling design approach, defining first the canonical views representing business entities such as customer, policy, claim, etc., by means of Denodo interface views. Later they provided an implementation to such views in every location by making use of bottom-up design, starting with importing data from the local data sources, refining and normalizing them and combining the data so as to finally match the top-level business view definition. The process was carried out in a different manner in every location, as the information was stored in different systems (e.g. local CRM, policy management system, claim management system, etc.), but the final result was exactly the same, a unified view of data comprised of tidy views of their key business entities.

As a result, they were able to provide the same kind of information from every location, and fulfill the requirements of both their business units and their corporate functions, in terms of agility and consistency when accessing the information.

The performance is managed and guaranteed by the Denodo Data Virtualization Platform that includes sophisticated techniques to optimize query execution in such a highly distributed environment that minimize the data transfer over the net through intelligent caching and incremental queries, and exploit the processing power of the sources as much as possible.

Once this abstraction layer was in place they realized the significant benefits coming from having a unified corporate data access layer, where they could easily define their business rules and policies for accessing data, enforcing secure and proper access to the information to every application and user. This unified data governance starts from a centralized definition of data to determine the use of the information, the criteria standardization and unification, the key business entities, the application access rules, safety and quality standards, and finally the business rules and policies for accessing the data.

Many people think of data virtualization as a quick way to get access to information that it is spread across disperse data sources mainly to solve tactical needs. However, the strategic use of data virtualization as an enterprise-wide data management tool is well understood by large organizations that make use of this technology to solve their very complex data integration needs.

Anastasio Molano

Comments

  • In the example above, you say “implementation to such views in every location by making use of bottom-up design”. I thought that data virtualization implied that data continues to stay in original structure etc. and the abstraction layer will handle the mapping. However you state that data should be moved / restructured to conform to the CBM, which would entail ETL etc.

    Doesn’t this defeat the premise of Data Virtualization ?

Leave a Reply

Your email address will not be published. Required fields are marked *