Our recent blog series on the data integration portfolio introduced a variety of new architectures that help the enterprise manage their data resources, including replication, virtualization and cloud data warehousing. Organizations are now able to integrate multiple data management solutions to address a variety of business sources and requirements. But it is important to understand that the foundation of any enterprise data management portfolio remains the same . . . a roadmap to data management must be created that is independent of the underlying technology. This series of blogs will examine the three main elements of the data integration roadmap: the logical data model, master data management (including metadata management), and the analytical data model.
At its core, the logical data model (LDM) is a business model. It divides the business into entities and their relationships. Customers, vendors, products . . . all are example of common business entities. A customer’s location (which may be a state, region, etc.) would be represented as a relationship between the entities of Customer and Location. Note that the Location entity may also have a relationship with Vendors, Services, even Employees. These standardized and reusable entities are what become conformed dimensions in the enterprise data model (this will be explored further in master data management). Entities have attributes . . . for example, the Customer entity may have the attributes of name, customer number, phone number, etc. It is easy to see how these correspond to columns in a table, but we are not talking about the physical database at this point. We are merely creating a map that will drive how the physical model takes shape.
LDM’s may not be needed or suitable for all projects, particularly agile projects. In its place, data modelers may employ a conceptual data model that is a more abstract representation of entities and relationships, devoid of any detailed notation. These abstractions are in the form of objects and roles, their descriptions in pure business language. Conceptual data models are useful in requirements-gathering involving non-technical business stakeholders . . . the disadvantage is that a LDM is often still needed after the fact to provide mapping guidance for the data modeler.
In today’s modern data warehouse environment, with abstracted virtualization layers and cloud-hosted compute centers, are LDMs still needed? Yes, the business needs of enterprise data still require an understanding on how to structurally present that data. Data virtualization is little more than a logical data model over a staged physical data architecture. Cloud data warehouses such as Snowflake provide compute resources for querying that are designed via a logical model. In the end, the logical model is a business model, independent of data architecture, and it must be approached as a business solution.
In the next installment of the Data Integration Roadmap we will look at the role of master data management and metadata in governing the enterprise data foundation.
About the Author
Joe Caparula is a Senior Consultant at Pandata Group with a focus on delivering data architecture projects for our clients across several industries.
