Data Model is necessary to convert the business specific functionalities to a structured form. It is the backbone of any applications or data repository to hold the data for day-to-day operation, tactical and strategic need of business. A good design in data model would bring scalability and robustness of a solution. A data model is the skeleton of the data warehouse. A robust design in data model brings the scalability and flexibility to the data warehouse. Typically, there are two different modeling techniques heavily used in the data warehouse. Based on Kimball’s bottom-up architecture, the dimensional data model is used to support analytical reporting requirements and based on Inmon’s top-down architecture, the normalized data model is used for the data warehouse to consolidate enterprise data in a single integrated design. While the normalized design is heavily based on the business process, the dimensional data model design is based on Key Performance Indicators (KPIs) requirements.
Many industry standard data models are available such as FSLDM, IFW, OFSA, etc. mainly focusing at the enterprise requirements. These are mostly in normalized data models and are suitable for Inmon’s suggested enterprise data warehouse. These are cost heavy, technology specific, and need heavy customization before implementing for a specific data warehouse solution. For Ex: FSLDM is in Normalized structure and not suitable for providing analytical business insights. Similarly, OFSA is suitable for Oracle specific products. Because of their enterprise focus, these cannot be used for a specific module or business subject area.
In recent time, most of the organizations are implementing Data warehouse in agile approach. Coforge has experienced with such implementation and has followed Module and sub-module wise solution in each sprint. By following the available industry standard data model, it is quite difficult to implement the data warehouse using agile approach. Due to this reason, there is a need of a data model which is easy for customization and is easy for deployment. Coforge suggests a technology agnostic, highly modularized industry standard data model which is a right fit for the warehouse implementation.
Apart from the above drivers, Coforge considers following parameters to bring a quality data model and its associated solution components.
Coforge follows certain best practices while designing these models like: Process oriented approach against Data centric or Source centric approach. Below are the salient features of this framework built from vertical specific LDM:
While the pre-defined data model brings the quality solution to the data warehouse implementation, it saves time and cost to the implementation to a large extent. It is solely dependent on the coverage of the data model based on the business requirements. There is no ideal solution to address 100 percentage of the requirements as most of the business processes are very much localized to the specific organization and the local compliance rules. The difference of the requirements from the available functionalities need customization / modification in the existing design. Based on the previous implementation, Coforge finds there is 30 to 50% of cost reduction to the implementation by following the business data model with associated components.
LDM is built while analyzing the business / process specific functionalities. At higher level, it is at subject area level connectivity aligning to the process flow or the way different processes are interdependent of each other. By drilling down the higher level details to granular level, the subject specific data model would be entity relationship, then the entity configuration to highlight how the attributes are configured within the entity. The business definition, data dictionaries, business rules are critical while defining the data model. The functional domain experts provide the details to the data modeler to work on the LDM. Later, the LDM is converted to PDM for a specific database by understanding the availability of datatypes, data lengths, indexing techniques, partitioning techniques. This is how a data modeler play a vital role in bringing a data model by understanding both the business and technical aspects.
Following are key considerations for building the logical data model:
Organizations for a specific business domain operate more or less in same way and hence, their business functionalities are same. Even though the requirements at business users level could be different, they align to same set of industry standard KPIs. Building the data model for those KPIs would be applicable to those organizations and would address most of the requirements. A business data model brings following benefits: