The Evolution of Data Warehouses—From Data Analytics to AI and Machine Learning
When data warehouses first came onto the scene in the late 1980s, their purpose was to help data flow from operational systems into decision-support systems (DSSs). These early data warehouses required an enormous amount of redundancy. Most organizations had multiple DSS environments that served their various users. Although the DSS environments used much of the same data, the gathering, cleaning, and integration of the data was often replicated for each environment.
As data warehouses became more efficient, they evolved from information stores that supported traditional BI platforms into broad analytics infrastructures that support a wide variety of applications, such as operational analytics and performance management.
Data warehouse iterations have progressed over time to deliver incremental additional value to the enterprise with enterprise data warehouse (EDW).
|Provides relational information to create snapshots of business performance
|Slice and dice, ad hoc query, BI tools
|Expands capabilities for deeper insights and more robust analysis
|Predicting future performance (data mining)
|Develops visualizations and forward-looking business intelligence
|Tactical analysis (spatial, statistics)
|Offers “what-if” scenarios to inform practical decisions based on more comprehensive analysis
|Stores many months or years of data
|Stores data for only weeks or months
Supporting each of these five steps has required an increasing variety of datasets. The last three steps in particular create the imperative for an even broader range of data and analytics capabilities.
Today, AI and machine learning are transforming almost every industry, service, and enterprise asset—and data warehouses are no exception. The expansion of big data and the application of new digital technologies are driving change in data warehouse requirements and capabilities.
The autonomous data warehouse is the latest step in this evolution, offering enterprises the ability to extract even greater value from their data while lowering costs and improving data warehouse reliability and performance.
Find out more about autonomous data warehouses and get started with your own autonomous data warehouse.
Data Warehouses, Data Marts, and Operation Data Stores
Though they perform similar roles, data warehouses are different from data marts and operation data stores (ODSs). A data mart performs the same functions as a data warehouse but within a much more limited scope—usually a single department or line of business. This makes data marts easier to establish than data warehouses. However, they tend to introduce inconsistency because it can be difficult to uniformly manage and control data across numerous data marts.
ODSs support only daily operations, so their view of historical data is very limited. Although they work very well as sources of current data and are often used as such by data warehouses, they do not support historically rich queries.