Data Cube
We have written several prior articles about data modeling and data warehousing. These articles include: Dimensional Modeling Part 1; Data Warehousing using Star Schema; Bitemporal Modeling; and Common Challenges with Star Schema. All of our prior articles were concerned with storing and retrieving data from a database, for the purposes of business intelligence. One business intelligence tool that we have not yet discussed is the data cube, also known as an OLAP data cube.
Data cube
One typical implementation of data cubes is via a multi dimensional database (MDB). The data cube consolidates and aggregates relevant data, then allows: drilling down; slicing and dicing; pivoting data to view it from different angles.
Today, cloud computing, in some ways, makes data cubes obsolete. The power and scalability of cloud based data warehousing tools such as Amazon Redshift, Azure Synapse and Snowflake, and columnar databases such as Vertica, ClickHouse and BigQuery allows most business intelligence use cases to be run directly against the full warehouse, rather than against a data cube extracted from the warehouse. Data cubes were common back when CPU performance and memory were limited, because only aggregated data could fit into memory. Now memory and CPU are readily available so there is less need for creating data cubes.
For startups, since you are building from scratch, you are almost certainly better off using a cloud based data warehouse or columnar database rather than building data cubes for your analytics purposes.
Up next
An example of an existing application serving its users by using a data cube.