Columnar databases, also knowns as column-oriented databases, are databases that store data on disk as columns, as opposed to rows. Data stored as columns reduces disk I/O and seeks for aggregate functions (sum, average, count, and so on), which can make columnar databases optimal for analytical data and data warehouses. However, operations that operate on entire rows of data are less optimal.
For example, take the following data:
While a relational database would store the above data by row, for example:
1: James Red 2: Kim Green 3: David Blue 4: Liz Blue
A columnar will store the above data as columns:
James: 1 Kim: 2 David:3, Liz:4 Red: 1 Green: 2 Blue:3,4
To retrieve the most popular favorite color in the above data requires fewer steps in the columnar format than in the row format. However, retrieval of a specific record requires fewer steps in the row format than the columnar format.
Deeper Knowledge on Columnar Databases
A columnar data warehouse solution on AWS
Broader Topics Related to Columnar Databases
Organized collections of structured data