What is a Column Oriented Database?

A column oriented database is a database management system that has been specifically built for data warehousing, business intelligence, analytics or any query-intensive application. Simply put and illustrated below, a column oriented database stores data in columns as opposed to legacy databases that typically store data as rows.

This paradigm shift in approach to storing and retrieving data has a significant impact on how fast data can be queried. For example, image a table with 100 columns defined within it. An average query accesses about 20% of the columns within a table, so in our example, 20 columns. In this instance, a row-based database is forced to perform I/O against all 100 columns, resulting in 80% of the I/O being wasted against columns that are not even participating in the query. Conversely, because a column oriented database performs I/O only against the columns that are asked for in the query, there is no wasted I/O and in this example, a column oriented database will do five time less work to arrive at the same answer. This I/O advantage directly translates to improved query performance and because a typical analytics use case may access millions or billions of rows, this performance advantage will tend to widen as the volume of data accessed is increased.