Why use a columnar database for analytical workloads

In this session, we’re going to discuss how columnar databases the improve performance and efficiency of analytical workloads. We’ll begin by explaining why transactional queries (e.g., return every column in a single row) benefit from row-based storage whereas analytical queries (e.g., return the aggregate of a single column in every row) benefit from column-based storage. We will walk through the storage and query processing architecture of MariaDB AX, an open source columnar database, to show how columnar databases work. In addition, we will show how massively parallel processing, combined with column-based storage, not only improves the performance and efficiency of analytical workloads, but scales to support interactive, ad hoc analytical queries on terabytes of data and billions of rows in real time.