Why You Shouldn't Use Postgres or MySQL as Your Data Warehouse

Columnar databases are more suitable for data warehouses than OLTP databases due to their optimization for BI-style queries that involve aggregating large numbers of rows into a few rows, and selecting only a few columns. This is in contrast to OLTP databases which optimize for selecting many columns but only a single row. Columnar databases can be significantly faster (up to 1000x) than OLTP databases for BI-style queries and are designed to scale across multiple nodes, providing additional performance benefits. While it's possible in theory to operate two databases under the hood to support both styles of queries, in practice, it is better to keep them separate due to different availability requirements. Major cloud providers offer columnar data warehouses such as Amazon Redshift, Azure SQL DWH, and Google BigQuery, which are supported by Fivetran.