Oct 10, 2018

[HDGEM] Normal relational database system is ineffective for many analytical use-cases

Normal relational database system, like Postgres and MySQL, store data internally in row form: all data rows are stored together and are usually indexed by a primary key to facilitate efficient accessibility.
This setup is perfect for most transactional operations, like reading rows by id to display your user profile. However it's ineffective for many analytical use-cases.

In analytics, we're rarely interested in specific rows but in the aggregates of all (or most rows). For example, computing the average age of all of our users, grouped by location.

Performing these queries on classic row-oriented databases requires them to read through the entire database, along with all unused columns, to produce the results.

--
Posted By Blogger to HDGEM at 3/23/2017 11:18:00 AM