Skip to content

Why a column store reads less

A table has 40 columns and 100 million rows. A query reads two of those columns and aggregates them.

Why does a column-oriented store handle this far better than a row-oriented one?