Traditional query processing algorithms are based on "iterator" or "tuple-at-a-time" model where a single tuple is pushed up through the query plan tree from one operator to another. Each operator typically has a next() method which outputs a tuple or record and the latter is then consumed as an input record by the caller operator... Continue Reading →
Why Analytic Workloads are faster on Columnar Databases?
In this post I will briefly summarize why analytic (OLAP) workloads perform better on columnar (aka column-oriented) databases as opposed to traditional row-based (aka row-oriented) databases. Introduction Storage Organization Vectorized Query Execution CPU Cache Friendly Late Materialization Compression Introduction Analytic workloads comprise of operations like scans, joins, aggregations etc. These operations are concerned with data... Continue Reading →