An Experimental Analysis of Different Key-Value Stores and Relational Databases

Abstract

Nowadays, databases serve two main workloads: Online Transaction Processing (OLTP) and Online Analytic Processing (OLAP). For decades, relational databases dominated both areas. With the hype on NoSQL databases, the picture has changed. Initially designed as inter-process hash tables handling OLTP requested, some key-value store vendors have started to tackle the area of OLAP as well. Therefore, in this performance study, we compare the relational databases PostgreSQL, MonetDB, and HyPer with the key-value stores Redis and Aerospike in their write, read, and analytical capabilities. Based on the results, we investigate the reasons of the database’s respective advantages and disadvantages.