Implementing big OLAP Cubes using a NoSQL-Based approach: Cube models and aggregation operators

Abstract

OLAP is a technology that allows answering multidimensional analytical queries in purpose of decision making process, and It is considered as the core concept and a part of a broad trend in business intelligence. OLAP specification is generally platform-independent, although current implementations rely very heavily on relational and multidimensional databases (R-OLAP and M-OLAP implementations). Terms like data warehouse and OLAP cube are related to processing a huge amount of data; Unfortunately, these classical approaches become obsolete with big data phenomenon and cannot scale to meet the need of big data analysis. In this paper, we study the implementation of OLAP cube under key value-oriented NoSQL database, and for this purpose, we proposed two approaches to fetch data from a relational data warehouse and load it into NoSQL-OLAP cube implemented under Oracle NoSQL database following two models: The first one uses the tabular data structure layered on top of key value store, and in the second model, we proposed an aggregation operator to retrieve and load data into the key value storage API provided by Oracle. In order to validate our proposal, we conducted an experiment using Docker Engine and Docker Swarm to deploy Oracle NoSQL as a container in single-node setup and in a cluster mode. Thus, experiment had shown an interesting perspective on using key value store in OLAP engines.