Migration of relational databases to NoSQL - methods of analysis

Abstract

The amount of data to store, organize and manage in any organization, is very high and increases every day, fact well-known by companies as Facebook, Google or SAS. With this current growth rate, technologies must adapt to the amount of disposable data, and a new approach to information processing is required. Big Data technologies are more focused, and this is a reason for a greater spread of NoSQL database models. The purpose of this article is to validate the existing (and already used) migration methods and to adapt them, to understand the most efficient method to migrate a relational database to a NoSQL database. We will show the methodology used and what were the steps followed for the implementation, as well as the configuration of the environment used during the tests. Results show that in this migration process, the most efficient method is what is referred to as automatic offline migration. However, it requires a window of unavailability greater than the method of online migration, which in turn requires more resources from the operating system to migrate. Therefore, the most efficient method to migrate a database will depend on the application availability, and the computational resources available for it. We hope to make an important contribution in helping to choose a migration method to use, and the metrics that can be collected to better evaluate the performance of a migration.