Availability and reliability study of NoSQL data stores on commodity hardware

Abstract

Modern application development and the delivery of these applications have changed drastically during the last few years. Applications are deployed on every mobile device to cloud devices hosted on servers and because of this change, users expect much faster response times from servers. To determine which data store, to store information and which data structure to choose, for a high availability and scalable architecture, is still a challenge for developers. Modern applications need to follow the reactive manifesto approach to be more responsive, elastic, resilient and message-driven to be classified as a failure-tolerant system. Four NoSQL categories were chosen to be studied using a common programming language driver. Our research strategy conducted an experiment and this work followed an experimental design approach to send objects using Create, Read Update and Delete (CRUD) operations to measure the read metrics and write metrics per data store. Our research results showed which NoSQL database can be used as read model and which database as write model for a Command Query Responsibility Segregation (CQRS) application, using the reactive manifesto approach.