Performance Evaluation of Distributed Systems in Multiple Clouds using Docker Swarm

Abstract

The design of distributed systems in multiple clouds have been gaining popularity due to various benefits of the multi-cloud infrastructure such as minimizing vendor lock-in, data loss and downtime. Nonetheless, this multi-cloud infrastructure also poses several challenges such as compatibility, interoperability, complex provisioning and configuration due to the variation in technologies and services of each cloud provider. Consequently, it is a tedious task to design distributed systems in multiple clouds. Virtualization is regarded as the base technology of the cloud and therefore, most cloud-based distributed systems are based on it. Nevertheless, virtual machines require substantial resources and cause several issues across multiple clouds such as provisioning, configuration management, load balancing and migration. Docker Swarm is a container-based clustering tool that resolves some of these issues and supports the design of distributed systems in multiple clouds. It has also incorporated several inbuilt attributes of the distributed system, however, it is still evolving. This paper initially presents the simulated development of a Docker Swarm-based distributed system which can be easily replicated in multiple clouds. Subsequently, based on the simulated Docker Swarm-based distributed system, it performs an evaluation of several attributes of this distributed system such as high availability and fault tolerance; automatic scalability, load balancing and maintainability of services; and scalability of large clusters.