Containerization for shipping Scientific Workloads in Cloud

Abstract

In Cloud computing environment, researchers are actively looking for opting containerization technology using various tools to achieve better performance in High Performance Computing (HPC) applications execution. In virtualization world, containers are getting more popular and found suitable in comparison of virtual machines as they are giving better performance. Along with agility, support micro services and integrated easily with management and monitoring tools. To have containerization in place, Docker is one of the most suitable open source platforms with Operating System (OS) level virtualization. Singularity is also one of the popular solutions to work with HPC applications. Before directly start using the technologies and tools, they must be analyzed, explored and have a proof of concept for performance first. This paper presents (1) Containers evaluation using open source platform Singularity and Docker (2) monitoring of containers using orchestration and monitoring system Kubernetes (3) feature analysis of scientific workloads using Containers in cloud. Feature analysis reports having performance result sets are primarily aimed for helping the DevOps on making the decisions for choosing the right technology and tools to run their parallel and high performance computing applications.