Framework for Containers Orchestration to handle the Scientific Workloads using Kubernetes

Abstract

Thelightweight nature of Operating System virtualization give birth to thecontainers, and their high efficiency for application deployment prefers itsusage in Cloud computing. Containers encapsulate and bundle the requireddependencies for application development and deployment in libs packages as asingle entity. The nature of containers eases application migration within theHigh-Performance Computing environment to handle the scientific workloadssmoothly. Singularity containers are aimed to be only for High-PerformanceComputing (HPC) applications. The existing HPC workload managers are good withcontainer monitoring, scheduling, and resource management, but containerorchestration is always a concern. This study proposes a framework that willease container orchestration using Kubernetes to handle the scientificapplications workload. HPC cluster is built using interconnected containers ina private cloud environment. The architecture is derived based on the requiredconfiguration for containers to deploy the application and scheduling jobs.