Microservices platforms, such as Kubernetes, have received a great deal of attention lately as they offer elastic and resilient orchestration for services implemented as portable, light-weight, software containers. While current solutions for cloud-based Kubernetes deployments are mainly focusing on maintaining long-running infrastructure, aimed to sustain highly-available services, the scientific community is embracing software containers for more demanding data analysis. To this extent, the pay-per-use pricing model of cloud resources represents an interesting opportunity, and one-off deployments are relevant when running analysis on demand. Hence, in science it has emerged a need for in-cloud one-off setups of systems like Kubernets, which poses a challenge both in terms of deployment speed (as a serious cost impact factor) and in terms of required technical skills. Here we introduce KubeNow, which provides a seamless mechanism to set up ready-to-use Kubernetes-based research environments, aimed to support on-demand scientific workloads. Throughout the paper we present KubeNow design, user stories and deployment time evaluation (in comparison with a well-known Kubernetes installer). KubeNow supports the major cloud providers and it is available as open source: https://github.com/kubenow/KubeNow
View on arXiv