Cloud-Native Application Validation & Stress Testing through a Framework for Auto-Cluster Deployment
| Autor: | Nikolaos Astyrakakis, Ioannis Kefaloukos, Charalabos Skianis, Evangelos K. Markakis, Yannis Nikoloudakis, Evangelos Pallis |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
Computer science
Process (engineering) business.industry Distributed computing 020206 networking & telecommunications 020207 software engineering Cloud computing 02 engineering and technology Stress testing (software) Automation Toolbox Upload Software deployment 0202 electrical engineering electronic engineering information engineering Cluster (physics) business |
| Zdroj: | CAMAD |
| Popis: | The proliferation of cloud-native applications (applications built specifically for the cloud) has taken cloud computing to another level, but it also revealed several issues such as the lack of automation and complexity of deployment. Extensive research on the subject, revealed many endeavors towards that end. Nonetheless, there are still issues to be solved, such as the cluster monitoring and the cluster deployment automation. Most of the existing solutions, are semi-automated or completely manual approaches, with some of them targeting specific cloud providers/vendors. Furthermore, there is also very limited research that tackles the validation of such applications. Untrusted developers, are able to develop and upload applications to marketplaces, without being audited and verified. Towards addressing the above-mentioned issues, this paper presents a novel, completely automated tool for deploying and monitoring a Kubernetes cluster over OpenStack. Additionally, we propose a tool that provides automatic validation of cloud-native applications. The evaluation of the proposed toolbox resulted in the deployment of Kubernetes clusters with remarkably low overall times, compared to other, manual approaches. The validation process lasted approximately 11 minutes for a containerized application with the Kubernetes Horizontal Pod Autoscaler (HPA) enabled and approximately 3 minutes for a containerized application with the Kubernetes HPA disabled. These overall times are relatively shorter than several other non-automated approaches. The afore-mentioned overall times, are analogous to the underlying hardware and network resources of our test-bed. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|