Next-generation DevOps for network and compute-intensive applications

Abstract

DevOps has emerged as a critical technology/practice in the era of rapid digital transformation, significantly enabling automation and streamlining the software as well as infrastructure development lifecycles. As next-generation applications demand more scale and distributed management of computing/networking resources, traditional DevOps practices must be enhanced with advanced tools/methods involving emerging technologies. This thesis proposes a novel approach to a next-generation DevOps framework, which is 'applicationinspired' and leverages advances in cloud-based testbed development. The key contributions of the proposed next-generation DevOps framework are: 1) a set of advanced network services involving software defined networking that enhance throughput of the application, while facilitating seamless communication between development and operations teams; 2) a robust security protocol using lightweight Kubernetes for mobile edge cloud computing, which ensures data integrity and system resilience through network microsegmentation; and 3) a highly usable training module for improving knowledge and skills in deploying machine learning based applications using KubeFlow on a public cloud platform. We evaluate these contributions both quantitatively and qualitatively, and demonstrate how our next-generation DevOps framework helps in developing and deploying cutting-edge applications involving video content delivery, drone swarm orchestration and healthcare delivery.