AMI IECON Edge Management Infrastructure

Automatically deploying applications and updates to the edge is vital to the performance and efficiency of a development team. When new constraints are introduced to the edge architecture, such as device-level zero-trust, applications that are both containerized and native, operating in a private network where ports cannot be opened and static IP addresses cannot be guaranteed, make deployment more challenging. There are many technologies, such as Kubernetes or ioFog, that facilitate the automated deployment of applications and updates on the edge; however, none of these technologies address the unique constraints that the developers at the Ambient Intelligence research lab at Saxion are facing. Thus, updates and new installations are handled manually, which is slow and error-prone. This graduation report presents the research results and design of a new proposed architecture that can automate edge application and software update deployments for private homeowners, while guaranteeing device-level zero-trust security, decentralization, and the installation of both containerized and native applications. Research has shown that common edge orchestration technologies, such as Kubernetes, ioFog, and Docker Swarm, can manage the installation and updating of applications. However, none of these frameworks can install native Linux applications, nor do they implement device-level zero-trust. Docker Swarm appears to be the best technology for this use case, as the current framework the developers use relies on the Docker Engine. To reduce refactoring time, Docker Swarm is the easiest drag-and-drop system to integrate into the existing architecture, which already facilitates all the unique constraints of this domain. Since Docker Swarm is also capable of device-level zero-trust and managing Linux-native applications, a hybrid approach was designed, where the original framework is kept in place and Docker Swarm is added as a wrapper on top of the existing architecture to facilitate centralized container management and log viewing via Portainer. To automate application deployments and updates, Linux services were designed to automate the deployment of both containers and native applications, leveraging the existing framework. This was achieved with a GitOps architecture that utilizes a centralized Git repository with application manifest files, such as ArgoCD, which tracks their installation directives and version. The static public IP address limitations and port requirements for Docker Swarm were overcome by utilizing VPN tunneling for the Docker Swarm overlay network. The designed solution was implemented in a lab environment. During the project’s implementation phase, Docker Swarm, Portainer, and OpenVPN were successfully integrated with the existing edge infrastructure. The results were evaluated using the DORA metrics, with observations made by the developers. The metrics showed a positive improvement in the number of manual steps and the time required for developers to install new applications and roll out updates. In conclusion, a hybrid architecture combining Docker Swarm, the existing infrastructure, and a new system with Linux services and a GitOps approach proved to be successful in enhancing the efficiency and throughput of a development team aiming to roll out updates and install new applications at the edge.