Interactive application deployment planning for heterogeneous computing continuums

Abstract

Distributed applications in industry are modular compositions involving containers, functions and other executable units. To make them deployable and executable in production, they need to be assigned to heterogeneous resources in computing continuum’s, consisting of multiple clouds, devices and other runtime platforms. Existing assignment processes are neither transparent nor interactive. To overcome this limitation, we introduce the Continuum Deployer, a tool capable of reading application descriptions (e.g. Helm charts for Kubernetes), interactively performing and comparing assignment algorithms (e.g. to multiple K8s and K3s cloud/edge/fog/device resources), and exporting deployment files (e.g. Kubernetes manifests). We evaluate the tool with empirical package analysis, exemplary deployments and a synthetic experiment to appropriately address scalability concerns.