Converter: A CEAML Reasoner Python package to Streamline Orchestration Across Cloud and Edge Continuum

TL;DR

The paper tackles the challenge of translating CEAML, a TOSCA extension for cloud-edge modeling, into concrete deployment configurations for Kubernetes and Kubevirt. It introduces Converter, a Python-based CEAML reasoner that outputs deployment, termination, and scale-out plans, enabling both single-cluster and multi-cluster orchestration workflows. The approach provides structured inputs (e.g., CEAML model path, registry tokens, cluster IDs, external IPs, GPU inventories) and generates comprehensive artifacts such as namespaces, secrets, PVs, and VM deployments. The work demonstrates practical impact through the ACCORDION platform, showing autonomous, cross-cluster orchestration across edge and cloud environments, thereby simplifying and accelerating CEAML-driven deployments.

Abstract

In recent years, there has been a concerted effort in both industry and research sectors to innovate new approaches to DevOps. The primary aim is to facilitate developers in transitioning their applications to Cloud or Edge platforms utilizing Docker or Kubernetes. This paper presents a tool called Converter, designed to interpret a TOSCA extension called CEAML and convert the descriptions into Kubernetes or Kubevirt definition files. Converter is available as a Python package and is recommended for use by orchestrators as an auxiliary tool for implementing CEAML.

Figures (5)

• Figure 1: Converter input and outputs
• Figure 2: Code snippet to provide access tokens and CEAML models
• Figure 3: Code snippet for deployment plans
• Figure 4: Code snippet for termination plans
• Figure 5: Code snippet for scale out plans