A Framework for Cognitive, Decentralized Container Orchestration

Abstract

This perspective paper introduces a novel framework for container orchestration called CODECO. The CODECO orchestration framework relies on a data-network-computing approach to define the best infrastructure that can support the operation of next-generation Internet applications across a mobile, heterogeneous Edge-Cloud continuum. The selection of such an infrastructure is aligned with target performance profiles defined by the user, such as resilience or greenness. CODECO proposes to rely on decentralized Artificial Intelligence approaches to provide the most suitable infrastructure to an application deployment, considering infrastructural challenges, such as intermittent connectivity and node failure. This paper explains the current CODECO framework and gives insight into operational use-cases where CODECO is being deployed, as relevant examples of application for such a framework. Recent developments in the creation of the open-source CODECO framework are described and explained, allowing the use of the framework by the research community. The paper then provides a thorough analysis of CODECO’s features in comparison with existing orchestration frameworks, explaining the benefits introduced with this dynamic orchestration approach.

Keywords

• Edge-cloud
• orchestration
• Kubernetes
• AI/ML
• heterogeneous networks
• data observability