Towards Scalable Federated Container Orchestration: The CODECO Approach

abstract

Most container orchestration technologies remain cloud-centric, suffering from limited cross-layer awareness, centralized control assumptions, and a lack of federated abstractions for multi-provider collaboration. This paper presents CODECO (COgnitive De- centralized Edge-Cloud Orchestration), a federated orchestration framework that extends Kubernetes to address these limitations across the Cloud-Edge-IoT continuum. CODECO introduces a data-compute-network co-orchestration model, partition-based fed- eration through application neighborhoods, and AI-assisted decision support, enabling context-aware placement and adaptation of microservices across distributed clusters. A hybrid governance model combines centralized policy enforcement with decentralized execution and learning, preserving operational coherence while supporting far Edge autonomy. Privacy-preserving Multi-Agent Reinforcement Learning provides placement recommendations while maintaining data sovereignty across administrative domains. Validation through the CODECO Experimentation Framework across six heterogeneous partner infrastructures demonstrates that CODECO reduces deployment operational complexity by 87.5% while introducing acceptable resource overhead (+7.23% CPU, +4.7% energy) relative to a vanilla Kubernetes baseline. Federated coordination across geographically distributed clusters and cross-layer energy observability are further validated. The CODECO codebase is published under open-source licenses via the Eclipse foundation.