MIMO: a Framework for Extensible and Flexible Intent-based Workload Meta-Orchestration

The widespread adoption of cloud computing and its inter-connectivity with edge systems have enabled the creation of a true Computing Continuum, in which data-driven applications can be deployed dynamically to their optimal placement within the continuum, satisfying user-specified service-level requirements, commonly referred to as intents. In fact, intent-driven orchestration and meta-orchestration are increasingly popular across various scenarios beyond workload management, underscoring its versatility and potential. In this context, the FLUIDOS project aims at the creation of a decentralized meta-operating system integrating Kubernetes' orchestration capabilities with the multi-cloud and multi-cluster abstractions provided by Liqo. This combination allows Kubernetes clusters to seamlessly share computational resources across continuum members. We here present Model-based and Intend-driven Meta Orchestrator (MIMO), a novel approach to achieve dynamic and customizable intent-based meta-orchestration in Cloud-to-Edge Computing Continuum that leverages an ensemble architecture to combine intent-specific model-based and algorithmic-based orchestration capabilities. MIMO embraces the requirement to adapt decision making capabilities to the intents provided by the user.