AI-Assisted Synergetic Orchestration Mechanisms for Autoscaling in Computing Continuum Systems

The dominance of microservices-based development frameworks and the development of complex distributed applications, along with the massive production of powerful Internet of things (IoT) and edge computing devices, are paving the way toward the transition from centralized to distributed computing paradigms, where applications are deployed across the computing continuum. In the computing continuum, we consider the management of resources from the IoT to the edge to the cloud part of the infrastructure. To achieve effective management of resources and application workloads, a set of challenges have been identified. These challenges revolve around the need to introduce automation and distributed intelligence characteristics into emerging orchestration mechanisms. To treat this, the design of synergetic or cooperative orchestration mechanisms that take advantage of artificial intelligence techniques is arising as a promising approach. In this article, we describe the design and development of a synergetic orchestration mechanism, powered by multi-agent reinforcement learning (MARL), to guide the autoscaling of distributed applications that are deployed across the computing continuum. A MARL environment and a corresponding agent have been developed and applied in real and simulated environments to guide scaling actions, while indicative evaluation results are provided.