Energy-Efficient Federated Learning Scheduler for VM Allocation in Edge-Cloud Continuum

As the world becomes more connected through smart devices and the Internet of Things (IoT), managing how tasks are scheduled and executed across cloud and edge environments has become increasingly complex. The need to efficiently allocate resources, such as virtual machines (VMs), while keeping energy consumption low is critical for both performance and sustainability in such distributed systems. Recently, the integration of Federated Learning (FL) with energy-aware scheduling approaches is drawing attention as a means to address these challenges. FL empowers collaborative intelligence across edge and cloud nodes without centralizing sensitive data, thereby enhancing privacy and responsiveness. When combined with energy-aware schedulers, FL makes it possible to dynamically assign tasks and virtual machines on real-time assessments of resource availability and energy constraints, taking into account the diverse capabilities and limitations of participating devices and networks. This synergy enables the development of intelligent scheduling mechanisms that balance computation loads, optimize communication, and prolong the operational lifespans of edge devices, all while adapting to continuously changing system conditions. This research demonstrates that FL-based, energy-aware schedulers for VM allocation can significantly reduce task delays, conserve energy, and enhance overall service quality in heterogeneous edge-cloud. This work proposed an enhanced Federated Learning based Energy-Aware Scheduling (FL-EAS), which reduces latency as well as conserves energy while meeting the SLA rate. This work achieved a total of 95.17% SLA success rate. In terms of accuracy of the federated learning model convergence, this work achieved 99.91%.