Dynamic CPU Frequency Scaling for Efficient Resource Allocation in Heterogeneous Fog Networks for CIoT Applications

Consumer Electronics (CE) devices in healthcare, smart homes, entertainment, etc., generate a substantial volume of computing tasks in Consumer IoT (CIoT) networks. However, CE devices typically have limited computational resources to ensure the minimum latency and energy consumption in task execution. Thus, CE devices can offload their computational tasks to fog nodes, which are the computational resources between CE devices and the centralized cloud. Hence, CE devices can efficiently overcome the dependency on far-located cloud servers. However, the resource limitation of fog nodes imposes new challenges in resource allocation in fog networks. The computational tasks do not always need the total available computational resources of the fog nodes. Hence, in this work, we propose an optimization strategy that dynamically adjusts the fog nodes’ Central Processing Unit (CPU) frequency with variations in the task arrival rate, data size of the task, and CPU word size. An optimization problem is formulated using the proximal operator and the Lagrange multiplier. The numerical results show that the proposed algorithm outperforms the other strategies by 36% at least.