Fog-Enabled Self-Organizing Networks for Efficient Task Partitioning and Resource Allocation

Cloud servers provide necessary computing and storage resources for IoT applications. However, the remote location of cloud servers is the primary challenge in achieving low end-to-end latency and energy consumption. The fog networks can meet the demand for end-to-end latency and energy-efficient task processing for IoT applications. However, fog networks have limited computation resources and are decentralized compared to cloud servers. This makes the fixed resource allocation strategy challenging in fog networks. Therefore, Self-Organizing Networks (SONs) can provide promising solutions to these challenges of fog networks. This article investigates dynamic task partitioning between fog nodes, which enables self-organizing capabilities in fog networks, mitigates the overloading of a single fog node and utilizes other underutilized fog nodes. First, we provide a decision on task offloading. Then, an optimization problem is discussed, which mitigates the requirement of prior future information on task arrival rate for task partitioning. We determine the optimal task partitioning ratio based on the queue backlog and computation capacity of the fog nodes. Results show that, on average, the proposed strategy reduces the queue delay and queue length by 55% and 17%, respectively, compared to the existing baseline strategy. Specifically, the task partitioning problem with resource allocation is generic and realistic for several latency-sensitive IoT applications.