Dynamic Network Slicing for Fog Radio Access Networks

Fog radio access network (F-RAN) has been recently proposed to satisfy the quality-of-service (QoS) requirements of the ultra-reliable-low-latency-communication (URLLC) IoT applications, hence fog nodes are empowered with computing and storage resources to independently deliver network functionalities at the edge of network without referring the users to the cloud. However, due to their limited resources, fog nodes should utilize their resources intelligently for low latency IoT applications to leverage the complementarity with cloud computing. We consider the problem of sequentially allocating fog node's limited resources to various IoT applications with heterogeneous latency needs. We formulate the problem as a finite-horizon Markov Decision Process (MDP), and present the optimal solution, known as the optimal policy, through dynamic programming. The fog node learns the optimal policy through interaction with the IoT environment, which enables adaptive resource allocation in different IoT environments. Comprehensive simulation results for various IoT environments corroborate the theoretical basis of the proposed MDP method.