Ensuring Fairness in Edge Networks: A GNN-Based Media Workload Migration Scheme With Fairness Guarantee

As the number of mobile and IoT devices grows, an edge network faces considerable difficulty in collaborative service provision. Computing and bandwidth resources in edge environments are restricted and unevenly distributed. Besides, video streaming requests from users may vary over time. These factors pose significant challenges for users to acquire a consistent Quality of Experience (QoE). Previous studies concentrate on a resource allocation problem with a single objective: maximizing total QoE for all users. Few studies have focused on a fairness issue for each user's QoE in an edge environment. This work designs a resource allocation algorithm by allocating computing and bandwidth resources fairly for all users. We leverage the relationships between entities in an edge network to model the graph embeddings of each entity using a graph neural network. Our proposed Path-embedding Learning (PEL) method learns from the network topology comprised of links with bandwidth resources and nodes with computing resources to generate feature representations of workload migration paths. Our proposed Fairness-ensure Backward Propagation (FBP) scheme can guarantee fairness for users. Our theoretical derivation reveals that the proposed algorithm achieves fairness with approximate Pareto optimality. Simulation results demonstrate that the algorithm can obtain better QoE than existing algorithms.