GNN-Based QoE Optimization for Dependent Task Scheduling in Edge-Cloud Computing Network

With the increasing diversity of user demands for network resources, efficient and flexible task scheduling schemes have gained greater importance. Given that existing works about dependent task scheduling primarily focus on optimizing QoS objectives without considering the impact of user preferences on decision results, and the majority of prior research neglects the underlying relationships among dependent tasks. In this paper, we introduce a Graph Neural Networks (GNN) based dependent task scheduling algorithm (GDTA) to enhance user satisfaction and propose a QoE model to assess user-centered quality of experience. This novel approach incorporates GNN for the purpose of generating embeddings for tasks and networks, leveraging its inherent capability in extracting graph-based features. Compared with baseline algorithms across diverse task parallelisms and network topologies, our method achieves higher QoE scores and shows superior stability and generalization on unseen graph datasets.