Offloading Demand Prediction-Driven Latency-Aware Resource Reservation in Edge Networks

The wide applications of edge computing has brought dawn to terminals with limited computing resources and energy supply. Terminal completes its task through computing offloading to reduce energy consumption and improve performance. In order to improve the efficiency of resource scheduling and utilization, this article proposes a delay-aware resource reservation strategy based on the prediction of spatial–temporal correlation task offloading demands. Due to the strong time-varying characteristics of terminal task offloading demand, this article designs a spatiotemporal task offloading demand prediction model (STOD). It divides the region into multiple subregions and models them as a graph structure so as to predict the task offloading demand of each region separately by considering the complex spatial and temporal dependencies of regional task offloading demands. With this model, we propose a regional edge server resource reservation (ESRR) algorithm to minimize the terminal task offloading delay. The experimental results based on the real data sets show that ESRR can reduce the average offloading time consumption by 30% under different scenarios and verify its feasibility and effectiveness.