Cost-Minimized Partial Computation Offloading in Heterogeneous Edge-Cloud Systems

Nowadays, Internet of Things Devices (IOTDs) support numerous applications that require extensive computational resources and are sensitive to delays. Nevertheless, IOTDs are constrained by limited computational power and battery life, preventing them from processing all tasks in real-time. Computation offloading provides a solution to these problems where IOTDs can offload part of their tasks to edge servers for execution. Small Base Stations (SBSs) are located closer to IOTDs, which act as edge servers. However, SBSs have limited computing resources compared with a Cloud Data Center (CDC). Therefore, a heterogeneous edge-cloud system is often deployed in urban areas for computation offloading. In addition, attaining the lowest cost in such a system while satisfying the delay requirements of tasks presents a significant challenge. In this work, a computation offloading strategy aimed at minimizing the overall system cost is proposed. Initially, an optimization problem with real-world constraints is defined, leveraging the hybrid architecture as its basis. Then, a novel swarm optimization algorithm named Grey wolf optimization embedded with Simulated annealing and Genetic learning (GSG) is proposed to solve this optimization problem. GSG optimizes resource allocation and task offloading among IOTDs, SBSs, and CDC. Simulation experiments involving real-life tasks demonstrate that GSG achieves a significantly lower system cost than its existing peers.