Trusted D2D Offloading with Energy Optimization for Workflow Applications in Mobile Edge Computing

Jiabao Fan,Cheng Qian,Zheng Gong,Haoyu Luo

DOI: https://doi.org/10.1109/CBASE60015.2023.10439102

2023-01-01

Abstract:Device-to-Device (D2D) offloading is a promising complement to the Device-Edge-Cloud hierarchical offloading paradigm, which enables end devices to offload their computational tasks to nearby devices with idle resources. However, since the end devices are owned by individuals, people may not be willing to share resources with devices they do not trust. Excessive battery consumption can also dampen their enthusiasm for participation. Unfortunately, existing works on D2D offloading rarely jointly consider trust relationships and energy consumption, especially for the offloading of workflow applications where interdependent tasks are assigned to different individual devices. In this paper, we propose a novel D2D collaborative offloading approach for workflow applications in mobile edge computing environment, aiming to minimize the energy consumption of workflow execution while meeting the on-time response requirement of workflow requests and trust constraint of end devices. Specifically, we first partition the workflow into branches to meet the performance of single device and reduce data transfer among different devices. To achieve effective matching between the workflow branches and end devices, we introduce the concept of multi-hop trust, which helps to build a larger collaboration network with more candidate collaborators. Subsequently, we propose a distributed stable marriage matching algorithm to make the offloading decision based on the real-time information of the devices in the collaboration network, including their idle time, computing and communication capacities. We evaluate the proposed approach by conducting comparative and ablation experiments, and the simulation results demonstrate that our approach achieves the lowest energy consumption while ensuring reasonable response time of workflow requests.