Distributed Optimization for Computation Offloading in Edge Computing

Rongping Lin,Zhijie Zhou,Shan Luo,Yong Xiao,Xiong Wang,Sheng Wang,Moshe Zukerman

DOI: https://doi.org/10.1109/twc.2020.3019805

IF: 10.4

2020-12-01

IEEE Transactions on Wireless Communications

Abstract:Edge computing is a promising technology that offers data analysis and computing for Internet of Things (IoT) services at the network edge. It has the potential to significantly reduce the latency and improve the reliability of IoT services by allowing computation workloads and local data generated by IoT devices to be offloaded to edge nodes. This paper aims to develop algorithms for efficient provision of both job assignment and resource allocation for edge computing networks. The main objective is to minimize the long-term average of the response time delay subject to constraints on computation resources and power consumption. We apply a drift-plus-penalty based Lyapunov optimization approach to convert the original problem into an upper bound optimization problem. We then relax the latter to a convex optimization problem. Finally, a distributed algorithm based on branch-and-bound approach is provided and the gap between the distributed algorithm solution and the optimal solution of the original problem is theoretically analyzed. Numerical results based on extensive experiments have demonstrated that our distributed algorithm can achieve the required performance of edge computing that supports IoT systems, under static traffic conditions as well as under dynamic environments with time-varying traffic.

telecommunications,engineering, electrical & electronic