Multi-Edge Computing Offloading for Ultra-Reliable and Low-Latency Communication

Abstract

Recently, the Ultra-Reliable and Low-Latency Communication (URLLC) has been a hotspot and difficulty in the fifth-generation (5G) communication. With the ever-increasing demands on the accuracy and immediacy of data, the size of computing tasks is growing rapidly, which will exceed the rate of data processing in local devices. To tackle this problem, the Mobile Edge Computing (MEC) systems can provide a flexible and efficient approach to offload the learning tasks from capacity-limited local devices to its connected computation-intensive edge nodes. To provide satisfactory Quality of Service (QoS) for the URLLC tasks, it is significant to investigate task offloading schemes. However, lots of related works focus on the offloading schemes in single edge node cases. To further improve the computing efficiency, we propose a novel framework to jointly optimize the task offloading among multiple edge nodes in this work. Specifically, we at first propose a task division rule based on minimum granularity decomposition. Secondly, we formulate a task offloading optimization problem to minimize the total delay with limited resources, in which both the task queue status in edge nodes and the transmission circumstances are considered. To solve the NP-hard problem efficiently, we prove that the original problem can be transformed into an equivalent mixed integer programming one, and then introduce the Branch and Bound (B & B) algorithm to obtain a global optimal solution. Finally, we provide the simulation results to evaluate our model. The results show that the B & B algorithm outperforms the greedy, average and random algorithms, especially for the cases with computation-intensive tasks, limited number of available edge nodes and unevenly distributed of edge nodes' computation resources.