Delay Model-Based Computation Offloading Scheme in Edge Collaboration Framework

Recently, the widespread use of the Internet of Things (IoT) devices has brought people’s lives more convenient However, it has limited computing resources for processing tasks. For sufficient computing resources, a computation offloading scheme has been proposed using an edge server. Nevertheless, when many devices are connected to the edge server, the processing efficiency of the task is degraded. To solve this problem, the computation offloading scheme has been studied on the basis of the edge collaboration framework. The existing computation offloading scheme does not consider other edge servers’ computing resources or communication overhead. It leads to a high completion time and low success rate. In this paper, we propose a delay model-based computation offloading scheme in an edge collaboration framework. First, we determine task offloading based on a probabilistic model and formulate the edge collaboration problem using a delay model. The formulated edge collaboration problem is solved using a greedy algorithm. Second, we allocate computing resources to assigned tasks in the edge server. Computing resources for task processing are calculated based on computation and buffering time trade-offs. Experimental results show that the proposed scheme achieves a high success rate and low completion time compared to the existing schemes.

I. Introduction

Internet of Things (IoT) is widely used by people, which has brought convenience to life [1]. IoT-based applications require different deadlines, bandwidths, and computation performance [2]. However, IoT devices do not provide high computation performance for processing tasks because of limited computing resources. To provide sufficient computing resources, a computation offloading scheme has been proposed using cloud centers or cloud services [3]. Cloud server achieves high computation performance compared to the IoT devices. However, it is located far from the devices, so the communication overhead is high. High communication overhead by cloud affects the task requirement such as Augmented Reality (AR), Virtual Reality (VR), and video analysis. In addition, the cloud server occurs backhaul network congestion caused by many connected devices [4]. To solve this problem, edge servers have been developed.