Cost Efficient Scheduling for Delay-Sensitive Tasks in Edge Computing System

Edge computing, as an emerging computing model, can offload delay-sensitive computing tasks from Internet of Thing (IoT) devices with limited computing resources and energy to the edge cloud. In the edge computing system, several servers are placed on the network edge near the IoT devices to process the offloaded tasks. A key issue in edge computing system is how to reduce the system cost while completing the offloaded tasks. In this paper, we study the task scheduling problem to reduce the cost of edge computing system. We model the task scheduling problem as an optimization problem, where the objective is to minimize the system cost while satisfying the delay requirements of all the tasks. Then, we prove that the proposed optimization problem is NP-hard. To solve this optimization problem effectively, we propose a task scheduling algorithm, called Two-stage Task Scheduling Cost Optimization (TTSCO). We validate the effectiveness of our algorithm by comparing with optimal solutions. The results show that the approximate ratio is less than 1.2 for 95% of the data sets we use. Performance evaluation shows that our algorithm can effectively reduce the cost of edge computing system while satisfying the delay requirements of all the tasks.

I. Introduction

With the development of IoT technology, the number of delay-sensitive applications (e.g., health monitoring [1], location-based augmented reality games) are increasing rapidly [2]. As the computing resources and energy of IoT devices are limited, many processing-heavy tasks should be offloaded to remote servers to be processed. Cloud computing with powerful computing capacity is considered as a potential way for processing the offloaded tasks. However, due to the long distance between the conventional cloud and IoT devices, sending a large number of tasks to conventional cloud for processing will cause long response time and serious network congestion. To deal with this issue, edge computing is recently proposed as a promising computing model [3], [4]. Edge computing provides an additional layer of computing infrastructure which consists of some servers at the network edge (i.e., base stations). For the computing tasks offloaded from IoT devices, edge computing provides the computing services and returns the results to the devices. In this way, the transmission delay of offloaded task and the traffic load of core network will be greatly reduced. Because of these benefits, edge computing has drawn increasing attention from researchers.