A Constrained Static Scheduling Strategy in Edge Computing for Industrial Cloud Systems

Introduction

Nowadays, Information Technology and Internet affect everything from communication to industrial (Keshanchi, Souri, & Navimipour, 2017). As the emergence of recent exponentially growing technologies, such as big data (Mourtzis, Vlachou, & Milas, 2016), cloud computing (Varghese& Buyya, 2017), edge computing (Satyanarayanan, 2017), networking, artificial intelligence (Acemoglu & Restrepo, 2018), the Industrial Internet has attracted great interests.

The Industrial Internet is often understood as the application of the generic concept of Cyber Physical Systems (CPSs), within which the information from all industrial perspectives is closely collected, monitored from the physical space and synchronized with the cyber space (Li, Yu, Deng, Luo, Ming, & Yan, 2017). There are many successful applications for CPS in the industrial, especially with the emergence of Industrial Internet of Things (IIoT), it becomes possible to achieve real-time big data collect, storage, access, and processing in the cloud platform (Kaur, Garg, & Aujla, 2018). In IIoT, the data is generated by various sensors, which are distributed in the industrial sites. This data will be on the order of zettabytes in the near future (El-Sayed, Sankar, & Prasad, 2018). Therefore, uploading this massive data to the remote cloud platform for further processing will result in latency issues which affect the overall QoS (quality of service) for various applications in IIoT (Hoang & Dang, 2017; Song, Yau, & Yu, 2017). In addition, considering various types of IIoT devices, cloud computing cannot fully conform with delay-sensitive applications (Malik & Om, 2018). Also, the size of the cloud computing system is restricted due to high costs of communication. Fortunately, a novel computational paradigm called edge computing has emerged, which extends the cloud resources to the edge of network in a distributed way (Yang, Puthal, & Mohanty, 2017; Brogi & Forti, 2017). The emergence of edge computing makes it possible to perform the computations at the edge of network. The cooperation among cloud and edge devices can reduce latency and maintain the QoS for various applications in the Industrial Internet environment and improve the processing capability of system (Munir, Kansakar, & Khan, 2017; Yousefpour, Ishigaki, & Jue, 2017; Ning, Kong, & Xia, 2019). Since many tasks can be completed in edge, the resource management and task scheduling of edge computing have witnessed a boom of development in recent years (Fan, Cui, & Cao, 2019; Shao, Li, & Fu, 2019; La, Ngo, & Dinh, 2018).

This paper considers an edge computing supported industrial cloud system (EC-ICS). Edge computing nodes are usually distributed in the industrial site, so it can reduce latency by assigning the task to edge computing nodes (Shi, et al, 2016). However, the dynamic scheduling strategy is no longer suitable for system requirements. On the one hand, when using a dynamic scheduling strategy, the running time of scheduling algorithm increases as the number of tasks increases. On the other hand, dynamic scheduling strategy means that the same task may be processed on different nodes. So, the system needs to transfer data files frequently between different edge nodes which is obviously unrealistic due to the high costs of communication. In order to overcome the defect of dynamic scheduling and minimize task completion time, the authors introduce the queuing theory and genetic algorithm into task scheduling problem in edge computing and propose a constrained static scheduling strategy which assigns tasks to fixed nodes. The main contributions of this paper are summarized as follows:

• •

  In order to accurately describe the task scheduling and resource management problem in industrial internet, a system model is proposed where the task requests are sent by clients and processed by edge side or cloud side.

• •

  In this paper, the task completion time is described from the perspective of probability according to queuing theory. And the task completion time minimization problem is transformed into an integer non-liner programming problem.

• •

  A multi-elites-based co-evolutionary genetic algorithm (MEB-CGA) is proposed in this paper for the task completion time minimization problem. Experiment results show that the scheduling scheme obtained by MEB-CGA is reasonable and effectively in reducing task completion time. In addition, the MEB-CGA can reasonably balance the task workload on both edge and cloud sides.