Auto Scaling in Mobile Edge Computing to Reduce Response Time and Operational Costs

Abstract

Abstract—Internet of Things (IoT) technology is rapidly advancing and evolving. Currently, a wide range of stakeholders are closely studying how to distribute the computational infrastructure and mobile edge computing networks that are activated by cloud computing services. The mobile edge computing (MEC) environment is rapidly becoming a suitable framework for IoT-related infrastructures and mobile devices that include real-time applications. However, in recent years, with the increasing growth of devices and the need to offload tasks to edge servers in the MEC environment, there has been an increase in task response time, to the extent that some tasks may even be abandoned when their deadlines expire. Given that the main reason for the emergence of the MEC environment is to reduce task response time compared to other cloud environments, the issue of increasing response time in this environment is unacceptable. Also, given the limited processing capacity of edge servers compared to other cloud environments, edge server resources must be used optimally, so reducing operational costs must also be on the agenda. Therefore, the main objective of this paper is to reduce task response time and operational costs in the MEC environment. To achieve this goal, in this paper, we propose an automatic scaling method using the LSTM algorithm, which is a deep reinforcement learning algorithm, to resize the number of instances allocated to tasks, so that the proposed approach can effectively and timely maintain an appropriate number of instances in response to changing dynamic traffic, while minimizing response time and also reducing operational execution costs. The present research is descriptive-analytical and relies on library resources.