Joint Optimization of Server and Service Selection in Satellite-Terrestrial Integrated Edge Computing Networks

The convergence of terrestrial and non-terrestrial (e.g., satellite communications) infrastructures has been attracting more and more attention from connectivity stakeholders given the forthcoming future sixth-generation (6G) communication paradigm, thanks to ubiquitous coverage and broadcast/multicast capabilities offered by SatCom. With the advent of promising concepts such as the Internet of Space Things (IoST), the potential for applying edge computing in space has become particularly appealing. In particular, in-orbit computing services are provided employing edge servers placed on satellites, which can inherently offer storage and computation capacity. However, an open problem is to find out onboard which satellite edge servers should be placed and how distributed edge computing services could be offered. In this respect, this article proposes satellite-terrestrial integrated edge computing networks (STECN), aiming at minimizing the average response latency and average energy consumption by jointly selecting the optimal deployment of servers and services. First, we propose a general network framework for the STECN. Then, we develop a heuristic algorithm based on K-medoids clustering and non-dominated sorting genetic algorithm-II (NSGA-II) to solve the optimization problem. Specifically, we achieve the optimal placement of edge servers by exploiting the K-medoids clustering method. In terms of service deployment, the conducted performance analysis shows that the proposed solutions can approach the Pareto Front using the NSGA-II technique. Finally, the proposed heuristic algorithm is compared with two benchmark algorithms by means of simulations, verifying the effectiveness of our proposed algorithm. To the best of our knowledge, this article is the first work considering server placement and service deployment in STECN.

I. Introduction

In Order to overcome the digital divide experienced in remote areas such as rural areas, mountains, and oceans, satellite technology has become an attractive solution to connect the unconnected [1], [2]. At the same time, driven by the ubiquitous connectivity and global coverage of the satellite-terrestrial integrated network, edge computing [3], [4] has undergone a shift from terrestrial edge computing to air/orbit edge computing in recent years [5], [6], [7], [8] especially to offer ubiquitous, broadband, and reliable computing services to remote areas. In this perspective, key performance drivers are the placement of edge servers [9] as well as the deployment of computation services [10], as they require a careful design.