A Dynamic Predictive VM Resource Scaling Strategy in Satellite-Ground Computing Networks

Abstract

Combining satellite-ground network with the edge computing, an emerging research direction is to use low-orbit satellites as edge nodes to provide computing services for ground users and space missions. Due to the motion of satellites around the earth, the ground region covered by the satellite changes constantly over time, and the service traffic also changes accordingly. Therefore, the method of running a constant computing resource will lead to insufficient service capacity or high energy consumption. In this paper, we proposed a two-step dynamic resource management strategy SRTMS, which makes use of the certainty of satellite orbit and historical service data to predict the business traffic of future service region and dynamically scale the amount of in-orbit virtual computing resources. Through the strategy, energy consumption is reduced by 73% compared to the traditional mode in which all resources are operated at full capacity, saving resources that can be used for other payloads.