An Attention Mechanism-Based Microservice Placement Scheme for On-Star Edge Computing Nodes

In the context of high-speed networks with 5G and 6G, the influx of user requests under variable usage scenarios puts great pressure on the monolithic architecture, and quality of service (QoS) is gradually not guaranteed. Placing low-coupling, high-efficiency microservices on satellite edge computing nodes with wide coverage is a good solution, but the exponential increase of users and edge nodes accessing communication networks in recent years has gradually highlighted the importance of proper placement and effective management of microservices. The existing studies generally fail to achieve autonomous management of microservices in a variable and complex network environment, and the few studies on autonomous management of microservices are limited to achieving autonomous placement without constraints among microservices. The quality of service and operation cost will not be guaranteed when facing a large number of network requests at the same time. This paper addresses the much-needed problem of modeling microservice placement in satellite edge nodes as a network embedding problem and effectively captures the features that affect microservice placement performance using the attention mechanism in graph neural networks. Simulation experimental results illustrate the effectiveness of the research content of this paper for the automatic management of microservices in satellite networks, while the proposed scheme in this paper performs well in terms of success rate and the benefit-overhead ratio of microservice placement.