Network-aware container scheduling in edge computing

Kubernetes, as a powerful open-source container orchestration platform, enables the automated deployment and scaling of containerized applications. It has fundamentally transformed the deployment of large-scale applications, providing robust support and infrastructure for the transition from monolithic to microservices architecture. traditional container scheduling algorithms primarily focus on resource allocation. However, In the landscape of Internet of Things (IoT) edge computing, nodes often exhibit diverse network performance due to factors like geographical location and network topology. Meanwhile, modern applications relying on microservices architectures impose higher demands on server network performance. Consequently, scheduling methods aimed at improving infrastructure resource utilization efficiency are gradually becoming less applicable to such complex edge environments. The article proposes a Network-Aware Container Scheduling (NACS) algorithm, based on the TOPSIS method, which dynamically obtains real-time network and resource information of nodes in a cluster. This algorithm aims to provide applications with a high-stability layout characterized by low latency and low packet loss rates, thereby enhancing overall system performance. The experimental results indicate that compared to Kubernetes’ default scheduler, NACS significantly improves the network performance of the system. Furthermore, NACS can increase the throughput of web applications by 26% and the throughput of Redis databases by 29.6%.