Extending the Kubernetes Platform with Network-Aware Scheduling Capabilities

Kubernetes is today the de-facto standard container orchestration platform for the lifecycle management of microservices-based applications on Cloud environments. However it is not yet ready to be adopted on node clusters distributed in the Cloud-to-Edge continuum. In particular its scheduling strategy is not suitable for the placement of modern latency-sensitive applications on Edge environments, characterized by frequent node failures and network partitions. In this work we propose a network-aware scheduling extension of the default Kubernetes scheduler that takes into account the ever changing infrastructure network conditions and the dynamic communication interactions between microservices, with the aim to optimize the placement of application containers.