An architecture and implementation of automatic network slicing for microservices

Abstract

The concept of "Network slicing" enables us to provide an optimized logical infrastructure for each service. We construct multiple isolated logical infrastructures, slices, on a single physical infrastructure. Each slice accesses appropriate virtual network functions, a logical topology, isolated logical com-putational resources, and isolated logical network resources for the service to be provided. However, designing a service-specific slice is generally complicated. In general, service providers and slice providers are different. Therefore, service providers must design a slice optimized for their service and request slice construction from slice providers; slice providers need to understand a service to design a slice. This creates excessive time and cost overheads. We target automatic network slicing for services from the slice providers' point of view. In this paper, we assume that service providers develop their services from microservices. We show an architecture for automatic network slicing for microservices and implement it. We also discuss the issues revealed by our implementation efforts. This knowledge can be used to realize more general automatic network slicing for other service development methods. Automatic network slicing will provide slices more quickly and cheaply.