Slicing and Allocation of Transformable Resources for the Deployment of Multiple Virtualized Infrastructure Managers (VIMs)

Abstract

In the context of 5G networks, the concept of network slicing allows network providers to flexibly share infrastructures with mobile service providers and verticals. While this concept has been widely investigated considering mostly the network issues, in this work we focus on a slice as a service model that takes into account the data center (DC) perspective. In particular, we propose an architecture where DC slices are created over transformable (compute and storage) resources, which can be virtualized or de-virtualized on-demand. Then, on top of each slice, an on-demand VIM is instantiated to control the allocated resources. As a realization of this architecture, we introduce the DC Slice Controller, a system able to deploy and delivery full operational VIMs based on generic templates. We evaluate the effectiveness of the proposed system deploying three VIMs (VLSP, Kubernetes, and OpenStack) over commodity hardware. Experimental results show that the DC Slice Controller can timely provide a slice even when dealing with sophisticated VIMs such as OpenStack. As an example, we were able to delivery a fully functional OpenStack in four nodes in less than 10 minutes.