On edge cloud architecture and joint physical virtual resource orchestration for sdn/nfv

Abstract

5G networks will incorporate virtualized network infrastructures and new technologies such as Software Defined Networking and Network Function Virtualization. The developments have predominantly been made in the core and back-haul network segments. However, the ambitious Key Performance Indicators (KPIs) set for 5G networks will necessitate a renewed focus on the network edge in view of the virtual-ized infrastructure and SDN/NFV adoption. Several challenges are antic-ipated to be addressed for cloud at the edge, prime among which is the heterogeneity that spans, among others, hardware, software, radio, networking and virtualization domains. In this paper, we investigate cloud at the edge architecture, its unique challenges and the technology enablers. We present a prototype heterogeneous edge cloud implementation with focus on joint physical and virtual resource management and orchestration while supporting multi-tenancy for edge services. A scheduling & orchestration module is presented that interfaces with Kubernetes cloud management system to support service deployment per two scheduling polices of load balancing and energy saving. We present details of the considered edge cloud platform, the scheduling & orchestration module and its functions inside the edge cloud. Finally, we present some prelim-inary results and comparisons of the implemented orchestration policies in the context of heterogeneous edge services.