Virtualization-based Evaluation of Backhaul Performance in Vehicular Applications

TL;DR

This paper evaluates the performance of SDN/NFV-based vehicular networks using virtualization tools, focusing on delay, overhead, and energy consumption to determine their suitability for safety-critical applications.

Contribution

It presents a virtualization-based emulation framework for assessing backhaul performance in vehicular networks with real-world mobility data.

Findings

SDN/NFV networks can meet vehicular safety delay constraints

SDN controllers add moderate energy overhead

Controllers significantly increase processing delay

Abstract

Next-generation networks, based on SDN and NFV, are expected to support a wide array of services, including vehicular safety applications. These services come with strict delay constraints, and our goal in this paper is to ascertain to which extent SDN/NFV-based networks are able to meet them. To this end, we build and emulate a vehicular collision detection system, using the popular Mininet and Docker tools, on a real-world topology with mobility information. Using different core network topologies and open-source SDN controllers, we measure (i) the delay with which vehicle beacons are processed and (ii) the associated overhead and energy consumption. We find that we can indeed meet the latency constraints associated with vehicular safety applications, and that SDN controllers represent a moderate contribution to the overall energy consumption but a significant source of additional delay.