Network Function Virtualization (NFV) brings a cloud service automation paradigm to demand-driven elastic flexing of infrastructure resources. However, it is essential to characterize the impact of hardware and virtualization options on the virtual network function (VNF) performance and load on underlying infrastructure.
In this paper, we present a VNF characterization case study with an open-source VNF platform, Clearwater. We demonstrate that VNF characterization is vital for optimizing VNF performance as well as efficient utilization of infrastructure resources. We use the lessons learned from this case study to design and implement a VNF characterization framework, NFV-VITAL, to automatically instantiate VNF based on user preferences and available resources. We demonstrate how NFV-VITAL can automatically identify optimal configurations under different workloads with Clearwater and the Snort and Suricata intrusion detection systems.
Fig. 1 shows the OpenStack Icehouse infrastructure and fig. 2 shows how Clearwater is deployed on our OpenStack testbed.
Fig. 1: Testbed infrastructure
Fig. 2: Clearwater deployed on testbed
Clearwater case study plots
Click here for more workload and CPU plots of Clearwater VNF case study including scaling out, scaling up and hybrid scaling.
NFV-VITAL sample user input files
Example 1 and example 2 show the user input files for the two demonstrations in section III.B of the paper. Click here for more detailed explanations on user input options.