Network functions virtualization (NFV) promises to reduce CapEx and OpEx, increase service agility, optimize resources, and transform the network into a next-generation platform with carrier-grade services. But creating the right infrastructure to achieve these results requires a complicated series of interdependent decisions. A new white paper from Ixia outlines the main components of NFV infrastructure (NFV-I) and examines configuration options to help users get the most out of NFV.
The ETSI Industry Study Group (ISG) for NFV outlined the main requirements and specifications for NFV-I in its initial 2012 white paper. Since then it has published more than 15 other papers that address different areas of the infrastructure. The NFV-I is made up of the compute platform, network interface cards (NICs), a host operating system, a virtualization layer (hypervisor), and virtual networking technology (vSwitch). On top of this platform, VNFs typically are deployed by virtual machines (VMs) that are assigned CPU, memory, and storage. A virtual infrastructure manager usually manages and orchestrates the system.
According to the white paper, “Examining Factors of the NFV-I Impacting Performance and Portability,” these components are like numerous puzzle pieces that may or may not fit together. With all the different component options available, simply looking at the variables a vendor must test in order to verify their VNFs on several platforms is a significant challenge. Every hardware, software, and configuration option can affect system performance and, in turn, overall service.
To fully reap the benefits of NFV, the infrastructure must strike the right balance between performance and portability of virtualized network functions (VNFs). While it would be natural to have a bias toward maximizing performance, over-tuning for high performance puts more demands on the NFV-I and affects portability. The Ixia paper examines factors that affect performance and VNF portability, and it provides expert tips on how to test and configure the most important NFV-I components:
- Compute platform: CPU/motherboard and storage, CPU pinning, logical processor support, QuickPath interconnect (QPI), packet processing enhancements, acceleration, and storage
- Network interface cards: PCI passthrough, single-root I/O virtualization
- Host OS and hypervisor: Open-source NFV-I platforms, performance curve
Ixia examines each of these NFV-I components so users can effectively develop, test, and achieve predictable performance. The paper explains how defining Key Performance Indicators (KPIs) can give the platform the resources it needs to deliver the required performance and dynamically scale up/down when needed.
Read the white paper to learn more.