Archived Content

The following content is from an older version of this website, and may not display correctly.

As NFV continues its evolution from lab to proofs-of-concepts (POCs), early field trials (EFTs), and implementation, early adopters are discovering both the power of virtualization and its limits. NFV generates capex and opex savings because its architecture enables open platform ecosystems to replace dedicated hardware with commodity servers and virtualization. But we believe that in order to fully reap the benefits of the technology, organizations have to address the issue of NFV performance.

In theory, NFV’s more agile software-based framework makes it easier to build and deploy new service features, reduce deployment costs for services, and reduce reliance on proprietary devices. In practice, implementing NFV cost-effectively on complex legacy networks requires organizations to address fundamental data plane and packet-processing performance issues when they migrate to NFV architecture. Some early adopters are finding in POCs and EFTs that they must solve networking performance and specifically NFV performance issues before they can move forward.

For example, standard virtualized architectures that may work well for computing applications don't help networking performance. This causes bottlenecks in the hypervisor, virtual switches, and the VMs themselves that can degrade overall networking performance by an order of magnitude compared to physical implementations, according to Eric Carmes, CEO of 6WIND, which addresses the NFV performance issue with advanced packet processing software.

A logical first place to apply NFV is in Layer 4-7 network services — services that are often performance-driven — making it all the more important to address NFV performance challenges from the outset. A number of companies have begun tackling these challenges. Here’s a sample of a few:

  • Qosmos: Qosmos uses deep packet inspection (DPI) and Layer 7 network intelligence technologies to recognize communication protocols and metadata attributes. The company’s DPI module, the Qosmos Service-Aware Module (SAM), is essentially DPI integrated with a vSwitch to provide real-time traffic intelligence and classify flows at the hypervisor level.

Just as with any other promising technology, what the technology is matters less than how well it solves network challenges. By addressing NFV performance challenges early, organizations will more fully see the benefits to their business. Through customer interviews, consulting engagements, and trends on SDxCentral, Roy and I have been working on and hearing a lot about NFV applications, use cases, and deployment scenarios. To learn more about what kinds of NFV apps we’re seeing in the market today, check out our recent post and find out which apps we think will reach broad market adoption first.