Products and concepts such as Cisco's NCS, F5's LineRate Systems, Brocade's Vyatta, and Mellanox's OpenEthernet Initiative are bringing the building blocks of NFV to the network administrators.
The above and a plethora of others allow for the construction of Distributed NFV, where features are pushed to the edge onto offload cards, modules and other generic hardware.
Distributed NFV is one of the ideas coming out of the ETSI NFV Working Group. The concept is straight forward: put offload engines such as x86 cards in current hardware and run features there.
Distributed NFV is essentially another way to roll out NFV. Rather than the clusters of multiple functions in one location, it allows for deployment that matches with current architectures.
Distributed NFV also allows for providers to roll features directly to customer sites, if the customer's equipment can handle the function.
Distributed NFV requires some key parts of SDN including traffic steering to make sure that the traffic you want to utilize the feature is sent to the right device.
The above diagram shows how a network operator could utilize offload engines built into their edge devices to add services.
The idea is not new, Cisco has offered cards for it's Integrated Routers for many years, offloading voice, video, and even allowing for virtual machines to be run. Juniper has been pushing Distributed NFV since they joined the ETSI Working Group.
How Network Operators Use NFV
NFV has been integrated into many Internet providers current networks, the providers may not even realize it. Virtualized proxies, firewalls and routers are commonly rolled out into networks. Multi-function line cards are used to extend the capabilities of many high-end routers, providing the extension of available network functions within a network device.
NFV, SDN and terms like Service Chaining are becoming a part of the network vocabulary.
Where NFV is Headed
The NFV ETSI Working Group has published their first
standards suggestions for NFV, vendors have been building and acquiring technology allowing high-performance packet processing on generic x86 hardware. Many-core processors are being integrated into generic x86 servers, allowing for higher performance function virtualization.
Network Vendors are releasing virtualized versions of their router and switch software, allowing network operators and cloud providers to provide on-demand, expandable services to their clients.
Action Item: CIOs should start looking at ways that NFV can help lower their CapEX costs. CTOs can look forward to a plethora of NFV focused virtual appliances, bridging the gap between multi-functional lower performance solutions such as small business routers and expensive high performance hardware appliances.
Footnotes: See Defining Software-led Infrastructure for key disruptive technologies (including SDN) that will allow for a simplified and automated next-generation data center.
Steve Noble is the CTO of Sideband Networks and sits on the board of the Network Device Education Foundation (NetDEF), Steven has more than 20+ years of experience designing and running large networks. Since 1996 he has been heavily involved in writing and executing test plans for networking devices. His professional experience includes VP of Technology at XDN Inc, Technical Leader at both Cisco and Procket Networks along with being a Fellow - Network Architecture at Exodus Communications.