May 21, 2014
For the last couple of years, SDN has been the hottest topic in networking. Over that timeframe we have been constantly bombarded by differing messages about exactly what SDN is, but those messages have had a constant theme: SDN is coming, and it’s coming soon. I don’t doubt that SDN is coming, but given recent events I have to wonder: Has Network Function Virtualization (NFV) become a hotter topic than SDN? And will NFV be here before SDN is?
The initial push to get started with NFV came from service providers such as AT&T and Deutsche Telekom, and these organizations are still some of the primary drivers of NFV. The reason service providers are pushing NFV is simple: if all network functions are available as virtual appliances that can be easily provisioned and integrated regardless of the vendor who provided the appliance or the hypervisor(s) on which it runs, these organizations feel that they can greatly simplify their operations and reduce expense.
In order to bring their vision of SDN to fruition, an Industry Specifications Group for Network Functions Virtualization (NFV ISG) was formed under the auspices of the European Telecommunications Standards Institute (ETSI). I have to admit that when that group was formed I assumed that, based on my experience doing standards work with the European Computer Manufacturer’s Association (ECMA) back in the early 1990s, ETSI would create a long, drawn-out standards process. How slow and bureaucratic was that ECMA work back in the 90s? At one meeting there was a half-hour debate over the proper format for footnotes. I wish I were joking.
My first surprise came when I realized that the ETSI NFV ISG is not going to create standards. In fact, as pointed out in Network Functions Virtualization: Network Operators Perspectives on Industry Progress, “[a]lthough ETSI is a Standards Development Organization (SDO), the objective of the NFV ISG is not to produce standards. The key objectives are to achieve industry consensus on business and technical requirements for NFV, and to agree on common approaches to meeting these requirements.”
My second surprise was that the first meeting of the ETSI NFV ISG group was held in January 2013 and less than a year later (October 2013) ETSI published a set of high level reference documents that are available on the ETSI website. One of those documents describes a set of NFV use cases, and another outlines a procedure for industry participants to influence the work of the NFV ISG to encourage growth of the NFV ecosystem through multi-party implementations of Proof of Concept (POC) demonstrations.
There are currently nine POCs underway, focusing on five of the use cases. One use case in particular caught my attention because I have always thought of NFV being applicable only in centralized locations, such as in a data center, or in a service provider’s central office. There is, however, a use case and a POC that is focused on distributed NFV. As mentioned in the description of that POC, “there is also a need to deploy some functions out at the customer edge — functions that don’t necessarily require the scale and elasticity available in a large-scale centralized NFVI [Network Function Virtualization Infrastructure], and conversely, require highly specialized and often localized configuration and deployment. The ability to support the deployment of virtualized functions at the customer edge requires a Distributed NFV (D-NFV) architecture. D-NFV enables the placement of virtualized network functions (VNFs) throughout the network, where they are most effective and highly customized to a specific application or user.”
I am impressed. The ETSI NFV ISG is moving much more quickly than I thought and it isn’t focused on standards, but on driving POCs and use cases that matter to service providers. It is also creating a somewhat expansive view of what NFV is, and what it could become. In terms of the title of this blog? Yes, I do think NFV is hotter right now than SDN, in part because there has been so much hype surrounding SDN for the last couple of years and so few production deployments that at least some people are growing tired of talking about SDN. Reasonable people can disagree as to whether or not NFV crosses the chasm before SDN does. However, if I were in Vegas and they were giving even odds, I would pick NFV.