Software Defined Networking has been a hot subject for a few years now, and from the beginning the discussion correctly pointed out that a key characteristic of SDN is the separation of at least some of the data control plane from the data forwarding plane. At other times, however, the discussion seemed to mistakenly equate SDN and OpenFlow or SDN and network virtualization. And, as so often happens in this industry, the initial SDN articles were flagrantly optimistic in terms of when the tech would be widely adopted.
We recently completed a survey of 246 IT professionals – The 2015 Guide to SDN and NFV – to get a better handle on what’s driving interest in SDN, the opportunities and impediments, and where, how and when companies plan to pursue the technology.
Regarding when they might pull the trigger, survey respondents were given a multiple choice question and asked to indicate the alternatives that described their approach to implementing SDN. Multiple answers were allowed, and their responses are shown in Table 1, as are the responses to the same question given a year earlier to a similar group of respondents.
Table 1 indicates that while the utilization of SDN in production networks remains limited, it has increased somewhat significantly in the last year and should increase somewhat significantly again next year. Which begs the question, what is driving the interest? Where will SDN be deployed? How will it be deployed? When will it be widely adopted?
What’s driving the interest in SDN?
Survey participants were asked which opportunities they thought SDN could address (see Table 2). Quite a few, as it turns out. IT is generally optimistic about SDN, but interestingly enough, relatively few believe SDN will help them reduce CAPEX or reduce complexity, which were early expectations.
Identifying the drivers of SDN is key to understanding its status, but so is identifying the impediments that stand in the way of broad adoption. With that in mind, survey respondents were given a list of impediments and asked to identify the two that would be the biggest inhibitors to adopting SDN sometime in the next two years. Given the state of the industry, it isn’t surprising that the immaturity of the available products topped the list (Table 3).
Obviously, the maturity of the products and enabling technologies will improve over time. However, some of the other key inhibitors won’t just naturally disappear. These inhibitors, which need to be aggressively addressed by vendors and network organizations, include concerns about how to integrate SDN into legacy infrastructure and lack of a compelling business case.
How will SDN be deployed?
While the initial discussion of SDN focused on implementation in the data center, the scope has now widened. In order to understand where SDN will likely be implemented, survey respondents were asked, “If your organization is likely to implement SDN sometime over the next two years, where are you likely to implement it?” Their responses (Table 4) indicate that the primary focus is likely to be in the data center, however there is considerable interest in the WAN as well as in branch and campus networks.
Respondents were also asked to indicate how pervasively deployed SDN would be in their campus, WAN and data center networks three years from now. Their responses (Table 5) show marked inroads in the data center, but also significant deployment both in the WAN and in campus networks. Given the current status of SDN deployment (Table 1), these penetration rates seem overly optimistic.
How will SDN be implemented?
One of the key architectural differences between vendor SDN products is how they are implemented. The overlay model proposed by companies such as VMware and Nuage Networks, focuses on the hypervisor and tunneling and encapsulation. The fabric-based underlay-based model focuses on a range of virtual and physical network elements and relies on the SDN controller manipulating flow tables in the network elements. This is the model used by companies such as Cisco, NEC and HP.
Survey respondents were asked what relative value the overlay- and the underlay-based models will provide over the next two years. Their responses (Table 6) indicate, by a small margin, IT shops perceive the fabric-based SDN model will provide more value. However, many have yet to form an opinion.
Another one of the architectural distinctions between the varying ways vendor implement SDN is the role of dedicated hardware. Vendors such as VMware believe it’s possible to fully support network virtualization in the data center without using dedicated hardware. In contrast, vendors such as Cisco believe dedicated hardware is needed in at least some instances. The survey respondents were asked if they believed that, with the current technologies and products, it’s possible to broadly support network virtualization in the data center without using any dedicated hardware? Those that believed it was not possible outnumbered those that believe it is by almost a 2:1 ratio.
At one time many equated SDN and OpenFlow. To understand how IT organizations currently view that relationship, survey respondents were asked how likely it was that OpenFlow would play a role in their SDN implementation (Table 7). Most seem to have a favorable view of OpenFlow, but almost a third are still on the fence. (See related article.)
And finally, we asked about the relationship between SDN and Network Function Virtualization. Until recently, the conventional wisdom was SDN and NFV were separate topics and didn’t need to be formally coordinated. That changed in March 2014 when the Open Networking Foundation (ONF) and the European Telecommunications Standards Institute Industry Specification Group for NFV (ETSI NFV ISG) announced the signing of a Memorandum of Understanding (MOU). As part of that announcement, the ONF and ETSI stated that “Together the organizations will explore the application of SDN configuration and control protocols as the base for the network infrastructure supporting NFV, and conversely the possibilities that NFV opens for virtualizing the forwarding plane functions.”
The survey respondents were asked to indicate the relationship that their company sees between SDN and NFV and they were allowed to check all that applied. Their answers (Table 8) indicate the vast majority of IT organizations believe SDN and NFV are complimentary activities and a significant percentage believe that in at least some instances NFV requires SDN. This implies that a relatively large percentage of IT organizations will tie, at least loosely, their SDN implementation with their NFV implementation.
Early in the discussion about emerging technologies IT professionals tend to be optimistic about when the new tech may be widely adopted, which happened with SDN. However, the further into the development the more realistic adoption estimates become. Our research indicates SDN won’t be broadly adopted until there are mature solutions, well understood and accepted business cases and a set of best practices relative to how to integrate SDN into the rest of the infrastructure. That won’t happen in 2015 and may not happen in 2016.
Metzler has done everything from running enterprise networks to designing networks, developing software and running a consulting organization. His current interests include Application Delivery, Software Defined Networking (SDN) and Network Functions Virtualization (NFV). firstname.lastname@example.org