In our conversations with network operators, networking software and hardware vendors, and investors over the last two months, we have routinely been asked: What’s in store for SDN – and, more broadly, networking – in 2015? Unlike the previous two years, when the questions were centered on architecture and technology, this year the questions are focused on use cases and viability. This shouldn’t be a surprise given the shifting use cases, a changing of the guard in terms of economic and technical buyers, limited new product differentiation, and mixed financial performance of most public and private networking companies.
When we speak with network operators, we hear angst. This angst can be broken down into three buckets: a) lack of confidence around the business value for many SDN cases; b) the long-term viability of the vendors and open-source projects; and c) the realization that SDN is about much more than networking – and that it’s really SDx or software-defined “everything” and SDN decisions that impact compute, storage, and security. In short, these network operators are concerned about making (and living with) long-term decisions in a world of increasing flux.
This angst is translating into longer-than-expected sales cycles and slower product adoption of SDN technologies, at the same time capping spend on traditional networking technologies. What does this mean for 2015? Here are our SDN predictions for 2015:
1. Repeatable SDN Use Cases
In 2014, we started to see a handful of use cases being deployed at a handful of network operators. Based on our network operator discussions, we expect three of these SDN use cases to be more broadly adopted by network operators in 2015:
- Network virtualization for cloud service providers: This form of network virtualization allows network operators to create virtual data center networks for multiple customers for the purpose of selling “cloud” services (like compute, storage, etc.). The driving force for this use case is new revenue generation for the cloud service provider
- Network virtualization for enterprise data centers: This form of network virtualization allows enterprises to create virtual security domains within their enterprise data center networks. The primary driving force for this use case is to reduce the cost to support and maintain legacy software applications.
- Network monitoring: This is virtualization of networking monitoring to a) expand the portion of the network that is monitored; or b) increase the utilization of current monitoring tools. The primary drivers for this use case are a) expanded coverage of network monitoring to improve network performance or uptime; and b) reduce the cost of monitoring the network.
2. Emerging SDN Use Cases
In the second half of 2014, we started seeing network operator interest in two emerging SDN use cases, for which we expect to see a number of proofs of concept and field trials in 2015.
- Big data networks: These are networks designed and built for big data applications or high-transaction applications. Think of networks designed around specific applications like Hadoop clusters, search engines, high-frequency trading, or other high-performance compute clusters or applications. We increasingly see these use cases shifting from custom-hardware-focused solutions to server– and software-driven solutions and know of a handful of private, non-published deployments.
- Virtual CPE (vCPE) and virtual PE (vPE): This SDN use case is most often associated with NFV in the press and vendor literature, though the trend we see with network operators is that this use case is equally dependent on the SDN to orchestrate the various network functions required for each vCPE or vPE. What we hear from network operators about this use case is that a) the current approach of a stand-alone SDN controller (even if clustered) is too complicated for large-scale vCPE and vPE deployments; and b) the current approach of enabling an end-customer Web portal to personalize vCPE configuration is insufficient. Network operators we talk to expect to see a model that combines the Cisco/Meraki approach to cloud-based vCPE management with SDN programmability for large service provider manageability.
3. SDN controllers redefined by SDN applications
While we hear from network operators that they are relatively pleased with the SDN controller capabilities in network virtualization applications, we have yet to hear of an operator satisfied with a stand-alone SDN controller. There’s a slew of factors here, though the biggest is the fact that these controllers are not focused on solving a specific problem for network operators. As a result, we expect to see network operators take a new, more application-based approach in which specific controller functions are directly built into a specific application designed for a specific offering – say, a virtual enterprise CPE. We see smart networking providers viewing creating apps, instead of controllers, as their opportunity to create sustainable differentiation and expect at least a couple of large vendors to use this approach to build software-driven businesses in 2015.
4. Open networking redefined by programmability
Contrary to many of the marketing messages from networking vendors, we don’t hear network operators complaining about vendor lock-in, proprietary code, or closed protocols – and nor do we hear about a desire to “throw out the incumbents” that have dominated the open-networking discussions. Instead, we’ve heard from operators the need for network programmability. In 2015, we expect this discussion around programmability to be limited to the largest network operators – whose business and revenue streams are dependent on the network — and who happen to be the largest spenders on networking technologies.
What makes networks programmable? First, they need to embody a number of open attributes (which we detail at length in part one, part two, and part three, as well as describing what that means to the network ecosystem in part four and part five in our open-networking series). Second, it’s about the use cases and applications – and identifying the attributes that need programmability for automation and/or personalization (more on that in a later post).
5. Emerging white-box business model is a big deal
The network operators we speak with talk about white box as a business model, not a technology change. From large, mainline network operators (think anyone smaller than AWS, Google, Azure, and Facebook), we are hearing the desire to purchase their bulk network hardware (think L2 switches) like they purchase their X86 hardware. However, we also hear that they desire to maintain the same network operating systems and support as they do on their networks today. Based on this feedback, in 2015 we expect to find an increasing number of network operators a) purchasing white box hardware from ODMs and server vendors, and b) purchasing their network operating systems and software support and maintenance from their traditional vendors (i.e. Juniper, Cisco, Brocade, Arista, etc.). This allows them to optimize their current networks at low risk since they do not need to modify their network operating procedures and processes. As greenfield deployments and emerging use cases are more adopted, there is opportunity for new network operating system providers like Cumulus, Big Switch, and Pluribus.
6. SDN becomes part of SDx
As network operators realize the impact of SDN on other parts of their infrastructures, we hear more conversation about SDx, where the discussion is shifting from SDN to software-defined compute, storage and security. For network operators, this means they should think more broadly than just networking when making SDN decisions. For vendors, this means they need to learn to explain how your SDN solution fits in a broader SDx context.
It’s always fun to predict the future – we’ll see next January just how right or wrong we are!