This doesn’t mean enterprises are far along the SDN path, however. Just 11 percent have it in production, according to the Network World 2018 State of the Network survey. But the tipping point is near. Another 49 percent of respondents are researching or piloting the technology.
No wonder some expect global SDN to achieve 28.66 percent CAGR through 2023, while others are even more bullish, looking at 25.4 percent CAGR by 2021. That for a market IDC says had already reached $6.6 billion last year.
Technology professionals of all stripes will need to get a handle on SDN. What are the most important things to know as we approach take off?
A decade ago, “management” was a dirty word when it came to networks, implying custom scripts to control complicated switch and router hardware. Now with SDN, network management is becoming an unavoidable reality.
The shift is arising as the control plane is separated from the data plane, enabling true programmability. We’re looking at a transition that will kill the original purpose of the Domain Name System (DNS) designed to record what workloads are running where.
With the emergence of cloud, the pending 5G roll-out, and the complexities of IoT, the hopes pinned on SDN are many:
·A centralised, network-wide view for management and provisioning
·The ability to dynamically configure, secure and optimise network resources
·Greater automation available with non-proprietary programming
·Management of physical and virtual devices using the same controller and APIs
·Native cloud-friendliness and easier-to-manage quality of service (QoS) for voice, video and multimedia
·Security features beyond traditional firewalls and filtering
·Network experimentation that doesn’t affect the production network
·The potential to reduce network staffing and rededicate resources
·Additional OPEX savings through efficiency and server utilisation gains
SDN supplies the capabilities required for continuous deployment, hybrid and multi-cloud infrastructures, and other emerging demands, but until recently, implementation difficulties have barred widespread adoption. In the coming months, various companies are introducing tools and products promising to make SDN much simpler to establish and manage, lowering the potential downsides and letting its advantages shine.
SD-WAN as a subset of SDN is rapidly gaining traction. The Network World survey indicates 63 percent of organisations have current plans for the technology, and Gartner predicts 50 percent of routers will be replaced with SD-WAN by 2020.
The reasons make a lot of sense. SD-WAN supports multiple connection types, including MPLS, fiber, cable, and LTE, and enables dynamic load-sharing across WAN connections. Some early adopters are seeing savings in excess of 50 percent. Why? In addition to delivering the many benefits of virtualisation and automation, SD-WAN can reduce organisations’ reliance on expensive, private connections, such as MPLS.
The only things holding SD-WAN back are existing investments. Long-lived ASICs devices and multi-year carrier contracts are delaying SD-WAN for some IT organisations. As hardware ages and internet contracts expire, enterprises will be able to move on to SD-WAN, so it’s best to put blueprints in place ASAP.
SDN relies on open standards, and that gets some in the tech industry excited about vendor-agnostic capabilities. Many hope SDN will increase the viability of white box networking equipment—the cheaper, unbranded hardware that could, conceptually, work just as well as OEM gear.
It would be nice to spend less on switches and routers, and as the intelligence in networking moves from the hardware to the software, some outfits will pull it off. Telcos, such as AT&T in the U.S., are looking to build wireless network infrastructure around white box equipment. Facebook and hyperscale CSPs are also driving open source.
The problem comes in for the rest of us. The fact is, most large enterprises aren’t designing and running their own networks. They’re reliant on professional services from the big names. A need for Cisco engineering, however, means Cisco networking products will be sold in.
SDN has been notoriously difficult to implement, although perhaps there is hope for DIYers with the new tools coming online. Still, enterprises that live or die based on availability will likely be slow to save pennies on hardware, only to lose pounds on downtime. For smaller businesses, outsourcing the headaches with such offerings as SD-WAN-as-a-Service could provide an alternative route to the white box, but with the associated mark-up of the provider.
The long and the short of it: expertise still matters, and it will still demand a premium.
The potential for savings through white box hardware may not be as high as some predict, but hardware costs can still be reduced. SDN moves innovation out of the hardware, so performance improvements don’t require CAPEX investment in the latest networking equipment. Companies will find their existing hardware can provide value for longer timeframes.
The savings can be increased if enterprises reexamine their support contracts and move networking gear to third party maintenance. In particular, enterprises should shun the concept of “end of support life” pushed by the OEMs, and make their own lifecycle management decisions.
Microsoft is tackling SDN management, incorporating it as a key component of its software defined data centre (SDDC) toolkit and Server 2019. The industry can now use familiar management suites and follow the screen shots to get to SDN.
Similarly, VMWare has positioned itself to be a leader, aiming to take microsegmentation to a new level. As perimeter security fails in the face of BYOD, IoT, and edge computing, it must be replaced with more granular measures. Microsegmentation 2.0 promises to combine NSX SDN with virtualisation and containers to make this happen. No longer should it be quite the same guessing game to determine whether network packets are legitimate.
Other familiar names are in the SDN chase as well. Cisco is not going to roll over for VMWare; the company’s Application Centric Architecture marks its foray into SDN. Moreover, Juniper and Arista are making similar investments.
Traditional network equipment manufacturers won’t die off in the SDN era, but the technological change poses a fundamental challenge to their operations. These entrenched companies are being forced to transform from device makers into software vendors.
It’s a different business model with different solutions required for development, testing, quality assurance, security patching, and so on. Software isn’t completely alien to the networking equipment leaders, but some experts predict that the emergence of SDN will add risk of failures, especially with regards to security.
Until we see more enterprises implement SDN, armchair predictions can only go so far. It’s likely that the market will fragment for some time as various contingents go in search of their own versions of SDN utopia. But from the chaos should emerge replicable SDN best practices, and that coalescence of theory, insight, and experience probably won’t be long in the making. Bets on 2020, anyone?