When we talk about disruption, it’s hard to point to any specific moment and be able to say it’s the one that changed things. You can point to the center of the ripples in the pond, and you’d be right to say those ripples were caused by the stone that fell in that spot, but that’s ignoring the arc of the stone, and whatever caused it to travel through the air. Learning what a stone is, then, would be the first step in understanding how it sent ripples through the pond.
The idea behind software-defined networking (SDN) is to abstract physical elements from networking hardware and control them with software. Part of this is decoupling network control from forwarding functions so you can program it directly, but the main idea is that this separation allows for a dynamic approach to networking – something that the increasing disaggregation in IT makes a necessity.
The benefits, aside from programmability and agility, include the ability to manage entire networks from a central point, which results in a drastically simplified approach to networking. SDN allows you to be responsive to the extent that it’s useful for all of today’s most exciting frontiers – IoT, cloud services, big data, and mobility.
I should probably mention that I have a horse in this race. The CloudRouter project is focused on supporting SDN infrastructure with an open source, software-based router distribution for the cloud. Our main concerns are offering something dynamic, programmable, and scalable; and the work we, and everyone else working in SDN, have done is changing the landscape of networking as we know it.
As Chloe Jian Ma, senior director of cloud marketing at Mellanox, says:
The traditional way of building networks with vertically integrated, proprietary, over-priced boxes is being disrupted, and replaced by the new and much more open way of disaggregation, abstraction, and easy composition of network resources. This transformation was driven by hyper-scale cloud and Web services providers and is being widely accepted to power Web-scale IT.
The rock is in the pond. It’s time to see how it ripples.
For about 30 years, networks have been built on proprietary hardware with dedicated software. That use to mean customers of networking hardware companies like Cisco and Linksys were limited by the features in the devices they bought. If they wanted new features, then they had to wait for the company’s proprietary hardware refresh, which could take up to three years.
This timeline caused a lot of frustration for customers that wanted more control over the flow of data through their networks, and those that wanted to iterate on the functions of their existing networks faster than the companies were doing it themselves. Since then, and especially recently, virtualization has become more prevalent, which has resulted in servers and storage being virtualized along with, now, networking.
Until recently, networking has consisted of switches and hardware patched together with cables. SDN adds a virtualization layer on top of the hardware, which allows for certain functions to be controlled with software, especially the management and flow of information to make it more efficient.
Concepts similar to this have been around for years, but what makes the idea so exciting lately is the shift to open architecture. Commercial off-the-shelf (COTS) hardware, without the constraints of proprietary hardware, allows us to manage network services without requiring that everyone in the network has the same logo on each box.
This gets at the crux of the disruption taking place – we’re taking networking control away from the major corporations that once managed it.
It’s hard to say whether SDN has disrupted IT infrastructure or SDN is a result of the disaggregation that’s been taking place. But either way, the disaggregation means network administration can be done remotely, which now includes making physical changes to the network with software.
So why is this a disruption? SDN allows for an evolution in business models. Since white boxes can be used with networking software to create dynamic, centrally managed networks, businesses are no longer required to pay the premium associated with traditional networking vendors.
There’s flexibility now – you can take COTS/white-box hardware and get software on it that lets it compete with any proprietary solution. And you’ve come away with something responsive, scalable, and dynamic. As data loads continue to increase, this flexibility is something we’ll see reflected in the goals of contemporary business models. IoT is going to change things in a big way over the next few years, and businesses will need a networking system that can handle the massive changes about to take place.
And we see this reflected in the numbers. A report by Infonetics Research in 2014 (even back in 2014!) found that 45 percent of enterprises will have used SDN in their data centers in 2015, and that percentage is expected to increase to 87 percent in 2016.
These numbers aren’t surprising, either. Most corporate networks can’t adjust to the requirements of advanced technologies, but SDN’s dynamism and reflexes can – virtual machines can be deployed in minutes.
I’ve touched on this a little already, but SDN is provoking a sea-change in terms of the hardware that’s used in networking. The idea behind switching to white boxes, or COTS hardware, is flexibility.
Organizations are increasingly finding that traditional networking hardware is too closed-off and too expensive for them. Traditional networking devices include a multitude of features and functions that most people don’t actually use, but they’re still paying for that functionality because it’s bundled into the device. These unused functions represent a significant technical debt as users are forced to consume updates and deal with defects from the portion of the device they never use.
By switching to SDN, users and organizations can choose hardware that works for them based on the hardware’s merits, and they aren’t locked into any dedicated operating system. SDN allows for a much more flexible implementation where the users can customize the device to suit their needs. They can iteratively update their environments instead of performing forklift upgrades.
It might sound as if SDN will be the biggest part of networking in the future, but according to an IDC report in 2014, the physical networking hardware will continue to be the biggest piece of the networking pie, even in 2018. This should help explain why the overwhelming migration to white-box hardware is so important, and might explain why even Facebook is getting in on white boxes with the Open Compute Project.
The networking hardware landscape of the next few years will be one based on the collaborative approach that’s been steadily gaining popularity. That collaboration will result in white boxes that are increasingly impressive from a performance standpoint, even as there is more standardization among them.
Financial Disruption and the Future of SDN
According to a report by the CSFB, the SDN alternative to a basic Cisco data center switch is roughly 70 percent cheaper. Companies have huge budgets for networking, and SDN has the capacity to save them significant portions of those budgets by eliminating the need for proprietary hardware and dedicated software. Those companies have a viable alternative with SDN – one that allows them to customize their networking hardware through abstraction and virtualization so it operates at peak efficiency according to their unique conditions.
Maybe that’s why SDN revenue was expected to grow 126 percent from 2014 to 2015, and is expected to keep growing to a $3.2 billion industry by 2018. It might not be easy, as this article on the commercialization of SDN and NFV discusses, but data centers are demanding SDN’s ability to handle increasing loads of information, and global cloud security services flock toward SDN for its scalability.
It’s no surprise that, according to the Infonetics report, 85 percent of midsized and large North American enterprises say they expect to be using SDN in 2016. And if none of that has you convinced – a report from Rayno Media states that the SDN market will be worth well more than $20 billion once the adoption is complete.
I started this article talking about a pond, and maybe I had the imagery wrong. Let’s make SDN the pond instead of the rock: It looks as if it’s time to jump in.