SDN SPECIAL: The $35 billion SDN question
03 September 2013 | Richard Irving
The telecoms industry agrees that Software Defined Networking (SDN) is a good thing. But network operators will need convincing before they divert capex budgets to yet another untried "bleeding edge" technology. So how can the wholesale market make money out of SDN?
John Chambers, the redoubtable helicopter-loving head of Cisco, is about to come down to earth with a bit of a bump.
Just three short months ago, Chambers rounded on the burgeoning community of engineers looking to use cutting-edge technology to reinvent the way that networks run.
In his sights were a raft of companies that want to strip all the intelligence from Cisco’s highly engineered switches and routers, handing control to powerful computers that will sit high above the network and oversee things virtually, by dint of very clever software, not very expensive hardware. Such is the potential of these computers that, ultimately, they will be limited in scope only by the imaginations of the developers that program them.
That process, in a nutshell, articulates the drive to a once-in-a-generation shift in network design that is currently under way among proponents of so-called software defined networking (SDN). So concerned is Wall Street that the movement could eventually put Cisco out of business that Chambers has been forced to go on the offensive. On a conference call to analysts in May, Chambers was scathing.
"History is littered with startups that have bet against Cisco and failed," he warned.
The barbs coincided with Cisco's annual global IT survey, published in the same month. The report concluded that one in three network engineers were as likely to see Bigfoot or Elvis as a fully-fledged SDN deployment.
Rob Lloyd – Cisco's president of sales and development and a loyal Chambers lieutenant – twisted the knife another full turn when he told more than 2,000 senior executives attending Cisco's annual Partnership Summit in June that "if you hear that SDN will kneecap Cisco and everything we stand for... it's absolutely a load of... fibs. I want you to remember that."
In truth, the Cisco chief's comments were probably always going to come back to haunt him. Cisco itself admitted as much in a throwaway line in its own survey, conceding that 71% of customers were expecting to implement some form of SDN technology in the next twelve months alone.
But even Chambers will be surprised at the speed with which Nuage Networks, a Silicon Valley-based startup that was spun out of Alcatel-Lucent earlier this year, has managed to prove him wrong.
Nuage, which takes its name from the French word for "cloud", set out to build a fully virtualised network, stretching from the enterprise customer right through to the cloud. The brains behind Nuage's groundbreaking design is a powerful computer – or "controller" in SDN-speak – that oversees everything from the way traffic is pushed through the network to a whole range of bespoke services and applications that can sit above it in the cloud.
Speaking shortly after Nuage emerged from stealth mode, CEO Sunil Khandekar said that the controller's job is to bridge the yawning divide between the physical network and the cloud, so that virtualised applications can be designed to consume the least possible amount of network for the shortest time and at the lowest cost.
"Essentially, we are working to ensure that the network doesn't get in the way of all the innovation that is happening around it," he said.
SDN sets sail
Nuage is riding a swelling wave of interest in all things SDN. According to a recent survey by Plexxi, a smaller SDN equipment provider, and Lightspeed Venture Partners, a private equity firm with a plethora of very successful investments in the sector, annual spend on SDN equipment will exceed $25 billion and could hit $35 billion by 2018.
The survey concluded that the number of companies offering SDN products has grown from nothing to more than 220 in a little over three years, while the amount of venture capital flooding into the sector has grown fifty-fold over the same period.
The stakes are potentially very big. Even as Cisco's Chambers was attempting to disabuse the growing weight of expectation massing behind SDN, a handful of carriers and service providers – including Telus, SFR and Exponential-e – were already trialling a second-generation version of Nuage's system out in the field.
Now Telus is ready to begin implementing Nuage's technology in earnest. Speaking exclusively to Capacity, Ibrahim Gedeon, CTO, takes up the story.
"We've been living with this problem since 2003, when we became one of the first carriers in the world to move both our wireline and wireless business over to a standards-based MPLS network. Within about two years, I would say, we realised that having a common language for all our transport was good, but not good enough. What we really needed was a common language at the control level. SDN hits at the very heart of that," he says.
Few question the validity of SDN as a concept. For one thing, it serves to show only too well the extent to which all that work going into making switches and routers highly powerful, autonomous points of intelligence on the network was all rather a waste of time and money. The only players that have really benefitted from all that effort over the last twenty years or so are the equipment sellers â€“ the stock market value of Cisco, for example, has surged from $1 billion in August 1991 to $136 billion today.
If not completely redundant, much of that intelligence serves only to hinder new innovation, because every time engineers try to tweak the system, they have to reconfigure the whole network – a process that can take weeks, if not months.
Nuage's Khandekar has a slightly different perspective. Significant development work has gone into virtualising the processing power of computers and servers. The missing piece in the jigsaw is the transport link between the two, he says.
"If you have a proper understanding of your computing power, your storage capacity and your network capabilities, then you are in a position to leverage those dynamics in a very interesting way," he explains.
Simon Pamplin, Director, Systems Engineering, WEST EMEA at Brocade, agrees.
"People are much more comfortable with the concept of virtualisation than they were," he says.
Given how the market has evolved in the server and storage arenas, he adds, it was only a matter of time before designers would turn their attentions to the physical network itself.
"The network is a big drag on innovation. It takes a lot of very clever people to reconfigure a network and consider the impact of various changes. If we could do all that with software, then we would make the business as a whole much more agile and therefore competitive."
And thus the pressure to change the status quo intensifies.
"Wholesale is a very critical part of our business, and much of the work that we want to do here requires us to be able to establish new connections on the network quickly and cheaply," echoes Gedeon.
"If we don't go down the SDN route, it is likely that we will struggle to be competitive either in terms of price, or more importantly, from a customer experience point of view."
One question troubling many wholesale operators, as they start to weigh up the pros and cons of going for "first-mover" advantage, is whether the flexibility and savings that SDN designs will bring to next-generation architectures will be enough to justify the investment on their own, or whether wholesalers will still have to come up with a slew of "killer apps" to seal the deal.
Brocade's Pamplin believes that most operators will probably look to tick both boxes.
"There's a very good reason why networks have evolved in the way they have – they do a very good job and we shouldn't forget that. But the reality is that the way some applications now want to talk to one another means that data flows through the network differently to how it used to," he suggests.
The ethos behind SDN, he says, is that the network should be flexible enough to enable new business, not inhibit it. Nowhere is this more critical than in the wholesale market, where margins are increasingly under pressure.
"If you think about the business model underpinning wholesale, it's essentially about selling bandwidth and minutes – and both of those are getting squeezed," Pamplin says.
"If you are a carrier and you have at your fingertips a controller that can let you reuse parts of your network several times over so that you can get more customers onto the same infrastructure, or you can put new applications onto that infrastructure without having to rip out and replace the old technologies, then you are quickly going to get to the point where you have a whole new range of services and business solutions that you can offer your customers."
The big question, of course, is what type of services and applications are going to offer the kind of returns that make the changeover worthwhile. And here, the SDN evangelists fall somewhat quiet.
It brings to mind the mood surrounding the launch of the iPad: pretty much everyone agreed that it was a good thing, it was just that no one quite knew what to do with it.
And like the iPad, it may be that SDN does not spawn a killer app at all, but rather a host of apps that customers never knew they needed until they were faced with the prospect of having to do without them.
Relieved of such pressures, software developers might tap some interesting niches. Among mobile operators ramping up their deployment of 4G, for example, there are growing calls for a suite of analytical tools that can monitor traffic in real-time, either for security, revenue or marketing purposes.
"The capability to manipulate traffic via software – in real time so that you could peel parts of it off to an analytics engine without interrupting the network itself – would be a very powerful tool. I could see a number of use cases coming out of that sort of functionality," says Pamplin.
Controlling future networks
Another potential area for wholesale providers might be network outsourcing. Carriers typically outsource network operations in regions where they simply do not want to run their own systems anymore, either for tax or regulatory reasons.
While the easiest solution is to buy capacity from another wholesale player, that typically means ceding complete control over the management of the traffic and that goes against the grain for many operators. Speaking on the condition of anonymity, a senior executive at one of Europe's largest wholesale operators explained.
"If you are buying a serious amount of capacity, you need to have some control over the traffic to make the numbers stand up. If you could separate out the management of the network from ownership of the assets out in the field through some sort of SDN configuration, then I think you would garner considerable interest – especially in Europe, where you see a lot of pass-through traffic spanning several jurisdictions."
Other potential user cases include apps that will allow wholesale providers to offload unpredictable traffic spikes on the network and the ability to set up and manage virtual networks for different types of traffic such as IPTV, VoIP and mobile.
Of course, the problem with all this talk of innovation at the application level is that a lot of it is, for now at least, just talk. Even some of the low-hanging fruit that SDN providers talk breathlessly of virtualising come with considerable challenges. As the European wholesale executive rightly points out, clever software will not solve everything in the brave new world.
"When it comes to something as conceptually simple as auto provisioning, SDN designers forget the very basic fact that you can't burst traffic on a network unless you have the hardware out there to carry it in the first place," he suggests.
Careful with that controller...
Perhaps more pertinently, the executive questions the extent to which customers – especially enterprise customers – really want all this hands-on control anyway.
"The reality is that if you are moving traffic around the network, you can get it wrong – and believe me, when you get it wrong it is very unpleasant indeed. If you're going to push that level of control out to the customer, you're going to need some very powerful software to make sure they don't mess up."
The executive saves his most damning thoughts on SDN to last.
"We are tracking developments – and we will take our lead from the vendors – but there is no head of steam building up behind SDN at the moment. As a wholesale provider, I am still searching for decent use cases."
But not everyone agrees. Telus's Gedeon concludes thus.
"You have to ask yourself why we are all in the wholesale market – after all, we are not all Akamais distributing content around the ether. In essence, we have enterprise customers who need massive wholesale links. And if you put yourself in the shoes of such an enterprise customer, wouldn't you want your carrier to be able to offer the kind of transparency that you can expect from an SDN architecture? And wouldn't you want your carrier's wholesale partners to be on the same page, so that you can benefit from that transparency end-to-end? I know I would."