Software defined networks: The future of infrastructure
Few initiatives epitomise the drive to make networks more intelligent than software defined networking. And the cutting edge work, as Richard Irving discovers, is not on the ground but in the cloud.
The fight back has begun. For much of 2012, leading equipment makers like Cisco and Alcatel-Lucent struggled to dispel mounting fears that they would be driven out of business by the march towards software defined networking (SDN) – a game-changing technology that strips away the need for expensive switches and routers by taking all the intelligence out of the network and handing it to a central computer that can be programmed in an almost limitless way.
Now they are looking to reassert their dominance in a market that will be worth more than $300 million this year and possibly billions by the end of 2016. And the battlefield on which they are mustering their forces is not, as most analysts anticipated, the wide area network, but in the cloud.
In recent months, Alcatel, Cisco and Juniper have all announced acquisitions or initiatives drawing together fledgling SDN projects and cloud-based services. This should come as no surprise. Most wholesale carriers readily admit that they view the cloud as a top priority over the coming year, so it follows that equipment vendors should be exploring ways to virtualise SDN architectures.
Cynics might counter that vendors have a vested interest in preserving sales of proprietary routers and switches to network operators for as long as possible and that they therefore have much to gain in shifting the focus to cloud-based services, where they have no legacy revenues to lose. But the reality is that telecoms companies themselves see more opportunity in the cloud than elsewhere and are clamoring for the technologies to support their own efforts.
This is not in any way to belittle the huge strides that SDN design has so far made on the ground. Last year, for example, Google announced that it had upgraded its entire data centre backbone, one of the largest stand-alone networks in the world, to an SDN architecture.
But as Mark Lewis, vice president of architecture and development at Interoute, operator of Europe’s largest cloud services platform explains, the search engine’s requirements were unique and in many ways less onerous than those facing most network operators. "Google had a very specific objective and a very specific use case. However, most wholesale carriers have a huge variety of customers who want very different things from a network. We look at SDN very much in terms of how it can help facilitate cloud services."
Valuations in the clouds
Perhaps the defining moment for the SDN movement came last summer, when VMware, the Palo Alto-based software maker, stumped up a whopping $1.26 billion for Nicira. According to one estimate, the deal valued Nicira at around 100 times historic earnings – arguably the highest price that a technology start-up has fetched since the very top of the dotcom boom. But if the price was high, the expectations were – and remain – immense.
VMware’s cutting edge software helped to create the backbone of cloud computing as we know it today, optimising computer and data storage through a suite of state-of-the-art software breakthroughs. It was perhaps inevitable that having featured so prominently in the move to virtualise computing and storage, VMware would want to complete the circle with a technology that could do the same for data transport and that it would be prepared to do so at pretty much any price.
In doing so, it puts VMware's majority owner, EMC Corp, on a collision course with Cisco, the world’s biggest maker of networking equipment and a long-standing partner. Moreover, this time around, VMware will find the market a little more crowded.
Ironically, a swathe of rival start-ups that began 2012 in stealth mode are breaking cover early, spurred by Nicira’s spectacular valuation. And as they emerge, they are being snapped up by hardware vendors keen to defend a market that is conservatively estimated to be worth around $40 billion a year.
In December, Cisco itself led the way, stumping up $141 million in cash to buy Cariden, a network mapping and cloud optimisation company that is moving aggressively into the SDN arena. Juniper followed suit a fortnight later, agreeing to pay $176 million for Contrail – a company that has been going for less than twelve months and is not due to launch its first SDN product until later this year. Then Alcatel-Lucent went public, confirming that a specialist unit within the business, known as Nuage Networks and run by Sunil Khandekar, has been working on an SDN solution for the best part of a year.
Like his peers, Khandekar sees the biggest opportunities in virtualised infrastructures and sees this manifesting itself around data centres: "Data centre networks are very static, very rigid and as a result they have been slow to respond to the push towards cloud services," he explains.
"The concept behind SDN is all about bridging the gap between the network and the applications that sit above it." In the past, these have evolved separately like ships passing in the night: network operators and applications providers have both assumed that either side will configure their products to ensure compatibility. But the market is moving swiftly towards a virtualised cloud-based infrastructure and that changes everything, Khandekar adds.
"Large telcos see tangible new revenue opportunities in the cloud services market. They have considerable network and data centre assets at their disposal. The key is to provide visibility and control to the applications that run over them and thereby encourage them to consume more and more network services."
A new type of spin
Nuage Networks takes the form of a controversial new investment model known as a "spin-in", where a technology company seeds a new start-up, quite often the brainchild of existing employees, and then buys the company outright if the technology succeeds. Khandekar says that Nuage, which was created over a year ago, decided to go public after extensive discussions with clients who are very supportive of the company’s approach to SDN. But the timing of the announcement comes as competition in the SDN market hots up.
In particular, fevered speculation surrounds another spin-in, this time at Cisco, where three of the most successful networking engineers of the last two decades are busy plugging some of the gaps in the group’s SDN strategy. The Silicon Valley giant, which arguably has more to lose from the rise of SDN than anyone else, has put $100 million into a new venture called Insieme, the brainchild of Mario Mazzola, and his long-time lieutenants, Prem Jain and Luca Cafiero.
The trio, lest it be forgotten, have struck the jackpot for Cisco not once, not twice, but thrice before: In the early 1990s, the team invented a corporate networking switch while at Crescendo Communications which Cisco bought for $94 million and turned into a $13 billion business; in 2004, they helped elbow Cisco into the storage-related switch market when they sold their second venture, Andiamo Systems to the group for $750 million; and in 2009, they helped Cisco flesh out its attack on the data centre hardware and services market when they sold Nuova Systems to the giant for $678 million.
Cisco has an option to buy the latest venture for $750 million if all goes well. For now, the equipment maker is saying little about Insieme, other than to stress that Mazzola’s work will compliment a wider effort, known as the Open Network Environment, or Cisco ONE.
Cisco’s ultimate goal is to defend its lucrative hardware business from a basic principle of SDN, namely that the data plane, which is responsible for shunting packets of information on their way along the network, should be separated from the control plane, which decides when and where that data goes next.
In so doing, it is possible to do away with Cisco’s intelligent routing and switching equipment and transfer all that power to a programmable computer in the control plane that would then talk to all the various components on the network through a protocol called OpenFlow. The resulting network is infinitely customisable, flexible and scalable – and can easily be optimised making it cheap to operate.
Few would argue with Dan Pitt, executive director of the Open Networking Foundation, when he evangalises about SDN as a "revolutionary approach" to design that will change the way every organisation with a network operates. Just a few lines of ordinary computer code are enough to change the way companies do business for good, he argues.
Opportunities and threats
But where he sees opportunity, others see a potential threat. A simple operating protocol such as OpenFlow, for example, might not only disenfranchise high-end equipment manufacturers, but also undermine the foundations on which long-term relationships with clients are built by dissolving the sticky relationship which links buyers of complex and proprietary network equipment to the vendors through upgrade after upgrade.
Cisco claims that its new initiative embraces the guiding principles of SDN, but goes further by including overlay technologies allowing each layer of the network, from the transport layer through to orchestration and management layers, to be fully programmable. Crucially, says Cisco, this means that network providers can design their own protocols to further customise their own systems.
Some supporters of the OpenFlow protocol have suggested that any attempt to undermine the ubiquity of what they hope will become a de facto industry standard for SDN, will set back network evolution by years. However, many network operators recognise that the future of SDN does not hang on the merits of OpenFlow.
For his part, Alcatel’s Khandekar says that Nuage will embrace OpenFlow, but warns that the debate around the protocol is turning into an unnecessary distraction. "When you start talking about the huge potential of SDN with customers, the discussion moves on very quickly from which protocol you want to use to talk to the network elements below. The whole promise of SDN – the real nugget here – is how you harness applications to allow you to quickly consume network resources. In other words, how you hand control to applications that can determine network behavior without having to drill down and re-tool the network itself."
Given that his company is named after the French word for cloud, it is no surprise that Khandekar is steering its effort towards virtualisation-based technologies. But he is not alone.
Interoute is poised to launch a new virtual router to work alongside its award-winning virtual data centre architecture,
a private and public cloud computing platform built into the company’s pan-European multiprotocol label switching/IP network.
The router, which is currently being beta tested with customers, is part of a big initiative at Interoute aimed at helping network operators exploit the ongoing integration between the business of connectivity and IT.
"The cloud on its own is a worthless commodity unless you can actually provide transport to and from it," says Lewis. "When you combine the cloud with connectivity, you have the makings of a very powerful product. And if you can do that at scale, you should be able to lower the cost, which then becomes the point at which enterprises become very interested," he explains. The virtual router is a natural progression in this regard: "Essentially, we can point a customer to a web page where they can download a router that automatically connects to our cloud," he says.
The revolution gathers pace
There can be little doubt that the telecoms sector is moving from a silo-based ecosystem populated by a plethora of disparate platforms and product lines, to a point where technologies are integrating very quickly.
And as Nicolas Fischbach, director of network and IT platform strategy and architecture at Colt, explains, the promise of SDN is that it will help smooth that transition: "What makes SDN so revolutionary is that it has caused people to think differently about how we build networks, bind them with applications and manage them in-life."
In other words, he says, the attractions of SDN lie not in a single protocol like OpenFlow, or in the various applications that it will spawn, but in the fact that it will allow operators to develop new end-to-end integrated architecture to deliver, operate and consume network capacity and managed IT services.
"At the moment, we have different network topologies to serve different customer needs – from Carrier Ethernet to IP and IP VPN", he explains. "Going forward, we will have a single hybrid network that will be capable of delivering all three of these solutions and that a customer will be able to change between during the life of its service."
By definition, of course, an SDN solution will only ever be as good as the apps that sit over the network, and some of the most exciting development work currently being undertaken by developers relates to the way various apps and the network itself will communicate with one another.
"This is very new," says Fischbach. "The whole concept of having the network be able to feed back to the app what it can and can’t do – and the app being able to process that and understand the network’s capability – is groundbreaking. It’s never been done before."
The challenge, adds the Colt network chief, is to ensure that the apps are up to the task as you build out the network: "SDN is probably 20% about the network and 80% about the apps and systems." The concept started as an engineer’s dream but the more you work on it, the more you realise that a big proportion of the whole solution is all about systems – details such as orchestration, automation, billing – how you push the SDN concept out from the OSS to the BSS layer, he explains. "We realised very early on in our own development works that we only had engineers in the room when what we needed was architects as well."
Fischbach says Colt is currently in the process of selecting the right "fabric" to support the OpenFlow protocol as well as continuing work on the apps that will speed up network automation: "We have a very long history of driving innovation in a number of areas such as Carrier Ethernet and managed services and we plan to push network programmability with the same energy," he says.
Spotlight on SDN
How big is the potential market?
According to the information provider IDC, sales of SDN gear as well as software and services will surge from $200 million in 2013 to $2 billion in 2016. Some analysts put the number closer to $10 billion over a similar time frame.
Do carriers really have that much budget to throw at SDN?
No one is building greenfield networks anymore so any move to SDN will necessarily have to be an evolutionary process. But everyone is looking to move as quickly as possible to a lower cost structure: the quicker SDN architectures help better utilise the network, the faster carriers can grow. That’s a compelling argument for most finance directors.
How long will it take before SDN comes of age?
Don’t expect to see carriers try any meaningful deployment in their wide area networks much before 2015 – SDN is currently undergoing rigorous testing among several prominent carriers including NTT and Verizon. It’ll take about 12-18 months to iron out any glitches and at least a further 12-18 months on top of that to scale up for a launch.
Who are the companies to watch?
The SDN arena is a hotbed of M&A activity at the moment. Nicira, whose chief technology officer helped devise OpenFlow, recently fell to VMWare for $1.2 billion while Oracle snapped up San Jose-based Xsigo Systems for an undisclosed sum. Other emerging start-ups include Big Switch and Arista Networks. Don’t be surprised to see IBM, Cisco or even Juniper step out along the acquisition trail.
Can SDN live up to all the hype?
You bet. Google has already migrated one of the largest networks in the world over to an SDN format and expects to see cost savings as a result. If that’s not enough to tempt carriers, the prospect of developing new services that can attract big premiums surely will.