Among the glitzy stands and hyped-up smartphone launches that characterise the annual Mobile World Congress fair in Barcelona, the topic of network functions virtualisation (NFV) could be described as something of a dry topic.
However, as the mobile world moves down the path to faster 4G networks and future ubiquitous 5G networks, it will become a vital element in a communication service provider’s (CSP’s) arsenal.
NFV works by decoupling network functions from dedicated hardware devices – such as routers, firewalls and load balancers – and hosting them on a virtual machine. This means that networks can be controlled using a hypervisor on a standard x86 server.
A recent report from the OpenStack Foundation claimed that NFV was a potential game-changer for mobile operators. This is because it can help them develop and deploy services in double-quick time, while reducing their reliance on proprietary networking hardware and freeing up datacentre capacity.
According to Jonathan Bryce, CEO of the OpenStack Foundation, NFV has emerged as the fastest growing area for OpenStack. He says this was not a use case he anticipated when the organisation was first set up.
“It turns out that some of the telcos wanted to automate their resources. They came to our community because they thought OpenStack had the potential to help them do that,” says Bryce.
He cites the report as a clear indication of “how important and pervasive” cloud technology is becoming to every industry, including networking.
US telco AT&T was one of the first to use OpenStack for NFV. It now runs a number of its datacentres, and many millions of calls over it. Where a company of its size goes, others tend to follow.
AT&T’s call to arms has Deutsche Telekom, Orange and Telefónica interested as well, says Bryce. BT is also understood to be evaluating the technology, as well as China Unicom and NTT in Japan.
Perhaps predictably, SKT from South Korea is well along using it in both private and public cloud and research and development for its future 5G deployment.
So why OpenStack and its competitors? “There are a couple of things specifically,” Bryce explains. “Firstly, cloud is the right way to manage resources dynamically and they need to do that for their networks.
“In our case the existence of the community was also key. If that hadn’t been there they may have moved in a different direction. The big driver for NFV is to lower cost models. Participating in open-source communities is very disruptive to entrenched business models that telcos have followed for decades.”
Bryce says that networks have traditionally been very carefully managed because they are so critical to the enterprise. Therefore, there has been a certain amount of resistance to the idea of automation and cloud-enablement.
However, he says, the technology is now solid enough that – whether an operator is provisioning infrastructure, managing failover or offering new services – there is a solid use case.
Community members – particularly those from software-defined networking (SDN) companies such as Big Switch or Nicira in VMware – are working on Docker, which can be a challenge in a network environment and containerised networking systems.
Brocade CTO of mobile networking Kevin Shatzkamer says CSPs are demanding openness. “They want integration, open APIs [application programming interfaces] into the infrastructure for network management, and provisioning. Our perspective is built entirely around open standards – our controller is OpenDaylight, and we have a large development team working with OpenStack,” he says.
Swisscom picks HPE for virtual networks
Armin Veit, senior network architect at Swiss incumbent Swisscom, is in the process of deploying NFV to offer virtual customer premise equipment (vCPE) to his business customers, and has a number of trial installations in the field. He is using Hewlett Packard Enterprise’s (HPE’s) OpenNFV platform to do so.
“In the past we had very traditional networks. If our customers wanted to change their firewall rules, they had to call us and we had to do it for them, which adds a lot of delay,” says Veit. “
“We want to let SME [small to medium-sized enterprises] customers change anything they want by themselves. That’s the basic idea, to get to lower operational costs than we have today.”
Up to now, like most telcos Swisscom deployed network functions at customers using dedicated on-site appliances, leading to significant capital expenditure (capex) and “truck rolls” for installation and service calls.
Because it will be able to manage its customers’ networks from a central location, offering on-demand services, its own costs will come down, its deployments will be quicker and its services will be more highly available, says Veit.
“We had a big debate about carrier-grade OpenStack and we’re trying to keep as much of it open as possible to prevent future problems,” he tells Computer Weekly.
“We want to build the services we need to have feature sets that are easy to maintain, otherwise we would just end up building new silos. We want to lever what we have already built, otherwise we would have to rebuild it and reintegrate it, which costs money. That was the point at which we decided we had to rely on OpenStack.”
David Sliter, vice-president and general manager of communications and media at Hewlett Packard Enterprise (HPE), is fully supportive of this idea.
HPE recently made it a requirement that any supplier wishing to sell cloud apps or services to the European public sector or commercial organisations via its Cloud 28+ catalogue must ensure their offerings are OpenStack-compliant in a year of joining.
“It’s best of breed. NFV allows the customers to say okay, I want X from these guys, Y from Cisco and Z from HP. Some are competitors and some are our partners. The network world is not a binary thing anymore,” he says.
“The part I liked about the Swisscom approach was they adopted an agile, business-focused methodology. They defined the business outcome, which was to launch vCPE as a service. They had business-driven success factors, as opposed to saying ‘let’s go build a virtual platform just because we can’.”
Ray Watson, vice-president of technology at network services supplier Masergy, agrees that the “biggest, most exciting thing” about the open approach to NFV is that it brings independence to the customer.
“Now I can spin up a Brocade or a Cisco router on a Linux machine in 15 seconds, I am not locked into their proprietary chassis. So by using OpenDaylight or OpenStack, enterprises will benefit hugely by either joining the whitebox revolution or by leveraging the technology to get better pricing from mainstream suppliers.”
“The best way to achieve multivendor at the control and orchestration layer is to lever open source and stay open at the technology level,” adds Brocade’s Shatzkamer.
Time running out
For Shatzkamer, the time for CSPs to act on NFV is now. He argues that changes to European Union (EU) roaming regulations – along with companies such as Apple and Facebook doing more over-the-top services and Google’s entry into mobile virtual network operation – necessitates a kind of networking arms race.
On the technology side, he says, there are other challenges, such as bandwidth demand outpacing available spectrum on mobile networks, traffic variance and building networks to support massive internet of things (IoT) sensor deployments.
“The net result is that you need flexibility in your infrastructure to support these use cases. The industry is not in a position to wait three to five years, there are SDN and NFV skills and operational best practice that they need to learn now,” he says.
“CSPs need to understand not only the network, but how to tune and optimise Linux, virtualisation of both virtual machines and containers, scripting and programming APIs and have a working knowledge of computer science.
“These people exist, but they are few and far between and Facebook and Google are recruiting them all. CSPs must ramp up acquisition of that expertise for themselves.”
Beyond the telco
To date, it has been mainly telecoms operators that have gone down this road. But it will eventually hit the enterprise in a more noticeable way – particularly in those with multiple, distributed sites.
Watson at Masergy, who has been doing SDN on the quiet for more than a decade, sees open standards as key to emerging hybrid, programmable enterprise wide area networks (WANs) – something he refers to as the fifth generation of WANs or 5GWANs.
“The biggest component is programmability, the ability to make real-time changes to aspects such as bandwidth and quality of service, and doing it all in an open and secure way,” he says.
Besides real-time service provisioning and adjustments, Watson forecasts that such networks will help enterprises address security concerns, meet the demands of their millennial employees for more seamless ways of working and address constrained IT budgets. This will also mean enterprises can internationalise their business more readily, he adds.
“A big surprise for us has been that some customers’ biggest use case is avoiding customs,” he says. “If a customer didn’t need a firewall in China, but suddenly they do, we can get it there in minutes.”
Watson predicts a number of factors will contribute to a so-called tipping point for wider enterprise adoption of such networks. He says they will be heavily dependent on adoption of emerging areas of technology, such as the IoT, that they will be required to support.
Read more about preparing for NFV
- Telecom expert Tom Nolle explains how to convince CFOs that the time is right for NFV implementation.
- Orchestration challenges are slowing NFV adoption, with a number of factors hindering deployments. Expert Lee Doyle explains and offers advice for service providers.
- Cisco’s NFV software stack will eventually run on any Intel-based x86 server. Will it be a game-changer for customers?