Your browser is not supported. Please update it.

11 June 2015

AirFrame completes Nokia’s transformation into a cloud vendor

Spending on LTE infrastructure will peak this year, and already vendors like Ericsson and Alcatel-Lucent are feeling the pinch as major customers in north America, Japan and elsewhere complete the bulk of their 4G roll-outs. Of course there will be new contracts to win, including continuing spend in China and the prospect of large Indian deployments, as well as steady expansion in Europe, including Russia, and Latin America. However, many carriers in those regions are focusing on modernization or step-by-step LTE migrations, rather than big-bang, big-buck projects like those in north America. This means there will be pressure on the vendors to seek revenue growth elsewhere.

Some of this effort will be focused on non-infrastructure businesses such as managed services. But while some of the big OEMs may, over time, transform completely into software and services providers, IBM-style, for now their business models require that they have significant revenues from physical hardware/software platforms. A slowdown in LTE spending will affect not just the access network but the amount of new backhaul and routing capacity being installed.

According to a new report by research firm IHS, LTE infrastructure spending will reach $23.2bn in 2015, but will decline after that. Stéphane Téral, head of IHS’s mobile infrastructure and carrier economics practice, commented: “We’re reaching the peak of LTE roll-outs.” Of course, that slowdown will be compounded by far more rapid decline in spending on 2G and 3G, and overall, mobile network infrastructure investment fell by 8% sequentially in the first quarter of 2015, to $11bn. LTE was up by 4% sequentially, mainly thanks to Chinese TD-LTE deployments.

But with Huawei and ZTE taking the largest share of those Chinese contracts, Ericsson and Alcatel-Lucent are both suffering from the anticlimax after delivering major LTE projects in north America and Japan.

So vendors will be looking eagerly for the next big architecture shift. Increasingly, that looks like being virtualization and cloud-based networks, and Nokia, in particular, is embracing the trend, as seen most recently in its announcement of its AirFrame server platform.

However, virtualization is a two-edged sword for traditional hardware vendors, threatening to replace their expensive proprietary appliances, routers and even base stations with commodity servers feeding low cost, stripped-down cell site equipment. Already, major SDN (software defined network) projects, notably at AT&T, are raising fears that Cisco, ALU and other major router suppliers will see their revenues from these key customers decline. They are trying to offset that with increased focus on SDN controllers and other essential technologies, and on new markets such as the internet of things. But those are major shifts towards immature markets, and require broad organizational and cultural changes, not just a new product line.

Of the tier one infrastructure suppliers, Nokia has been the most ready to adapt to new trends, particularly towards intelligent mobile networks in which the software adds most of the value (even before carriers get to full SDN, which for many is a project for the next decade). The firm’s Liquid Net was an early example of a network designed so that resources and capacity could be allocated flexibly where required, from RAN to core to transport. That was followed by Liquid Apps, which deployed certain applications and content on x86 servers close to the cell site; and this year, Radio Cloud. These appeared to plot a potential path to becoming a software-only vendor, which would put Nokia in a strong position to thrive in the world of SDN and virtualization, delivering intelligence platforms for commodity servers and routers. With the Alcatel-Lucent acquisition, Nokia will also gain the French firm’s experience of China Mobile’s and Telefonica’s Cloud-RAN trials, and its Nuage and CloudBand platforms for telco cloud services.

However, while Nokia has accepted the move away from specialized network hardware more readily than most, its AirFrame announcement shows it sticking to its hardware roots too because of course, cloud-based core and access networks still require hardware, just of a different kind. The ultimate vision may be that off-the-shelf x86 (or ARM-based) servers evolve to the stringent performance levels required by some network functions, but in reality, for many years, some of those functions will require specialized accelerators. Those will provide high value opportunities for chip and network providers, but will need to be integrated tightly with the servers themselves and with the IT functions that surround them.

Intel has been addressing that telecoms/IT convergence with a series of coprocessors, targeted at specific network activities including baseband processing, to act as companions for its Xeon server chips. Of course, making x86 servers the basis of the new telecoms infrastructure is the key goal, but Intel recognizes the need for platforms which are fully optimized for the demands of carrier networks. Now it has intensified its partnership with Nokia to create a telecoms/IT architecture which aims to offer the best of both worlds, in a highly integrated way – and to try to keep ARM out of the equation too.

Sandra Rivera, general manager of Intel’s Network Platforms Group, said in a statement: “We believe that combining Intel’s leading processors and virtualization technologies with Nokia Networks’ deep knowledge of the telecommunications market will accelerate adoption for cloud technologies deployed by leading operators worldwide.”

AirFrame is the result of that Intel/Nokia alliance, which started in earnest when they co-developed Liquid Apps. That combined x86 and base station platforms to support content delivery and other applications where they were most needed, and so reflected many of the goals of SDN – allocating resource flexibly on-demand, for instance – without actually being based on and SDN or NFV (Network Functions Virtualization) framework. AirFrame goes further and promises a complete solution for the carrier which wants to adopt a cloud-based approach to the network, and claims to be differentiated from rival solutions in several ways – combining the economics of off-the-shelf hardware and open software with accelerators for radio activities; plus a heavy focus on ultra-low latency and on security.

Nokia, celebrating its 150th anniversary this year, said this was “the first offering that combines the benefits of cloud computing technologies with the stringent requirements of the core and radio in the telco world”. Marc Rouanne, EVP of mobile broadband, said in an interview: “From today there are no borders – telecom and IT are merging, and we are recognizing this and acting”. He said operators want to embrace cloud technology in order to be more “We know how different the telco environment is from traditional IT”, Rouanne added.

That goes to the heart of the dilemma for operators – in moving towards the promises of SDN, such as lower costs and the ability to launch new, targeted services almost instantly, will they throw the optimized performance baby out with the bath water of proprietary hardware? A carrier network’s requirements in terms of reliability, real time application support, ultra-low latency and sheer data throughput capacity are in a different league from those of enterprises, and will only get worse with the shift towards the internet of things (IoT). Major cloud providers, like Google and Facebook, have some of the same challenges, and are taking increasing interest in the mechanics of an optimized delivery network.

Hence, they should be a target for AirFrame, which combines the Intel/Nokia hardware platform with a suite of professional services for implementing and operating cloud data centers that are specifically geared to telco needs.

The actual platform consists of pre-integrated racks with ultra-dense x86 servers, high performance switches and software defined storage. The server can also be slotted into pre-existing racks as a plug-and-play replacement for standard IT kit. An open interface means AirFrame can support third party software and hardware. Nokia itself already has a relationship with Hewlett-Packard for data center servers and switches though it is not clear whether that will embrace AirFrame.

While Nokia stressed that it was using open hardware and software components wherever possible (presumably AirFrame will be integrated with its existing SDN programs), much of its differentiation from Ericsson’s or Huawei’s telco cloud solutions (Ericsson ‘s hyperscale data center, for instance) lies in its claim for proprietary acceleration capabilities for radio activities, which “will help operator customers be more efficient and differentiate themselves in the market”, as Henri Tervonen, VP of mobile broadband architecture at Nokia, said.

This combination, Nokia says, can “effectively merge the telco and IT domains in one single solution” and is capable of laying “the foundation to deliver the latency and data processing requirements of the future”, even as far as ‘5G’.

“We are bringing together a hardware architecture optimized for low latency with virtual improvements to the cloud stack and acceleration techniques to offload some workloads from the main CPUs for special needs of telco customers – all built in the cloud for centralized and distributed data center contributions,” Tervonen said on a media call.

As well as low latency and high throughput, the other non-negotiable element of the next generation network will be security. This is another differentiator for AirFrame, says Nokia, which has integrated its Security Director, developed at its dedicated network security lab in Germany, into AirFrame. “This is an orchestrator for security policies, updating the security settings and capabilities as needed: If the security was included in the IT functionality of the servers, that could be a major overhead for an operator,” Phil Twist, VP of portfolio marketing, commented.

AirFrame is designed to support IT and telecoms standards so that it can run common IT applications in parallel with its more specialized telco cloud activities. This could see an intensification of an existing trend for server-based BSS systems to be more integrated with network-based OSS, so that carriers can gain a holistic view of a customer based on their real time network usage and location (from the OSS) and their billing records, service preferences and other information from the BSS. Some operators will start to virtualize all these processes on platforms such as AirFrame.

As well as supporting these kinds of internal change, Nokia envisages AirFrame giving operators the wherewithal to deploy standards like NFV, and to move into new business models enabled by flexible capacity allocation. In particular, as AT&T has outlined in its Domain 2.0 SDN strategy, it will be easier for carriers to support large numbers of service providers from their clouds and networks, in a more flexible way than the traditional rigid MVNO deal; or to rent cloud capacity for customers’ IT applications.

AirFrame is also the culmination of a gradual move towards a software-driven platform, which began with Liquid Net. Since then, Nokia has been the most prominent of the big vendors in supporting the concept of Mobile Edge Computing (MEC)– the subject of an ETSI project and expected to be a core element of 5G. MEC pushes IT and cloud computing capabilities in the RAN as close to the user as possible, as seen in Liquid Apps, and is a key focus for AirFrame.

Nokia believes MEC will be vital to achieving a flexible, agile and programmable network which can also achieve ultra-low latency and massive throughput. Hence there is an emphasis on distributed micro-clouds in AirFrame, as Nokia takes an early stand on an issue that remains highly contentious – how far data centers should be distributed rather than centralized within the carrier network.

For some applications, such as voice and various IoT services, low latency may only be achieved by having the NFV and cloud platforms close to the edge. Roz Roseboro, an analyst with Heavy Reading, commented: “It is very likely the telcos will use their existing central office locations to support NFV, and there are hundreds or thousands of those in the US. They’ll want modular solutions that can scale cost effectively in order to support lots of smaller locations as opposed to a few huge ones.”

Twist said in an interview: “The way that networks are now being built means you have these big data centers hosting all of your core network applications and you end up with a fairly inefficient way of doing it, in which every bit of data that comes from your mobile phone has to be transported across the network to this data center and then transported back out again. Our proposition is that we’re moving towards a mixture of distributed and centralized. So some of the network functions that you might use once per call – call set-up or whatever – can be in a central data center, but all of the other time-critical stuff would be far better suited if you push it out to the edge of the network.”

AirFrame will also live alongside Nokia’s other big software-defined move of 2015, Radio Cloud, which it unveiled at Mobile World Congress, and which also pursues a flexible mixture of centralized and edge-based processing.

This saw Nokia borrowing concepts from the enterprise utility computing era to offer the RAN the ability to use processing capacity from many sources – from a nearby cell to a central server – in a dynamic way.

Radio Cloud will use open source software where possible. It is built on three interconnected layers. The first is the cell site equipment – radio and antennas. These cells are dynamically grouped by the second layer, which handles traffic scheduling and coordination from the cloud. That activity then triggers capacity allocation from the third layer, the basebands, which are connected to, or even running in, the centralized data center.

The idea is that base level capacity is enabled near the cell sites, while for peaks of usage, additional resource is allocated on-demand from the data center. Nokia estimates that only about 20% of the typical RAN’s capacity is used at any one time, while the rest is active only in times of peak demand. Its Radio Cloud is designed to separate coverage and capacity, so that a defined amount of capacity is available all the time and everywhere, with the rest flexibly deployed when needed.

This is clearly one possible route to Cloud-RAN, which tends to cover a multitude of architectures from familiar base station hotels to full virtualization of baseband functions as applications on a cloud server. Full Cloud-RAN remains a medium term idea, at best, for the majority of MNOs, but the vendors are already preparing their platforms, no doubt hoping to tempt the early movers in Asia – where Nokia has gained significant experience through its close work with one of the pioneers, SK Telecom.  A combination of Radio Cloud and AirFrame, plus elements such as Nuage, which will come with ALU’s dowry, should put Nokia in a powerful position to provide a workable platform for operators to move towards Cloud-RAN, however they choose to define that.

Alcatel-Lucent announces Q1 results:

The highlights of ALU’s results for its fiscal Q115 – the last one unaffected by news of the Nokia deal, as Combes put it – included continuing improvements in two key metrics, margins and free cashflow. There was also the sharp uptick in operating income, which more than doubled year-on-year to €82m. Operating margin was 2.5%, while gross margin was 34.6%, and free cashflow was €332m, up from €66m a year earlier.

Group revenue, excluding managed services, were up 12%, with ‘next generation products’ delivering 25% growth. However, at constant exchange rates, the respective increases would have been less impressive, at 2% and 9% respectively.

Like Ericsson, however, ALU suffered from slower carrier spending in north America, where the major carriers are coming to the end of their first wave of LTE roll-out.

Combes said in his statement: “Our first quarter 2015 results reflect not only strong growth in our next generation products, but also the diligent efforts we have made to turn Alcatel-Lucent around and build a more resilient organization … Execution of the Shift Plan is and will continue to be our top priority, underlined by our commitment to reach our positive free cashflow target in 2015. These results help pave the way for successful execution of our goals as we prepare for our combination with Nokia.”

Adjusted operating income totaled €82m, up from €33m in Q114, though operating income from the vital core networking unit was down from €96m last year, to €41m – the blame lying with the softer north American and Japanese markets, plus “variations in product mix”. By contrast, the access unit generated operating income of €67m, reversing a loss last year, with fixed access and managed services being particularly strong.

Despite the progress in key areas, ALU still reported a net loss, of €72m, similar to last year’s €73m, largely because of a higher tax bill resulting from activation of deferred tax assets in the US in the year-ago quarter.