Your browser is not supported. Please update it.

8 May 2014

Does NFV need a new IT architecture, and who will control it?

It is now clear that the hallmarks of the new telco network will be dramatically different from the current generation, and will be geared to supporting billions, not millions of devices; thousands, not tens of service providers; and hundreds, not tens of new operator applications per year. Growth for the mobile carrier will depend on extreme flexibility, in order to reduce costs but, more importantly, to launch (and kill) new services, new MVNOs and new devices at the flick of a (virtual) switch. The cornerstones of this new flexibility will be virtualization; and use of the cloud to analyze the data from the billions of ‘things’, to deliver services to consumers and other organizations; and to manage the network itself.

All kinds of players can see opportunities in the shift to running telecoms activities on IT platforms, in software. The most obvious are those with experience in the enterprise data center itself. IBM and HP are extending their telco managed services to include networks, not just back office systems; Oracle and Amdocs are buying companies which operate deep in the network to complement their IT-centric solutions for carriers; EMC is beefing up its cloud architecture to cope with the massive data mobile and embedded devices will generate.

Various questions arise though. Will specific cloud platforms be required to meet the particular needs of carriers? Already, NFV (Network Functions Virtualization) is an approach heavily geared to those needs, but Japan’s NTT is also leading a push for sector-optimized hardware. And in many years to come, when the telco’s cloud and virtualization architectures are fully established, who will be controlling those clouds ‘ traditional operators, or players coming in from the vertical segments which will create so much of the growth driven by the internet of things (IoT)?

Espousing the argument that NFV in particular, and carrier SDN (software defined networking) in general, will need a new class of IT platform, a group of Japanese players have set about defining how that would look. A consortium led by telco NTT Group is working on a distributed server system optimized for NFV. NTT and its partners, Alcatel-Lucent Japan and Fujitsu, call their project VNPaaS, adding to the endless variety of ‘aaS’ acronyms with ‘Virtual Network Platform as a Service’.

The partners argue that virtualizing large-scale telecoms systems involves so many functions that a more scalable, flexible server architecture is required, which can then manage many carriers’ networks on a hosted basis. VNPaaS is actually a use case within the ETSI specifications for NFV, and NTT’s proposal has been approved as a proof of concept by the ETSI Industry Specification Group (ISG) which leads work in this area.

The VNPaaS platform is based on Alcatel-Lucent’s CloudBand offering for virtualization and orchestration, while Fujitsu has provided the middleware to ensure reliability and scalability. Such efforts are important because nearly all players are clear that the challenge for NFV is not virtualizing the functions per se, but orchestrating and managing them once they are deployed in increasing numbers and complexity. Companies like Cisco are placing heavy emphasis on orchestration in their NFV strategies, as is the more open CloudNFV initiative.

However, while one of the attractions of virtualization is the cost efficiency achieved by using off-the-shelf server hardware, some companies ‘ and not just IT hardware vendors ‘ argue that such servers are not carrier-grade. There are more robust (and more expensive options) like NEBS (Network Equipment Building System) compliance, but many, like NTT, argue that a whole new design of server hardware and operating software is required to get the most from NFV.

‘Conventionally, reliability and scalability functions should be realized by applications. As advances are made in virtual network construction technology, the software scale of each service application will increase due to the numerous functions that the software will have in the future,’ says the statement from the NTT consortium. ‘However, the proposed platform technology supports reliability and scalability functions for each service application and enables faster development of service applications. Moreover, the platform that consists of distributed hardware resources enables faster system recovery, for instance, in the event of natural disasters in the future.’

While Alcatel-Lucent’s CloudBand has been prominent in the push for telco-optimized IT platforms, all the big network vendors have similar developments, hoping to keep their precious carrier hardware customers loyal, even if the type of hardware has changed. It is clearly in their interest to argue that specialized IT systems will be required, not just a roomful of standard boxes from Dell or HP. Most companies (even Intel) do accept that some network processes will never run effectively on standard x86 boxes, but will continue to require specialized DSPs and accelerators. But where the line is drawn can be a question of commercial and political interest rather than technology.

Nokia is drawing its own line in partnership with Juniper, one of the companies which seems set to benefit from the operators’ shift to virtualization and the cloud ‘ if only because it can exploit some of the inevitable conflicts of interest for arch-rival Cisco. Cisco has an extensive virtualization and SDN program, but clearly has some tough dilemmas to face with the erosion of its traditional network hardware base ‘ it has been accused of promoting a ‘hardware based networking’ approach which would keep its own architectures at the heart of an otherwise open SDN strategy. Juniper has less to lose, and was recently anointed by AT&T as an important part of that carrier’s closely watched SDN roadmap. Now the router maker has also extended its close alliance with Nokia (so close that it would not be too surprising if the newly cash-rich Finnish firm should make a bid).

The two firms cooperate in several areas, and have now added a telco cloud managed platform specifically for mobile broadband operators moving to NFV. The joint platform unites Nokia’s Liquid Core, which includes NFV and carrier-specific cloud management and orchestration, with Juniper’s MetaFabric data center architecture, and Contrail, its NFV and cloud network automation service. The bundled offering will support managed services based on OpenStack and will be available later this year.

The latest iteration of Liquid Core implements most of its elements as virtualized functions (for example, virtualized MME, IMS and Home Subscriber Server. It also includes the a cloud version of the NetAct network management solution, plus Cloud Application Manager, which provides orchestration for end-to-end services like VoLTE or evolved packet core in the cloud. Meanwhile, MetaFabrice and the Contrail SDN/NFV controller enable an automated and scalable mobile edge, so that new services are created as required, based on network intelligence.

‘By collaborating with Juniper, we are taking the next step together, enabling operators to take full advantage of the cloud with a clear path towards the robustness and scale of interconnected data centers,’ said Rajeev Suri, CEO of Nokia, in a statement.

Being included in AT&T’s software defined cloud initiative is particularly important for Juniper, given Cisco’s strong base in the carrier’s current networks. The smaller router vendor, and back office software provider Amdocs, have joined the first announced partners in AT&T’s User Defined Network Cloud project ‘ Ericsson, Tail-f, Affirmed Networks and Metaswitch. The line-up (with the exception of Ericsson) tells a powerful story of the opportunities for an entirely different line-up of suppliers, even including start-ups, to penetrate the big telcos via cloud and virtualization migrations.

While network and server vendors alike are focusing on creating platforms for telco NFV, the other big pressure on operators’ servers will come from the explosion of data coming from rising use of conventional mobile devices, and from IoT gadgets. This was the focus of EMC’s annual user conference this week, at which the firm optimistically proclaimed that the data from mobile devices would force carriers and enterprises to upgrade their legacy infrastructure to the cloud in order to rationalize the costs of supporting the data boom, and to harness big data services to make sense of it all.

The firm’s particular interest is in the huge rise in demand for storage ‘ its key business – which is sparked by these trends. Jeremy Burton, president of products and marketing, said: “We think these kinds of applications will redefine business in every industry. We think it’s going to have a knock-on effect on infrastructure inside IT and with storage being a part of that infrastructure that is going to get redefined too.’

Given their deep understanding of mobile consumers and devices, large cellcos believe they will be well positioned to provide mobile-focused cloud and big data services to enterprises and web giants. However, in the IoT, the companies in control of the networks and big data centers will often be a new set of rivals ‘ vertical giants like GE, which have the same access and understanding of embedded devices such as factory machines as cellcos have of their smartphones; as well as the inevitable over-the-top big data purveyors. Where cellular connections are used to connect the ‘things’ (which will certainly not be in all cases), the MNO will still have a role, but as in consumer services, it faces the risk of being the bitpipe, while others take the high value revenues in analytics and endpoint management services.

GE is a good example of how a vertical powerhouse can become a dominant player in the IoT, where many of the revenues will flow not from consumer gizmoes like smart glasses, but from specific industries with critical business needs that involve collecting data from large numbers of embedded devices. GE has created its Industrial Internet Initiative (III), which started life as a platform to manage all the connected products it builds in a common way, but which now has the bigger ambition of becoming an industry standard for interconnecting and communicating between objects.

The body set up to achieve that goal does include a carrier (AT&T) as well as GE itself, Cisco, IBM and Intel. They are working to establish standards-based interoperability for gadgets and sensors, particularly those embedded in machines, based on GE’s existing work and on specifications like the IBM-backed MQT messaging protocol.

Another important aspect of III is GE’s Predix, a platform for IoT analytics, and for creating apps for vertical market ‘things’, from aeroplane engines to smart meters, and remotely managing them. There are four elements ‘ sensors, analytics, device management and developer APIs ‘ all based on standards. This is cloud-agnostic and currently contains 24 data predictive capabilities, most of them geared to key M2M verticals like healthcare and energy. It claims to be the first data analytics platform that can manage massive scale industrial machines in the cloud, and therefore well placed to provide a framework for the whole IoT world.

Some of the partners in Predix are interesting because they show, like AT&T’s program, the ways in which suppliers from non-telecoms worlds can extend their reach once networks are run as IT systems, and once they address things as well as consumers. The main Predix supporters are Amazon Web Services on the cloud side, Accenture, and EMC’s Pivotal unit for data analytics and tools. GE is an investor in Pivotal, having plowed $105m into the EMC/VMware spin-out a year ago, and that alone should put the analytics firm into pole position for the potentially massive market for IoT big data. It will be the main partner to deliver Predix as a service to industrial customers, in tandem with III where relevant.

Industrial giants are set to be kingmakers in the IoT, as are large utilities, transport firms and other companies which manage vast numbers of things. With smart metering one of the most lucrative early applications for IoT (aka new generation M2M), it is unsurprising that close links are emerging between the telecoms and utilities worlds.

Telefonica is even reported to be planning to sell electricity to customers along with broadband services in Spain ‘ an event which would highlight two important trends, the shift of broadband and mobile networks, eventually, to be utility services; and the chance for operators (or, indeed, utilities themselves) to be the primary point of contact for a whole range of home-based services. The key will be a hub that brings together all those services ‘ broadband access, security, smart metering and so on ‘ which could be housed in the gas company’s meter, the broadband provider’s gateway or the broadcaster’s set-top box, among other items.

For vendors and managed services providers, all this cross-sector convergence should open up bridges into untapped verticals. Ericsson has been particularly active here, developing IoT managed services for carmakers (sometimes bypassing its traditional cellco customers in doing so). This week, it announced a deal to provide managed services for the internal networks of gas and electricity supplier Con Edison in New York City ‘ the Swedish firm’s first utility deal in the US. The service will handle dispatch of technicians serving 3.3m electricity and 1.1m gas customers, with Ericsson taking over Con Edison’s private network based on (ironically enough) the iDEN technology invented by Motorola. While a small project by the standards of Ericsson’s managed services (24 cell sites and a core), this is important as an indicator of how a telecoms vendor can use cloud IoT trends to expand into new segments, just as vertical players can tread on their telco turf.

VMware drives standard for virtualized networks:

Microsoft, VMware, Intel and Red Hat have jointly submitted a would-be standard to support interoperability between network applications on a virtualized network. Their base technology is the Geneve protocol, which has been submitted as an informational draft to the internet standards body, the IETF.

The technology currently bridges two prominent protocols in virtualized networks, VMware’s VXLAN (Virtual Extensible LAN) and the Microsoft-backed NVGRE (Network Virtualization using Generic Routing Encapsulation). This would ease the burden for operators and software developers, which currently have to support multiple protocols in their designs.

Microsoft played down the initiative and told LightReading: ‘The industry needs to provide customers with choice and broad interoperability across software defined networking implementations. Consequently, Microsoft is in the early stages of collaborating with others. The Geneve draft submission is simply one amongst multiple ongoing initiatives we are involved in. For another example, check out our work with the Cisco ACI OpFlex Protocol.’

VMware expects to see Geneve included in Open vSwitch and Linux platforms by the summer, but while the signs of a truce with Microsoft are hopeful, others are less willing to play nice. Cisco, Google and Huawei engineers are all said to be resisting ‘yet another protocol’ and arguing for simply extending VXLAN to become a de facto standard.