Wireless Watch has covered considerable column inches in recent months demystifying how the new RAN will look as operators progress ‘ or sometimes stumble ‘ towards the true HetNet (heterogeneous network), somewhere over the horizon. But of course the new-look RAN ‘ small cells, super cells, Wi-Fi, Cloud-RAN and so on ‘ is only prevented from being a chaotic mash-up of untargeted capacity when it is combined with a new-look back end. All that fragmented capacity will only make sense ‘ not to mention profit ‘ when it is driven by software that allocates resource exactly where it is required, and in just the right amount. No underused spectrum, no over-resourcing of little visited sites, no wasted bandwidth ‘ that is the real goal of the HetNet, and while its wheels may be the new cells, its engine is SDN (software defined networking), applied intelligently to a carrier environment.
Vendors and industry bodies are rushing to influence this nascent trend and put their own stamp upon it. Players with IT origins like Intel and IBM push a pure picture in which nearly all network functions are virtualized, on routers or, better still, x86 servers. Network providers stress the specific requirements of the carrier market and claim to be best placed to create SDN platforms that are optimized for this (and which may require rather more specialized hardware and software tools along the way).
Operators are also staking their claims to be scene-setters, at least where they can afford to invest heavily in defining software-driven architectures to match their needs, and in doing so, reduce their lock-in by traditional suppliers. AT&T recently put SDN at the heart of its new procurement structure, China Mobile is the cheerleader for Cloud-RAN, Deutsche Telekom has proclaimed itself a ‘software defined operator’.
Some of the early stepping stones are being laid down, from networks which share out resources flexibly and dynamically from end user to core (NSN’s Liquid family for example), to the early stages of virtualizing base station functions in a Cloud-RAN (as at China Mobile/Intel). Intel, Cisco, Huawei and others have unveiled ambitious carrier SDN architectures and even some early products.
These are baby steps. Like any major technological change, SDN will need the confidence of the majority of operators, not just a few groundbreakers, to become the norm, and that will require the confidence that comes from standards, plentiful product competition and robust testing and interoperability. Those elements are years from being mature, but the industry recognizes the need, and at least can learn many lessons from the enterprise environment, which is a lot further down the SDN track. This month has seen some important steps towards making carrier SDN – and one of its particular manifestations, NFV (Network Functions Virtualization) ‘ part of the mainstream in mobile and converged operator planning.
The first was standards body ETSI’s publication of a set of specifications for NFV. Just 10 months after it formed the NFV industry specification group group with a selection of operators, it has released the first five specs, designed to ‘clearly identify an agreed framework and terminology for NFV, which will help the industry to channel its efforts towards fully interoperable NFV solutions’. The body’s statement continued: ‘This in turn will make it easier for network operators and NFV solutions providers to work together and will facilitate global economies of scale.’
In fact, the ETSI NFV group will not set standards itself but will make contributions to ITU-T SG13 work on future networks, so a full official platform is likely to be years away, but these documents are likely, in practice, to become de facto standards to underpin current developments.
NFV aims to make it easier and cheaper for operators to bring new services to market and so encourage innovation, by using IT virtualization techniques. These separate hardware from software functionality, and network control plane from forwarding planes, as well as enhancing the capabilities of the software layer. The result is to consolidate the functions of an array of network equipment in software on standard servers and switches, located in data centers or network nodes. This could, in time, replace a range of expensive dedicated hardware ‘ many operators, such as Orange, think some mobile core elements like policy/PCFR platforms will be first to go, although C-RAN enthusiasts are concentrating on the base station.
Four of the five new documents cover fundamental aspects of NFV deployment – use cases; requirements; architectural framework; and terminology. The fifth defines a framework for coordinating and promoting public, proof-of-concept demonstrations to showcase aspects of NFV. More specs will be released during 2014 plus further detail on some of these.
“The documents, and the PoC framework in particular, send a strong message that we want to encourage multi-party interoperability and the growth of an open ecosystem,” said Prodip Sen, chairman of the ETSI NFV industry specifications group, and an executive of Verizon. The body says some pre-standard NFV roll-outs are already underway and more than 150 operators, vendors and other ecosystem players are working on the technology.
While NFV can be implemented separately from SDN, it is effectively a carrier-specific approach to that more generalized approach, and is likely to complement other SDN frameworks, as well as contributing to the overall debate, as the telecoms sector, particularly on the mobile side, starts to overcome its insecurities about something so radically different.
The key objective for NFV and SDN supporters is to convince the industry that these technologies will make operators’ lives cheaper and easier, rather than adding a new level of complexity and disruption when many of their profit models are hanging on by a thread anyway. However, the days when some companies could play on those nerves and pretend SDN would just go away are really over. Every major supplier has to have a strategy for SDN and NFV, however reluctantly, because at least a powerful subset of carriers is determined to pursue the promised benefits of reducing cost, aiding flexibility and avoiding supplier lock-in.
And some operators are not just supporting SDN out of theoretical interest, or to pressurize their vendors ‘ the approach is starting to impact near-term contracts. Last month, AT&T announced its Supplier Domain Program 2.0, an update to the original initiative of 2009, which categorized all its equipment purchases into categories or domains, typically with two suppliers each, in order to streamline purchasing and network integration. The second iteration positions AT&T firmly as an SDN company and is specifically designed to ease the carrier’s path to a software-driven, ‘all-IP broadband, all-wireless, and all-cloud infrastructure’, as Tim Harden, head of supply chain, described it. The original domains have been redefined to reflect this focus, with a specific one for NFV, and contracts under the new system will start to be awarded early next year. The initial focus will be on elements of the wireless and wireline core.
Meanwhile, Deutsche Telekom said it had started the process of becoming a “software-defined operator, a shift Axel Clauberg, VP of aggregation, transport and IP believes will be essential to survival. He told the recent SDN Open World conference: “Current networks are not ready for the traffic growth that’s coming and competition is getting even tougher. ARPU levels are not going to be going up. We have a lot of complexity and it’s hard to be fast.’ That means moving to
“dramatically simplified networks” which are more comparable to those of the web giants, whose over-the-top frameworks are already heavily software-driven and cloud-based.
DT labels its own project TeraStream and envisages a simplified IP backbone infrastructure supported by virtualized appliances. One of the key changes that is needed to achieve the vision is a reworked OSS (operations support system) that is slimmer and more dynamic to detect and respond to network needs. Eventually, the OSS needs to evolve to become the control plane for the entire network, a key focus of R&D at Ericsson, among others. The German operator is doing its bit to sever the ties with proprietary platforms and vendors by basing its new network software on openly available, and often open source, technologies such as the KVM hypervisor and OpenStack. The first DT subsidiary to test the TeraStream model is Hrvatski Telekom in Croatia, which is working with Alcatel-Lucent and Cisco.
However, Clauberg admitted there was only so far a telco could go currently, since there was no carrier-grade SDN on the market, and he urged operators to keep up the momentum behind creating one, as DT did when it initiated the activity that became the ETSI NFV project.
Vendors need to respond rapidly too ‘ not just the big infrastructure and data center players but the providers of OSS, mobile core elements and, perhaps most critical of all, security platforms. Companies are increasingly recognizing this ‘ we have seen major roadmap announcements geared specifically to operators from Intel, Hewlett-Packard, Cisco, Alcatel-Lucent to name but a few this year.
Ericsson refined its own vision last week when it introduced its concept of ‘network slicing’, which carries the burden, along with the cloud-based announcements made at Mobile World Congress, of keeping the Swedish giant at the forefront of cellco infrastructure, even when that shifts to software. Network slicing combines a service provider SDN framework with a cloud platform ‘ in this case, Ericsson’s existing Cloud System ‘ so that functions are not just virtualized, but are then able to be deployed dynamically, wherever they are needed in the network, along with associated applications or data. This would enable operators to put key network applications in the cloud ‘ their own or a hosted service such as Ericsson’s ‘ and run cloud-based services for their customers too, hitting out at one of the most important businesses for OTT giants.
“What we try to do with Service Provider SDN is to provide opportunities for the networks inside the cloud to be connected to the WAN and let the WAN be controlled from applications in the cloud, something that is totally impossible today,” said Ericsson CTO Ulf Ewaldsson at an analyst day earlier this year.
This cloud/SDN combination is not new to Ericsson. Alcatel-Lucent and Huawei have been particularly prominent in unveiling such architectures, and many believe the cloud is the only way for operators to gain the full benefits of NFV, in particular, and to prevent it being controlled by proprietary vendors. It will also be important as a way to behave more like OTT providers, creating an environment where the platform does not distinguish between the network functions and the components of an application, and where a full ecosystem of apps and APIs can develop, covering everything from simple task automation to advanced network management. Earlier this year, a
group called CloudNFV was formed, highlighting the increasing association of virtualization with the cloud . Its idea is to run the OpenStack open source software, widely used in the enterprise world, on top of any server and create a private cloud environment for NFV. It believes the best way to keep NFV away from a new generation of proprietary implementations is to ensure it is tied to the cloud. In a blog post, its chairman Tom Nolle urged ETSI to “forget the notion of designing NFV to be capable of running on everything from bare metal to the cloud, and adopt the presumption of cloud implementation. That would simplify the process of specifying how functions could be deployed and operationalized, and take advantage of the enormous body of work cloud computing has produced.”
Ericsson will certainly not go that far unless it has to, but will cling to many of the hardware and software platforms it controls for a while yet ‘ at least while it builds up its own hosted cloud services. However, it is pushing the debate forward with ‘network slicing’, which aims to provision slices of capacity on the fly for different applications, adjusting to the qualities needed in each case ‘ different latency, speed, QoS and so on. In effect, a particular service gets a dedicated ‘slice’ of the network, adapted for its needs, for as long as it needs it. “Each slice of the network would have its own QoS, with the virtual machine able to be created anywhere and orchestrated by analytics and policy,” said CTO Ulf Ewaldsson at the launch.
At the heart of this would be an SDN controller, such as one under development at Ericsson, and orchestration software (in the Swedish vendor’s case, this will come from its BSS division, now centered on its Telcordia acquisition).
Ericsson says operators will be able to place their network applications in the same cloud or data center, where they already have back end functions like billing and BSS, integrating the two for greater efficiency. Ewaldsson said: ‘Most operators have investments in the private cloud, [mainly] to support billing and BSS systems, but there is also a pressure to put telecoms applications in the same cloud. This gives us the opportunity to connect those engines of SDN being used in the private cloud directly to the same SDN for the network.’
The aim is to use this combined platform, supporting wired and wireless access, to be able to deliver a far wider range of services and revenue streams ‘ more like an OTT player ‘ and to do this flexibly and on-demand. Ewaldsson continued: ‘With network slicing, we could have one stream that goes to something which needs very low latency. In 5G, we’re looking at devices needing latencies lower than five seconds. That is what would be required for remote controlling something. On the other hand, we might not need a lot of speed, but we might need new protocols to enable the cloud to know what is going on the road so it can control what’s happening there, for example. Each of these slices represents a certain quality of service and it’s a combination of cloud technology with virtualization of apps in the cloud that the operator would run.’
Despite the grand vision, there is still plenty of proprietary Ericsson hardware involved, notably a new multi-application line card for the SSR 8000 router. This runs on Ericsson’s SNP 4000 network processor and supports multiple virtualized network functions on the routers. It also claims to be able to “hyper-accelerate” the launch of applications when they are provisioned from the edge of the network. “Centralize what you can but distribute what you must,” said Jan H’glund, VP president of Ericsson’s IP and broadband unit, claiming that, despite the centralized visions of the ‘pure’ SDN and cloud proponents, many operators were asking for multifunctional edge routers too, in order to virtualize low latency tasks close to the user.
Much of this is futuristic and will take years ‘ and some fundamental shifts such as an entirely new OSS ‘ to achieve. However, Ericsson believes it must start work now, partly because some operators are starting to demand at least a few elements of the SDN/cloud vision, but also because it is the only way to stay relevant. Current network structures and the operators that run them will become redundant if they do not change radically ‘ on that big vendors like Ericsson and bit carriers like DT, so often at odds, can firmly agree. So Ericsson will try to ward off Cisco, and the cellcos will try to ward off Google and Amazon ‘ even as, in reality, they try to hold onto as many of their traditional architectural comfort blankets as they can.
NEC unveils first NFV mobile core:
NEC has shipped what it claims to be the world’s first commercial virtualized Evolved Packet Core (vEPC) supporting NFV specifications on off-the-shelf Intel-based hardware.
Part of the Japanese firm’s SDN Solutions portfolio, unveiled in July, the vEPC virtualizes functions such as MME, Serving Gateway and PDN Gateway on the x86 server. The first customer is a communication infrastructure project in Myanmar, which will go live in December.
The vendor worked with Intel, using its Data Plane Development Kit (DPDK) enhanced for virtualization, in order to enable high speed data plane processing and flexible configuration of system capacity between the control plane and the user plane.
Many operators believe the EPC will be the first network element to be virtualized and other suppliers are targeting this, both majors and start-ups.