Your browser is not supported. Please update it.

18 October 2012

Ericsson wants to rethink software networks for IP carriers

Significance: Last year’s Broadband World Forum gave mobile networks equal billing with wireline technologies, and Nokia Siemens, in particular, led the debate on achieving a fully software driven network. That discussion had moved on at this year’s event, and this time it was Ericsson whose voice was heard most loudly. With SDN and OpenFlow concepts now widely accepted in the 4G world, it believes a new definition of the platform is required specifically for service providers. And more broadly, they need to stop thinking about the differences between wireline and wireless, different devices or screens, and have an IP network which treats them all the same, and delivers new services flexibly, with intelligence in every element. That is a five-year goal, said Ericsson CTO Ulf Ewaldsson, but means the infrastructure vendors need to accelerate their move to being software suppliers.

The annual Broadband World Forum event, this year in Amsterdam, has become a traditional forum for the infrastructure majors to outline their vision of the new network ‘ increasingly one which regards wireline, cellular and Wi-Fi as a single platform and uses techniques like software defined networking, not just raw horsepower, to address the challenges of rising data demands. Last year Huawei and Nokia Siemens got their voices heard the loudest with their pronouncements on end-to-end IP systems with intelligence in every link of the chain. This year, Ericsson took the high ground, with CTO Ulf Ewaldsson claiming the era of ‘big data’ required a rethink of network design, and in particular, a software defined network (SDN) specifically for carriers.

He was more upbeat than Ericsson’s keynote speaker last year, EVP Johan Wibergh, whose soundbite was: ‘Today’s networks are not up to it, they won’t scale. We will need to move to fourth generation of IP networks.’ But the solutions proffered by the Swedish giant remain similar – streamlined architectures integrating LTE, wireline and Wi-Fi, with intelligence from end-to-end. Of course the details play to Ericsson’s own strengths ‘ the increased emphasis on Wi-Fi for wired or mobile operators reflects its purchase of BelAir Networks; a vision to integrate software defined networks tightly into back office systems points to its acquisition of Telcorida. Such takeovers have seen Ericsson evolve from being a mainly mobile infrastructure vendor to providing the platforms to support wied/wireline, all-IP convergence ‘ essential to retain its pole position with many large operators, and to fend off IP king Cisco and competitors like Huawei, which have always played on both sides of the fence.

Specifically, Ericsson’s version of the converged edge-to-core network embraces SDN, real time analytics, HetNet and a new generation of OSS/BSS. In his keynote address, Ewaldsson said: “In Sweden, the time spent on the internet every week increased from eight hours in 2008 to 19 hours in 2012, and the trend is similar in other countries. The demand put on the network has changed enormously … The entire app industry went from zero to 35bn apps in four years. We see increasing multiscreen viewing habits. This kind of never-ending innovation is a loop between consumer demand and network capacity and capability.’

As carriers struggle to prevent this loop turning into a vicious circle, he said they must stop treating wired and wireless networks as separate technologies and think entirely in terms of delivering relevant and personalized services to users over IP, whatever the connection or screen. This should be managed independently of the number of mobile or fixed subscribers, or the video bandwidth demand, he added. The cornerstones, according to the Ericsson world view, are ‘fourth generation IP’ (new wireline networks as well as wireless 4G); virtualization and SDN; a new generation of OSS and BSS; and heterogeneous networks, which include all cell sizes, spectrum bands and air interfaces.

The first IP network built for the initial internet boom was all about scale and horsepower, but its successor should be intelligent and ‘aware’ of different users, services, devices, locations and connection or browser types. This should lead to a combination of services which are delivered in an optimized way for the specific speed, screen and user preferences; plus networks which require less manual maintenance and planning because they are intelligent, pushing resource and capacity intelligently as required by the application or device.

This is partly delivered by individual tools such as traffic management and policy products for functions like deep packet inspection; or more intelligent BSS platforms; as well as self-organizing network approaches. But something more holistic is also needed, and that can be delivered by SDN, which supports the virtualization of the control plane from the data plane nodes. SDN, and OpenFlow ‘ the open source controller which underpins many initiatives ‘ have mainly been seen in the enterprise and web services sectors, but are becoming increasingly pertinent to carriers as they look for highly flexible ways to deploy pressurized network resources. There is even a mobile specific project based on OpenFlow, called OpenRoads or OpenFlow Wireless, which has been trialled in particular by NEC. OpenFlow, and SDN in general, take many of the concepts of virtualization, as applied to servers, and adapt them for IP networks. Virtualizing the network separates the routing intelligence from the hardware pushing the packets. Packet ‘flows’ are dynamically updated with each request for bandwidth from any RAN element or device (in a wholesale scenario, several providers’ RANs can be supported).

Many see OpenFlow as the missing link which will finally persuade carriers, beyond Verizon and a few others, to embrace IMS (IP Multimedia Subsystem) wholeheartedly. OpenFlow could help make IMS more palatable, because it can be included in the IMS fabric, improving flexibility, quality of service and cost-effectiveness when carriers are delivering a wide range of IP services to different networks and devices. The OpenFlow Controller manages network resources in a programmable manner, so carriers can set priorities and rules, and it controls all kinds of elements – routers, hubs, switches and so on ‘ as a unified and virtualized ‘slice’. The ability to manage these slices of resources from end-to-end is why many analysts believe OpenFlow works so well with IMS,in particular supporting many degrees of QoS in response to different types of traffic.

However, despite the promise of this vision, SDN was conceived for the enterprise, and Ericsson believes there is a need for a specific flavour for service providers. Ewaldsson called for the industry definition of SDN to be expanded, saying: “SDN in the telco-operator domain is much more than [the Layer 2 communications protocol] OpenFlow. It’s not about only managing the data planes in switches and routers, but about having fully integrated network control by connecting the SDN management layer to the cloud and network management. This offers a new view on networking. Operators can orchestrate network resources on different layers for different purposes under the same management system. In addition, the SDN architecture must also provide network awareness to the application layer through service exposure APIs.’ And in future, media and CDN (content delivery networks) would also be integrated, though Ewaldsson conceded this was a five-year evolution.

And he put in a plug for the Telcordia unit, saying this Ericsson subsdiary was working on another key aspect of ‘service provider SDN’ ‘ he explained: ‘We need innovation on the support system side. OSS and BSS and big data analytics provide the insight into the network and user behavior that allow the operator to adapt and manage the network to serve the user.”

The call for a service provider SDN is not just being led by vendors like Ericsson and Cisco. Verizon and some other operators are starting to voice similar wishes. Verizon has often used its market weight to push its suppliers beyond a base standard, to deliver a platform more optimized for its needs. As far back as 2006 it created the A-IMS initiative to build on the vanilla IMS standards and optimize them for real world carrier needs, especially in the CDMA community. And when it comes to SDN, Verizon also regards many of the standards as either inadequate or over-rigid. In particular, it has criticized 3GPP approaches to device virtualization, an important element of SDN in the mobile world.

As industry analyst Robert Syputa points out: ‘Verizon has viewed the way the implementation has been required under the 3GPP standards as being overly rigid, requiring EMS and other architectural components that are overly complex and limit innovation and cost reduction. I view this championing of OpenFlow as both being logical in its own regard and also a way to push for changes in the standards with a ‘built-in consensus’ that has developed from the cloud computing realm.’

Importing technologies from the data center and integrating elements of the wired and wireless networks in a common platform will be a long and difficult process, and carriers will be very reluctant to have to invest in new infrastructure. Ewaldsson recognizes this, envisaging a ‘hybrid SDN’ which runs a new architecture over existing routers and other infrastructure. However, where carriers are spending on new core network kit, the suppliers are racing to push their SDN credentials. Last month, Huawei looked ahead to ‘routers-as-a-service’ when it unveiled its VRP (virtual routing platform), which is SDN-enabled. It promises a high capacity hardware platform with “centralized control plane and software defined forwarding plane, enabling operators to build an SDN-ready network featuring fast service deployment, better resource utilization, and open interface for third party application support.”

New services can be introduced more quickly and cheaply because the routers can be adapted to manage any new offering, without the need for new forwarding hardware. In the metro router version, the packet forwarding function is separated from the service processing function, with the latter managed centrally using SDN tools. In the access network, Huawei offers network virtualization tools to simplify the management of multiple devices.

Huawei is riding a rising tide among router makers, looking to bring SDN to carrier networks. Cisco unveiled its software-based router in June, and Brocade integrated OpenFlow into its routers and switches in February.

SDN is just one example of how carrier networks are becoming software systems first and foremost. This is the concept behind Nokia Siemens’ Liquid Net, which sees all the intelligence and differentiation built into the software ‘ in a way that could theoretically be decoupled from its own hardware in future. Liquid Net virtualizes several network elements and uses beamshaping techniques to add capacity to an operator’s radio, transport and core networks on demand, a cheaper option than throwing more hardware at all these areas. According to Phil Twist, head of marketing at NSN’s network systems division: ‘The network is aware of the traffic running over it and can adapt to it. It can adapt minute by minute.’

The elements of Liquid Net all decouple key processes. In the access network, Liquid Radio centralizes baseband processing and divides resource between radios as needed; Liquid Core is an ATCA platform which runs core functions such as gateways and optimizes content delivery and caching; and Liquid Transport brings multilayer optimization and flexible control to the backhaul.

NSN has been redesigning its mobile gateways for years, moving its core elements to a common ATCA platform rather than an IP router. Its ATCA system can now support the functions of the entire evolved packet core, elements of the IMS for service delivery, plus policy management and specialized functions such as deep packet inspection and content delivery, says the firm. All these are linked by an end-to-end network management system called NetAct, which can reconfigure the virtualized network in real time as situations change.

It is a highly software oriented approach which moves away from creating a network around a series of purpose-built boxes and turns it into a set of applications built on common generic hardware. A ‘liquid network’ can move resource to the place it is needed, avoiding the wasteful requirement to provision every site for the moments of peak usage; and it can change its purpose ‘ for instance, majoring on content delivery when a big sporting event is being streamed.

So NSN will add more gateways and functions to Liquid Net, but the really radical idea is to move beyond ATCA and make the framework hardware independent. It could then work on other multi-purpose hardware platforms, in effect making NSN a software company, whose customers would load the Liquid Net network onto any off-the-shelf hardware. Of course, this could be an option for any of the infrastructure vendors as they put their smarts into the software, and see their base station sales and margins being squeezed. In future, several RAN suppliers will either exit the base station game or rely on OEM partners, and the same could happen in the core router space. At that point, they will become software providers, so the current moves to lead the long march to a fully software defined network will not only give them competitive advantage in their traditional markets, but may become vital to survival in future.