Few technology platforms have evolved as quickly as NFV (Network Functions Virtualization), at least in terms of a set of common specs which have been embraced rapidly by vendors eager to be part of cellcos’ next generation roll-outs. But to deliver the kind of impact on network economics that virtualization promises, it must be deployed on a wide scale by operators. Most carriers are interested and positive, but a few are determined to accelerate actual roll-out, including Telefonica, AT&T and now NTT Docomo of Japan.
The biggest Japanese cellco has worked with long-time R&D partner NEC, as well as Alcatel-Lucent and Cisco, to create a proof of concept for a virtualized evolved packet core (EPC). All three are also part of Telefonica’s huge multi-country NFV trials, and, with other vendors, are starting to accumulate important trial experience with major carriers. After the success of this test, Docomo says it expects to deploy services on a commercial virtualized EPC by the end of March 2016 (the end of its fiscal year).
The EPC is the element of the network which most carriers expect to virtualize first. This can be done relatively in isolation from the rest of the system and even in the cloud. It has gone some way to centralization already, and operators are deploying new EPCs anyway, with LTE, and to accommodate VoLTE.
NFV moves some or all of the functions of the network from dedicated hardware appliances to virtual machines running on generic hardware such as servers and switches, and creates tunnels between those VMs. That means VMs can be created or moved without reconfiguring the network each time, while there are savings on the cost and complexity of specialized hardware.
Docomo said it had successfully carried out joint verification tests, and that each of its partners had verified ‘cutting edge’ network control mechanisms, in separate exercises. These included highly efficient handling of user traffic and automatic network recovery – to ensure that, in the event of hardware failure, a back-up could be quickly and automatically created on different hardware.
The operator said it would collaborate with other vendors too, emphasizing ‘ like Telefonica at this year’s Mobile World Congress ‘ that one advantage of virtualization would be the end of being locked into just a couple of suppliers. Seizo Onoe, CTO at Docomo, also believes it will be essential to work with a range of partners to accelerate progress. ‘NFV is highly expected to change the ecosystem of network industries,’ he said in a statement. ‘But without a high degree of collaboration among the players, such hopes could end up like pie in the sky.’
Also on the virtualization front, Alcatel-Lucent said it was working with Mobily of Saudi Arabia on the first commercial deployment of the vendor’s virtualized wireless controller ‘ and the first NFV deployment of wireless control.. The lightRadio 9771 Wireless Cloud Element RNC is part of ALU’s vRAN portfolio, which is also part of China Mobile’s extensive Cloud-RAN proof of concept work.
Alaa Malki, CTO of Mobily, said in a statement: ‘We have a very technologically sophisticated customer base and we need to stay at the forefront of technological innovation to provide the level of experience that will keep them satisfied.’ And Amr El-Leithy, president of ALU’s Middle East, Africa, Turkey and Azerbaijan region, said: ‘This is the first deployment anywhere for the virtual RNC by Alcatel-Lucent and it authoritatively demonstrates that carrier grade cloud network solutions is a reality. Today, we have achieved a significant milestone with our virtual controller now handling live traffic in Mobily’s commercial network.’
The vRAN remains a medium term objective for most carriers, however, because so many have plowed large sums in recent years into new, but conventional, 3G+ and 4G infrastructure. The EPC is not the only early-stage target though. Certain routers, accelerators and content delivery networks are also in the frame, and Telefonica’s first proof of concept tests, within its huge Unica virtualization initiative, centered on the virtual CPE, as demonstrated in Barcelona in February.
At MWC, the vCPE was shown off in a set-top box/multiscreen TV application. In this approach, all the services currently defined in a physical customer premises box will be defined in the network. The virtual CPE is in trials in Brazil and could go commercial in Q414. Other phase one moves, as set out by CTO Enrique Blanco, will be in the EPC and IMS, as well as virtual deep packet inspection (DPI) elements, set-top boxes (STBs) and DNS (domain name system) servers.
In stage two, from 2015 to 2016, Telefonica aims to virtualize PCRF (policy control resource function), B-RAS (Broadband Remote Access Server) and SDP (service delivery platform) functions, and add a real time analytics platforms and SDN orchestrator. The firm also has a Cloud-RAN field trial in Spain, but in common with most European and US carriers, regards the virtualized RAN as a phase three activity. It has not been placed on the firm roadmap yet, and Blanco insisted: ‘Our initial focus is on the IT and signalling platforms.’
Blanco said that, using current network deployment approaches, he believed Telefonica could achieve 10%-15% increases in network efficiency, but since traffic growth was running at up to 30% a year, this would be inadequate ‘ hence the urgent need to move to new, software-driven architectures to aid cost efficiency, agile network planning, scalability and new services. With NFV, Telefonica believes it can reduce operating expenses by 30% or more.
He pointed out that the carrier started virtualizing data center capacity in 2011 and about 40% will be virtualized this year, rising to 80% by the end of 2015. It will now start to introduce virtualized network functions, one by one, and expects to have about 30% of its platform running in virtual machines by 2016. Blanco stressed that this would require heavy support from the IT departments, and a multivendor approach. He told reporters: ‘We have to avoid vendor lock-in, so working with multiple suppliers with an open approach is vital. Some of the vendors might not be happy because it is cannibalizing their business, but we are working with them on that.’
The Spanish giant has been working on three major proof of concept trials for NFV ‘ with Huawei in Spain, ALU in Mexico and NEC in the US, while other vendors including Ericsson, BroadSoft, HP, Cisco, Brocade, Intel and Juniper are also involved in various capacities. But it is now looking beyond the virtualized functions themselves and towards the orchestration layer – perhaps the critical success factor for most projects and the area where vendors like Cisco hope to play most strongly, but one where there is no standard NFV implementation available as yet.
This is one aim of Telefonica’s newly created NFV Reference Lab, which is focused on developing the upper service orchestration layers as well as the virtual functions themselves. It also shows how major operators are not just keen to open up their own ecosystems to many suppliers via virtualization, but to take the reins of the inevitable open source efforts too. Rather than just tapping into vendor-led standards and open source activities, they are being proactive in such areas, and some are contributing considerable amounts of IP into open source communities themselves.
Telefonica’s first major orchestration activity is in partnership with Intel and Red Hat. The trio will work on a virtual infrastructure management (VIM) platform running on Intel servers. It will harness open source software and many of its enhancements to that will be put back into the open arena.
The three companies will each commit engineering and testing resources to the lab, which will be important to UNICA as it will evolve a common framework to enable Telefonica to use many vendors’ offerings. The new software will tie together virtual machines and allow them to be deployed in specific locations, such as a particular rack or blade in a networked data center, rather than just ‘thrown into the cloud’. The partners say it will improve performance if virtualized network functions are positioned in a ‘deterministic” way, rather than the cloud stack deciding where they should run. According to LightReading, Red Hat will include the resulting orchestration software in OpenStack and in its own software.