Virtualising network functions, along with software defined networking (SDN), is part of the wider telecoms wide area network (WAN) infrastructure trend towards virtualisation and commoditisation.
The network function virtualisation (NFV) focus is primarily on optimising the network services operated by WAN infrastructure providers (here called telcos), whereas SDN seeks to separate the control and forwarding plane for a centralised view of the network.
The main drivers for NFV adoption by telcos are escalating infrastructure costs and the change in network usage driven by the mass adoption of cloud computing. Internet content providers such as Google and Skype are delivering new video and audio services and content to users much more quickly than traditional telco service roll-outs, and the rise of cloud services has boosted WAN network traffic loads. Telcos need to keep up or go down.
Change in mindset
NFV enables a fundamental change in WAN provider mindset, and addresses all WAN elements from customer premise devices to network access, across the service edge, and right through to the core WAN. Telcos are being forced to justify their investment priorities and operating costs in terms of business value rather than evolving their infrastructure to meet specific performance metrics, irrespective of cost.
As a result, many telcos acknowledge the need to speed up implementation of new services and features. Doing so with a dedicated hardware platform takes too long, takes up too much office rack space and is energy inefficient. It costs too much.
This new mindset shifted the focus from proprietary hardware performance features to open standard software residing on standardised hardware. The affected range of hardware is extensive and includes: routers, firewalls, gateways, content delivery network (CDN), session border controllers, network address translation, WAN accelerators, quality of experience testers and radio network controllers.
Along with cloud automation and SDN programmability, NFV aims to relocate all these device functions onto standard high volume servers, storage and switches with software from independent software suppliers.
More coherent approach
NFV emerged when WAN operators tried to deploy standardised IT virtualisation technologies, but found they needed a more coherent approach when confronting traditional network infrastructure suppliers.
In December 2012, this led seven major telcos – AT&T, BT, Deutsche Telekom, Orange, Telecom Italia, Telefonica and Verizon – to band together and create the NFV ISG (industry specification group) under the European Telecommunications Standards Institute (ETSI).
Since then, 30 other telcos have joined the NFV ISG, plus 250 WAN hardware and software suppliers and communications research bodies. The creation of ETSI NFV ISG resulted in the foundation of NFV’s basic requirements and architecture in 2013.
Once telcos embark on the NFV transition journey, the payoff comes as:
- Lower hardware costs by avoiding supplier lock-in.
- Extending the lifetime and functionality of hardware platforms with software upgrades and add-ons.
- Having a much wider range of performance management features.
- Expanding the number of possible providers, notably with smaller suppliers coming out of the IT and DevOps environment and competing against established players on both price and performance.
- Adapting infrastructure to a DevOps environment where development, testing and implementation is possible on operational infrastructure.
There is no single standard or coherent hardware architecture you can adopt when planning a shift to NFV, so a buyer’s guide to NFV is as much an implementation guide as a product review.
NFV Release 2
The ETSI NFV ISG Release 2 specifies NFV compliant hypervisors, functional requirements for management and orchestration interfaces, interfaces to acceleration mechanisms, and the virtual network function manager (VNFM) that handles scaling, changing operations, adding new resources, and communicating the states of virtual network functions (VNFs) to other functional blocks. NFV Release 2 came out in 2015.
The current work on Release 3 began in May 2016 and will stretch into 2017, adding specifications for actual service deployments, including: billing, end-to-end management, multi-site connectivity, lifecycle management and security. It also works closely with other NFV initiatives, notably the Linux Foundation open source reference platforms, which include: ONOS, OpenO, ECOMP, CORD, Open vSwitch and OpenDaylight.
Also, there are at least 13 different standards organisations, such as 3GPP and TMF, working on fixed and mobile WAN infrastructure standards with NFV elements.
From a fully standards compliant product point of view, it is still very early days. ETSI will not host the first Release 2 Plugtest to test the alignment of NFV end-to-end configurations with components from different providers before January-February 2017.
Get their act together
But equipment manufacturers are not waiting around for telcos or the ETSI ISG to get their act together. They are already offering NFV products that are compliant with the specifications that come from the NFV-related industry consortia under the Linux Foundation umbrella. So any NFV-compliant product comes with a lot of strings attached, and buyers must investigate thoroughly what compliance any NFV product is actually based on.
Let’s not forget that the rising importance of SDN and NFV strategies underpinning WAN infrastructures depends on continued improvements in hardware processing and transmission capacity from the chip level and up. So competition for hardware performance and efficiency will continue, but with the incumbent WAN infrastructure suppliers shifting their priorities towards optimising their architecture, simplifying operations and increasing processing velocity to operate in a standards-based multi-supplier environment.
Typically, major infrastructure suppliers such as HPE, Ericsson, Huawei, Nokia, Juniper Networks and Cisco are combining their SDN and NFV product ranges to address performance challenges on fixed and mobile infrastructures. Specialised NFV contributions come from a wide range of suppliers, such as Amdocs for billing, Ciena for orchestration and Red Hat for virtual infrastructure management (VIM).
HPE’s ContexNet is based on the Linux OpenDaylight carrier-grade SDN/NFV fabric and is being deployed in conjunction with Intel. It enables a virtual mobile core and access infrastructure that allows mobile cloud service providers to virtualise network functions and deploy services faster to accommodate rapidly multiplying mobile data sources. We are seeing many similar SDN/NFV products undergoing field testing.
Early telco NFV adopters include NTT Docomo and Indonesia’s XL Axiata and are working with several major NFV suppliers. On the fixed network side, Ericsson is providing them with its OpenStack-based cloud execution environment as the virtualisation platform and its NFV Cloud Manager for orchestration. The system initially supports a multi-supplier virtual evolved packet core (EPC) deployment and also provides system integration and support services. On the mobile infrastructure, NTT is trialling Cisco’s Application Centric Infrastructure (ACI) to support its long-term evolution (LTE) systems for handling the company’s mobile communications.
In the UK, BT’s access network division, Openreach, has contracted Huawei and Nokia to deploy an NFV compliant G.fast broadband infrastructure to reach 10 million UK homes and businesses by 2020.
In France, Orange Business Services has rolled out the SME EasyConnect VPN service. This takes advantage of SDN and NFV technologies to let customers quickly set up and manage an IP-VPN, as well as firewalls, web content filtering and other services, using a self-care portal. The services rely on Ciena Blue Planet technology for orchestration and VNF management. Orange has been working with Red Hat for VIM services, and tasking Juniper Networks Contrail to provide the SDN controller.
Across Europe, Vodafone is working with Amdocs, Juniper Networks, Aria Networks, Red Hat, Adva Optical Networking and Fortinet to build an advanced virtual private network (VPN+) using SDN and NFV components to increase performance and keep hardware requirements to a bare minimum. The aim of a VPN+ is to automate and simplify deployment and incident management processes, while providing information about multiple parameters (analytics, performance and security monitoring). Smaller point-product suppliers such as Sonus (session border controllers), OpenWave (media optimisation), Qosmos (DPI), Callup (mobile device management), Peplink (virtual SD-WAN routers) and ADVA (optical networking) see significant opportunities with NFV to target much larger WAN infrastructure deals than would otherwise be possible, because they don’t have to provide all the pieces.
Adva develops high-performance FSP 150 ProVM hardware that works with open software and hardware from other suppliers. Last year, it acquired and integrated the Overture Ensemble software suite into its optical networking hardware to launch its NFV strategy. Adva’s Ensemble NFV is designed to allow service providers to innovate quickly, explore experimental service offerings and create try-before-you-buy programmes, hosting virtualised functions on any commercially available servers.
Peplink’s Multi-WAN FusionHub routers simplify the use of MPLS, xDSL, fibre, 3G/4G/LTE and Wi-Fi as a WAN, and have been designed from the start as intelligent multi-WAN capable devices providing full WAN virtualisation capabilities. FusionHub runs on mainstream virtual machine software, including VMware, XenServer and VirtualBox.
Clearly, NFV and SDN represent a step-change for WAN infrastructure providers in their competition with internet-based digital service providers, and in meeting the expectations of their customers who are shifting massively to cloud services and mobile access to these services. And while there is no going back, moving forward in a still loosely standardised tech space has its own implementation and integration challenges. Some telcos find it too risky to introduce new processes and product simultaneously, and may virtualise their network and service platforms first.
There are also economic repercussions for telcos investing in an NFV infrastructure. The pace of adoption is not just a standards and technology issue, it is also about the ability to bill customers and the revenue sharing between NFV suppliers and the infrastructure providers.
Early adopters are concerned about the NFV licensing, where there is a “misalignment” between the way an operator wants to sell services and the way suppliers provide licences – something seen in the early days of cloud services. One cost-saving option may be to run multiple services on the same virtual network or container and introduce more open source technologies.
Bernt Ostergaard is a service director at advisory company Quocirca.