BY CAROLINE GABRIEL
This is a sample of our full paid service Wireless Watch, click the link below for 4 week FREE trial. Wireless Watch Trial
As outgoing CEO John Chambers hosted his last Cisco Live annual conference, it was fitting that the centrepiece of the event was virtualization. The somewhat abrupt timing of Chambers' retirement indicated that he felt the giant IP company was at an important turning point, which made it the right time for a change of leadership.
The move, by carriers and enterprises, from physical to virtual networks and towards software-defined networking (SDN) will define the next few years in Cisco's core business and decide whether it remains in the dominant position which Chambers, despite recent turmoils, carved out for it.
As part of a broad set of SDN announcements, Cisco executives spent some time focused on operators, saying that their progress towards virtualization was more rapid than many had expected. It was anticipated that carriers would largely confine their first-wave efforts to relatively discrete platforms like the packet core, but in fact, according to Kelly Ahuja, Cisco's SVP for the service provider business, they are actually trialling virtual network functions, as well as SDN, across a wide range of elements, including firewalls, load balancers and VoLTE infrastructure.
"Most operators are telling us - look, my network is a chaos, an IT construct is a chaos. Virtualized chaos is still chaos, so what you got to do is show me what value these things can add for me," Ahuja said.
In that context, Cisco is focusing on one of the areas with the greatest potential to turn virtualization into gold for operators - virtual managed services, which provide the ability to support large numbers of third party offerings, from flexible MVNOs to corporate customers' services. "Business customers are where we're starting to see most relevant application and deployment of that," Ahuja said, according to SeekingAlpha.
Building platforms to support large numbers of customers' and MVNOs' offerings is a key commercial goal of AT&T's Domain 2.0 SDN program. Cisco is working with the US carrier, and with rival Verizon, on their SDN projects, which are among the most advanced in the telco world. This will cheer investors, who have been concerned at signs that Cisco might start to be squeezed out of its central position in the infrastructure of AT&T and other major customers.
The first wave of Domain 2.0 vendors announced by AT&T indicated the disruptive effect SDN could have on its supplier roster, and included Metaswitch, Tail-f and Affirmed Systems, and excluded Cisco, (though Cisco subsequently hit back by acquiring Tail-f). Verizon, which recently announced its own SDN program, has stuck with traditional partners in phase one and will add specialists and start-ups later - its key vendors are Cisco, Alcatel-Lucent, Nokia, Ericsson and Juniper. These large companies will create the framework architecture and the interface specifications, and Verizon's first targets are the relatively low hanging fruit of SDN - the data center, packet core and IMS.
To help bolster its position in the carriers' SDN roadmaps, Cisco is steadily adding to its platform, and says it now supports 15 virtual network functions for telcos. In a long list of announcements at Cisco Live, the most carrier-focused were additions to the Border Gateway Protocol (BGP) EVPN. This is targeted at service providers and it has now gained a Virtual Topology System (VTS). For operators requiring a programmable fabric, VTS adds the ability to provision and manage a VXLAN, based on BGP EVPN, as a software overlay across Cisco Nexus switches. (VXLAN is a proposed protocol for running a virtual network on existing Layer 3 infrastructure, and an important tool for allowing operators and enterprises to migrate gradually to SDN, without sacrificing hardware investments.) VTS will support any virtual switch that supports the BGP protocol.
Cisco first introduced BGP EVPN on the Nexus 9000 switch early this year but it is now being extended to the Nexus 5600 and 7000 families, and to the modular Nexus 9500. However, the Nexus 1000v does not yet support BGP.
While programmable fabric will be important to large carrier SDN programs, some smaller ones may use the more pre-packaged Cisco offering, Application Centric Infrastructure (ACI), which is mainly targeted at enterprise customers and the SDN 'mass market'. New extensions to the fabric software include extensions to support integration with Microsoft Azure; plus a plug-in for VMware vCenter plug-in; an ACI toolkit for simplified network provisioning; and a 'stretched' fabric that can extend from 30km to 150km over DWDM, pseudowires and dark fiber for multi-site data centers.
Cisco also announced two Nexus 3000 Series switches based on merchant silicon. The Nexus 3232C is a 32-port 100G switch based on Broadcom's Tomahawk chipset, while the Nexus 3264Q is a 64-port 40G variant. Both will ship in the third quarter with prices starting at $35,000. They indicate another disruptive aspect of SDN - the shift towards hardware which is more commoditized in design, openness and, of course, price. That in turn will drive vendors like Cisco, which have relied heavily on ASICs designed inhouse, towards the economics of merchant silicon, boosting providers like Broadcom.
BY CAROLINE GABRIEL
The Small Cell Forum has unveiled the latest instalment of its Release Program, this time focused on virtualization, a trend which will have a deep impact on small cell deployment and may, in time, spur adoption of new HetNet architectures.
Release 5.1, like its predecessors, consists of case studies, technical primers, best practice guides, business cases and other resources, designed to set out technical and commercial options and to lower the barriers to deployment for would-be small cell adopters. It is a point update to Release 5, and adds 12 documents, five of them focused on virtualization (other topics include LTE-LAA and 5G).
The SCF put virtualization on its agenda a year ago, and now it has laid out its roadmap. At the heart of this is nFAPI (Network Functional Application Platform Interface), which underpins the Forum's preferred architecture split for a virtualized small cell RAN. It provides the interface bwteen the radio head, where most of the latency-specific RF activity remains, and the centralized media access controller, in which most of the intelligence is concentrated, and which may be in the cloud or more localized.
One of the big debates around vRAN, of course, is where the architectural split should occur, and how many functions should be centralized or distributed. Real world choices will depend on the individual operator's objectives and resources (for instance, access to dark fiber), but the Forum believes a common framework approach will help some carriers accelerate and simplify their projects. But it acknowledges that one approach will not suit all, and also plans to explore a second approach, centered on PDCP (Packet Data Convergence Protocol - the layer of the LTE stack which performs IP header compression and decompression).
As David Chambers explains in his ThinkSmallCell blog this week (http://www.thinksmallcell.com/Standards/small-cell-forum-become-virtual-and-paint-hetnet-roadmap.html?utm_source=feedburner&utm_medium=feed&utm_campaign=Feed%3A+ThinkFemtocell+%28Think+Small+Cell%29), the main advantage of the PDCP split would be to avoid increase in backhaul overheads. Backhaul and fronthaul capacity and latency requirements are among the barriers to vRAN deployments, but the Forum says its nFAPI approach limits the impact on this, as backhaul can still be carried over IP, though with lower latency tolerance. The body said the approach was "well aligned with the packet switched backhaul service conventionally used to support small cell deployment".
Forum chairman Alan Law said: "We've found clear benefits and drivers towards centralisation and virtualization of the small cell network. These facilitate the scalability of small cells and enable functions to be moved around depending on loading conditions or availability of compute and transport resources."
A Rethink survey of mobile operators in 2014 found a clear overlap between small cell densification and vRAN in carriers' plans for next generation network architectures. With about 50% of MNOs planning some elements of vRAN or Cloud-RAN in their five-year plans for LTE, 90% of those also had urban small cells on their roadmap. And over one-third of those respondents said vRAN activities would be specifically motivated by the need for densification. Reasons to kick off vRAN projects in the small cell layer include the belief that virtualization will enable resources to be assigned more flexibly and cost-efficiently; that large numbers of cells will require centralized and automated resource allocation to avoid over-capacity and network chaos; and that there will be less risk, and more immediate benefits, from applying virtualization to a new and somewhat discrete small cell layer, rather than the well-established macro network.
Sue Monahan, CEO of the SCF, said: "Some of our most influential work to date has been around integrating small cells with not just the macro network, but with service integration, WiFi technologies and enterprise networks. Our new roadmap will help operators accelerate the delivery of integrated HetNet deployments, while also leveraging new trends from virtualization through to 5G, M2M and LAA." [Licensed Assisted Access - LTE in unlicensed spectrum.]
The Forum says 11m small cells have been deployed as of March 31, though the deployment of public access networks remains in its infancy - 95% of the base is residential, while there are 450,000 enterprise small cells installed, 16,500 in urban environments, and 20,000 for rural and remote applications.
Some players, such as ip.access's CTO Nick Johnson, believe that deployment of public access small cells will only be jumpstarted when there are better platforms to support multi-operator services. The ability for multiple providers to offer services on a common small cell network requires better management and quality control than the standard, MOCN, enables, since it takes a 'free-for-all' approach. Frameworks should allow a primary operator to build out a network (possibly working with an integrator, 'as-a-service provider, or vertical market partner) and then open it up to other MNOs and MVNOs.
The Forum's next major Release will appear in about a year's time. Release 6 will focus on the broad issues surrounding the evolution of the HetNet. This may look at some of the emerging architectures which their developers hope will be part of '5G'. For instance, NTT Docomo's Phantom Cell concept integrates small cells and virtualization, in a currently proprietary way which could feed into standards.The Forum believes one of its contributions will be to collate and distil the requirements of a large base of operators, and then feed the common themes into the vendor ecosystem and into standards bodies and specifications.
By Caroline Gabriel, Research Director
Analysts are in a unique position to assess the way service providers are thinking, and every so often, there is a tangible change in that thinking, which impacts on the whole ecosystem. In the past couple of years, one of the key watchwords has been ‘multivendor’.
Mobile operators have tended to stay in the comfort zones of end-to-end systems supplied by one or two well-established suppliers. That has some advantages, notably a high level of integration and of vendor stability. But it also has many disadvantages, including the limited freedom for operators to shop around for the best solutions, and drive competition and better pricing with a vendor neutral approach.
The profound changes taking place in the mobile architecture as 4G evolves towards 5G are seen by many operators as an ideal opportunity to break the stranglehold of the traditional equipment suppliers. Heterogeneous networks made up of large numbers of small cells, and the move towards virtualization, both encourage a multivendor approach.
Indeed, in a recent survey of MNOs, conducted by Maravedis-Rethink, the ability to adopt a multivendor strategy emerged as the second most pressing reason to invest in a HetNet (after cost: capacity improvements).
However, a decision to mix and match network equipment has a knock-on effect throughout the procurement process. For instance, it will be essential to have tools which can support equipment from any supplier. Network optimization systems, for example, will need to be able to work with a widening range of RAN architectures, and with equipment from any supplier, present or future.
That makes a strong argument for a vendor-neutral approach to optimization tools, especially as these become increasingly critical to the successful implementation of the network. Functions such as self-optimizing networks (SON), cell planning and dimensioning, video traffic management - and the many other tools which are needed to make a modern RAN perform at peak efficiency – will no longer be useful add-ons, often bundled in by the network equipment provider. Rather, they will be essential for the operator to maximize capacity and flexibility from the complex new architectures, and so to ensure return on investment (ROI).
In the small cell world, then, multivendor deployment and high levels of optimization, especially SON, will often go hand-in-hand. This helps to explain a shift in thinking, visible in two surveys of MNOs, conducted in 2013 and 2015. In the earlier study, over 70% of MNOs planned to rely primarily on their equipment vendor for optimization. In 2015, only 15% were committed to an NEP solution for future projects.
So as the HetNet takes off, there will be an increasingly strong argument for vendor-neutral optimization solutions, whether tools or hosted services. These are better designed for multivendor networks and come from suppliers whose primary business is in optimization, not selling equipment. This significant change in operator thinking is set to open up a competitive new market for specialized network optimization vendors as the 4G HetNet evolves.