Most operators have started their moves towards virtualization and/or SDN (software defined networking) in the network by focusing on one or two fairly discrete elements, such as the packet core or even the CPE, often harnessing experience initially gained in the data center. However, major SDN proponents like AT&T, Telefonica and Verizon believe the maximum benefits will be derived if, eventually, SDN is applied to every layer of the network, enabling access, last mile backhaul, optical transport and data center platforms to be managed as one end-to-end system.
The past week has seen two important steps towards this goal. First, Alcatel-Lucent, whose advanced SDN activities will certainly be seen as key assets by its acquirer Nokia, claims to have implemented SDN at both the IP and optical layers for the first time in a commercial offering.
The company says its SDN-based Network Services Platform (NSP) is specifically targeted at carriers and can provision IP, Ethernet and optical services across all network layers, regardless of the vendor of the equipment. This aims to address a key challenge for telcos. At all layers, networks are provisioned at perhaps 40% capacity to allow for peaks of usage. That has led to successive waves of technology for dynamic resource allocation and for virtualizing various functions so they can share network capacity in a flexible way, which are now moving to the transport layer.
Virtualization moves some or all of the functions of the network from dedicated hardware appliances to virtual machines running on generic hardware, and creates tunnels between those VMs so they can be reconfigured on the fly without affecting the whole system. SDN separates the control and data planes, centralizing all the control of the huge number of virtual machines on standard servers and allowing the behavior of the network to be freely programmable, using standard interfaces and common developer tools. So a software defined transport layer would see all the resources centrally orchestrated, and then dynamically allocated by that control system. Links would be created and removed according to requirements, the process taking less than a minute.
By including transport, operators can move towards the goal of optimized multilayer data flows from end to end, not just in sections of their networks ‘ which could double network utilization, say supporters. However, until now, vendors have only supported the IP or optical domains, not both. By making that breakthrough, ALU says it is also bridging the gap between two key activities – service automation and network control – and so greatly reduce the time to activate new services. In some cases, that could be reduced to “seconds”, claimed ALU, rather than days or even weeks.
Overall, new offerings should be created about 58% more quickly, and using 56% less resource, it added – figures based on 40 customer trials and research by consultancy AGC. Once defined, services should be able to be provisioned almost instantly. Such improvements will be particularly important as operators meet their customers’ rising demand for on-demand, cloud-based applications and content, while seeking to husband their network resources efficiently to keep cost of delivery low.
The answer to that challenge is to combine the service automation, which enables the on-demand apps to be activated and delivered responsively, with the control of the network resources available, adding traffic engineering which can span all elements. To achieve the latter, the NSP provides a real time network view across the IP and optical layers, tightly coupling them using SDN. The NSP maps the network topology, the quality of the links, usage and performance, and then uses adaptive routing algorithms to work out the best route through all the layers. Optimization and real time path computation take place across the entire network, maximizing the operator’s flexibility to harness the best resource available. ALU says the NSP is designed to work with virtualized networks, as well as hybrid virtual/traditional set-ups.
Basil Alwan, president of ALU’s IP routing and transport business unit, said NSP offers “a simple interface that can deal with the complexity of network control across multiple technologies”. It will be available to customers from next month though ALU says it has already made one sale, with a tier one Asian operator. It may not have an actual contract with BT, but the UK telco was prepared to go public with its support for the approach. Its end-to-end network architect, Rob Shakir, said in a statement: “Tightly coupling service requirements and performance with network control, as in ALU’s NSP, promises to deliver real world benefits of centralized optimization whilst exploiting the strength of our existing distributed network.”
Manish Gulyani, the vendor’s VP of product marketing in the same business unit, told MobileWorld: “It enhances our SDN portfolio significantly with a carrier solution that is aimed at service providers’ wide area networks (IP and optical). We believe it is the first really unified platform that’s going to allow our customers to unify service automation, along with real-time network control. It will give them the capability to introduce an on-demand, simplified customer experience.”
The NSP is another piece in an increasingly complete SDN and virtualization jigsaw from ALU. It will complement another of the vendor’s crown jewels, Nuage Networks, a subsidiary which concentrates on data center and branch network SDN. And it will work with ALU’s CloudBand NFV orchestrator, which manages and coordinates all the virtual machines deployed by an NFV (Network Functions Virtualization) architecture. ALU insists the new offering is vendor-neutral and can be integrated with the operator’s existing OSS (operations support system), while working with any equipment that supports standardized interfaces. These are, on the southbound side, Netconf/Yang, PCEP, BGP-LS, OSPF, IS-IS Traffic Engineering (IS-IS)/TE, OpenFlow, SNMP and EMS/NS; and on the northbound, NetConf/Yang and RestFul API.
Gulyani claims the NSP is a significant differentiator for ALU, claiming that Cisco is trying to tie together disparate products from various acquisitions to achieve the same goal, while other competitors support only IP or optical (Cyan supports Ethernet and optical, not IP; Juniper only IP, and Infinera only optical).
That situation will spawn alliances, of course. This was already seen last year in Ericsson’s intensified partnership with Ciena, to be able to provide the entire hardware line to underpin a software defined, multilayer network. The alliance will integrate Ericsson’s IP play, the SSR (Smart Services Router), with Ciena’s WaveLogic packet optical platform, which has a stronger market position than the Swedish giant’s own offering. Ciena says it has started to develop an SDN controller based on sourcecode from the open source OpenDaylight project, in which Ericsson, among other telco equipment providers, has taken a prominent role.
The new importance of optical integration may also spark M&A. Many operators have bought optical equipment from specialists without the need to integrate it deeply with their routers, but as the network starts to converge, from Layers 0 to 3, we will see data and optical gear merging into a broader platform for virtualized wide area routing.
Such trends could see some optical transport specialists, such as Ciena, ADVA and Infinera, in high demand as partners or acquisition targets for the router giants like Cisco and even Juniper.
And some of ALU’s key rivals will be hard on its heels. Last fall, Huawei unveiled an SDN-based unified network controller which can automatically detect and maintain resources at the IP and optical layers, as well as the resources between the layers. The flow engine has been trialled by several operators, which claim network throughput improvements of 60%, and it is sure to be the basis of upcoming, broader IP/optical capabilities.
A recently published study by Bell Labs, ALU’s R&D arm, showed that by converging routing and optical transport technologies, operators could reduce their network resources by 40% and achieve the same service availability. The famous research unit has been vital to ALU’s SDN efforts, and it seems that, in order to enable the NSP, it must have cracked some of the significant mathematical and algorithmic challenges associated with multilayer network optimization.
Bell Labs also says that carriers can support 24% more revenue-generating traffic by using sophisticated algorithms to distribute new connections across the network intelligently and efficiently.
In the second step forward in this area, the Optical Internetworking Forum (OIF) has approved a transport SDN framework, based on results from a large-scale multivendor prototype demonstration held last year. This will help provide carriers with common tools and APIs (application programming interfaces) to simplify the application of SDN to the transport layer, reducing the cost and risk of this major technological and cultural undertaking.
It has published its approach in the form of a white paper, ‘Framework for Transport SDN: Components and APIs’. Lyndon Ong, co-chairman of the OIF’s market awareness and education committee, said in an interview: “This document will help both carriers and vendors place their ongoing SDN efforts within the context of a framework. It recognizes SDN is not a new domain, but rather is being used in multi-domain carrier networks and is integrated across diverse environments, featuring a diversity of vendors and technologies.”
An important part of the work was a multivendor demonstration, held last October, involving prototype gear at various stages of commercial development. Working with the Open Networking Foundation (ONF), the OIF has been overseeing testing in several carrier labs in North America, Europe and China and published its first results last fall, claiming the prototypes were stable and commercialization was not far away. For instance, AT&T aims to have a WAN SDN controller in place in 2016 before moving on to converging its IP routing and optical transmission layers around SDN.
The application tested in October was cloudbursting, or increasing the bandwidth between data centers to support cloud services, a common business case for carrier SDN. Based on this kind of work, implementation agreements can be submitted to standards bodies such as ETSI, IEEE and the ITU.
The next step will be to define the northbound interfaces between applications and the SDN control layer and this work will appear in further multivendor demonstrations and a white paper in late 2015 or early 2016.
To integrate the optical and transport layers, even with SDN, is a major project, and means applying centralized control, standards-based programmability and virtualization everywhere. That will need to be justified by more than just immediate cost efficiencies. The real justification for the huge disruption of SDN will lie in supporting new services and revenues, enabled by the endless flexible and reconfigurable network; and by harnessing the vast amount of traffic and user data that will pass across it, to improve performance further and to create new customer experiences.
All that should be a significant boost for the optical sector. As SDN shifts carriers’ investment from the RAN to the telco data center, the resulting centralized, virtualized systems will require very high quality transport to handle all the traffic; to deliver good services and customer experience; and to feed increasing levels of traffic on the mobile backhaul links. Therefore there will be a rising use of high capacity optical technology, which will, in future, be integrated with the switching function (such as label switch routing), and control layer/management software to create a programmable transport layer. That can then connect to the cloud and virtualization platforms through APIs for orchestration and automation.
Products like ALU’s NSP point in that direction, and optical specialists are also getting SDN-ready. Ciena’s WaveLogic Photonics SDN tools, for instance, can apply network visibility and intelligence for the firm’s OneConnect Intelligent Control Plane, to span the OTN (optical transport network) switching layer and the photonic layer. “All layers of the network need to become more programmable to application needs,” said Mike Adams, VP of product and technical marketing, at the launch last October. “The more visibility you have into the photonic layer, the easier it is to think of that layer as programmable.”
Infinera also unveiled multilayer control plane capabilities in software, indicating how even the rarefied optical areas of the network are starting to feel the effects of an SDN movement that started in the data center and the higher layers. Start-up Vello Systems has even formed a body, the Open Source Optical Forum, to apply OpenFlow and other open standards to optical gear.