With the recent buzz around SDN and NFV (network function virtualization), we at SDNCentral wanted to understand how real SDN and NFV was at service providers, and what the real-world implications were. So we reached out to some folks who both understand the operational implications and lead the movement within the service provider space. Javier Benitez is a senior network architect at Colt and represents Colt at the ONF and NFV. He also serves as a technical lead for major projects such as IPv6 and next-generation datacenter fabric. Nicolas Fischbach is director of network and platform strategy and architecture at Colt and is in charge of researching, defining and driving Colt’s mid-to-long term network and platform strategy.
With that, gentlemen, let’s begin the conversation:
SDNCentral: For those readers unfamiliar with Colt, who is Colt, what services do you provide and who are your primary customers?
Javier: “Colt is one of Europe’s largest providers of integrated business telecom and IT services, serving 30,000 customers across European markets. We provide these major organizations, midsize businesses and wholesale customers with a platform that combines network, data center infrastructure, managed IT services and expertise in integrated computing and network services.
Colt operates a 22-country, 43,000km fiber network that includes metropolitan area networks in 39 major European cities, with direct fiber connections into 19,000 buildings and 20 Colt carrier neutral data centers. Colt also operates a 27,000 km transatlantic and US East Coast network.”
Nico: “From a technology point of view, we have the full spectrum of networks and platforms: a long-distance optical DWDM network, metro optical, DSL and EFM (Ethernet-in-the-First-Mile) access networks, a business Carrier Ethernet platform and an IP/MPLS backbone for Internet and IPVPN services. On top of those we deliver TDM Voice, VoIP, IN (Intelligent Network) and UC services. Our managed IT services build on our compute, storage, backup and security platforms.”
SDNCentral: Why is Colt interested in SDN?
Nico: “Virtualization and cloud computing have completely revolutionised the IT world, delivering powerful features like elasticity, scalability, on-demand and automation. Colt was one of the pioneers to offer carrier-class cloud-based business services. However, in parallel to this flexible cloud computing world, we have to manage networks which – put simply – live and evolve at a different pace. They are often more static and rigid, have more physical constraints and require more manual interventions. SDN attracted Colt’s attention as a means to bring software programmability to the network arena, with the ultimate objective of augmenting networks, making them more elastic, scalable and automated. This is a critical element in enabling the transformation of our business from a provider of bandwidth and connectivity to a provider of IT services.”
SDNCentral: How do you define SDN? What’s the impact of SDN at Colt?
Javier: “We all know that SDN is difficult to define these days, and every person I’ve been talking to has their own definition and view on what SDN is – and is not. The way we like to think about SDN at Colt is in terms of “Network Programmability” and this is the internal name that we refer to inside the company. One thing is very clear to us: we want to be very pragmatic in our approach and avoid any sort of dogmatism. This is why we follow (and contribute where we believe we can add value) most developments, be they from ONF, IETF or the newly created European Telecommunications Standard Institute Network Functions Virtualization Industry Specific Group (ETSI NFV ISG). We will use the tools (technical and otherwise) that become available, regardless of their origin, as long as they help us deliver the right solutions that enable us to better meet the needs of our customer.”
Nico: “I believe that the one thing that SDN caused, particularly at Colt, is change in thinking. Most of us have been around for a few decades and previously we kept building networks based on the same architecture blueprint which we believed was the (only) right one. These changes will both drive and depend upon the evolution of our operating model – which is critical with the changing role of a service provider.”
SDNCentral: What business problems can SDN solve for Colt?
Nico: “The days when carriers and service providers were only required to sell basic connectivity are long gone. Business customers today demand fully managed integrated IT and network solutions. And here’s a real business issue: how do you sell integrated solutions, when one of the components (managed IT) can be automatically provisioned in a matter of seconds/minutes and is more dynamic, while the other component (network) requires manual provisioning, which can typically take days and is much more static. This is indeed the key business problem that we hope SDN will help solve so that we can offer real automation, elasticity, scalability and on-demand services, but not as a one-way function initiated from within the cloud and directing the network A joined-up approach between the network and applications will be critical. And with that, SDN becomes the real enabler to deliver integrated network and IT solutions as well as the underlying capabilities to deliver transactional network services at scale.
A second business benefit Colt is anxious to realise from SDN is rapid innovation. Those in the carrier and service provider space know how long and complex development cycles are when dealing with today’s networks and platforms. Let’s assume that a particular new feature/protocol of interest is already standardised. In a typical multi-vendor scenario, you would have to wait for all your vendors to release the new feature. Before going into the live network, you would have to test all your different hardware and software combinations, across all vendors, and obviously make sure that they are interoperable. Finally, you would need to plan for the deployment across your network, often dealing with what can be painful planned works constraints. At this point I think it is quite obvious that the current network environment is not the ideal incubator to support innovation. One thing that SDN promises is to deliver centralised software-based control of the network infrastructure, shifting to a new architecture where innovation will rely on much faster software development cycles instead of slow hardware ones. But let’s be honest, when it comes to wide-area networks it’s going to be hybrid scenarios, distributed control planes have interesting properties too.”
SDNCentral: Colt is also involved in NFV. What is your view of how NFV and SDN can work together?
Javier: “Colt was one of the contributing organizations to the Network Functions Virtualization white paper back in October 2012 and is an official member of the ETSI NFV ISG group. NFV and SDN are complementary and we see them as different tool sets supporting the overall goal of network programmability.
It’s been a few years since we publicly shared our thoughts and plans around virtualized network functions but we should probably reveal that we have been delivering Internet Access using a Virtual IP CPE since September last year (2012).
While we started with customer premises equipment, NFV targets the virtualization of many of the network functions most of us run today. The end goal is to have virtualised network functions so that they are developed in with software and run on standard high volume servers. Even though SDN techniques could be used to accomplish similar objectives, what NFV is proposing builds on standard IT virtualization technology. There is certainly a potential overlap, and SDN could play a key role in the NFV network component.”
SDNCentral: What are your thoughts about NFV evolution and the speed of uptake? Is it just a paper tiger, or real?
Javier: “The first face-to-face NFV meeting took place a couple of weeks ago and attracted an incredible amount of attention from the whole industry. It is not easy to bring together so many different people and organisations in such a short period of time, and we are already succeeding in defining a formal structure to drive this effort forward. It is important to understand that the objective of this group is not to define new standards but to drive other standards developing organizations and the industry in general to address the technical challenges. After two and a half days of intensive meetings it was agreed that the following four working groups would be created: Architecture of the Virtualization Infrastructure, Management & Orchestration, Software Architecture and finally Reliability & Availability. It was also decided to create two Expert groups (Performance & Portability and Security) to support the Technical Steering Committee.
If there is one message that operators and service providers want to make very clear with the NFV initiative, it is that this is not a talking exercise. It is to deliver real developments in a very aggressive timeframe (no longer than 18 months). Most of the founding operators already have prototypes running in their labs, so there is an early consensus that what is being proposed is feasible. Now, the time has come to deliver the key standard elements through a joint collaboration effort.”
SDNCentral: Some community members have suggested that for service providers, SDN isn’t anything new, and that this concept of the separated control and data plane has been around for some time. What is your view on this? Is SDN just a natural evolution of networking?
Javier: “Separation of the control and data plane is indeed nothing new for service providers. Just to give an example, transport networks have traditionally been run for decades following this paradigm. What is new with SDN is the intent to standardise the different components of this new architecture, especially the communication protocol between those two planes, and the fact that this is now being applied to packet networks. SDN seems the natural evolution towards network programmability, where decoupling control from the data plane allows for network intelligence and applications to be resolved at a centralised software level. But to us SDN can and does mean much more. I think that while this is what sparked the SDN “revolution” it may not be the end goal.”
SDNCentral: In terms of SDN, has Colt deployed any production services using SDN today?
Nico: “Colt is currently executing a project called the “Cloud Centric Network” to evaluate and deliver Colt’s next-generation data center network architecture, the network interconnecting those data centers as well as stronger integration with our cloud platforms. SDN is considered a key enabling technology to deliver this new target architecture. Colt is currently running the evaluation, considering both pure OpenFlow and overlay solutions in this space. The fact that the initial Colt application is in the data center area seems to be pretty much in line with what is happening at other service providers and reflects what is currently available in terms of solutions.
On the NFV front, after delivering Virtual IP CPE for Internet Access services we plan to expand it to IPVPN and more complex scenarios (e.g. hybrid networking). The next step is to fully virtualise the L3 CPE in the cloud.”
SDNCentral: What do you envision for SDN in 2013? What hot areas do you think will emerge beyond datacenter network virtualization?
Nico: “The first commercially available OpenFlow based SDN solutions address network virtualization in the data center, typically using an overlay approach. What makes that attractive is the fact that being an overlay, it works on top of the existing infrastructure. What we expect to see in 2013 is the same concept being extended beyond traditional data center boundaries and into the WAN, but not all wide area networks and possibly only in a subset of the sub-networks or regions/areas. For example, we expect to see SDN being extended to optical transport networks. Indeed we already see discussions happening in the ONF, as well as some optical vendors proposing OpenFlow based overlay solutions for optical transport networks.
Finally, another area of the network which is not in the spotlight today is the access network, but in order to deliver an end-to-end service and technology wrap, the capabilities of the CPE and the customer access line are critical too. This is an area of focus for us (e.g. Modular Carrier Ethernet) and we expect more developments in the industry this year.”
SDNCentral: ‘Thank you, Gentlemen, for your time today and sharing your thoughts around SDN and NFV. It’s been a pleasure!”