 | Issue: | Latin America 2015 |
| Article no.: | 8 | |
| Topic: | Network Virtualization: a path to innovation in the telco industry | |
| Author: | Antonio J. Elizondo Armengol | |
| Title: | Head of Innovation & Technology Strategy | |
| Organisation: | Telefónica | |
| PDF size: | 257KB |
About author
Antonio J. Elizondo Armengol, Head of Innovation and Technology Strategy, Telefónica
Since 1998 Antonio is working in Telefónica I+D, the Research and Development company of the Telefónica group. His areas of expertise include IP network architecture, NFV and SDN technologies, traffic analysis, QoS, network planning and dimensioning, routing and network performance, and techno-economic analysis. He is head of the virtualization strategy and technology area within the Telefónica Global CTO unit since 2013, coordinating the efforts of Telefónica in defining and standardizing NFV technologies and leading the team that has developed the NFV reference implementation lab and launched OpenMano as an Open Source project.
Antonio earned a Master’s degree in Telecommunications Engineering at the Universidad Politécnica de Madrid (UPM) in 1998 and a Master’s degree in Economics at the Universidad Nacional de Educación a Distancia (UNED) in 2005
Article abstract
The communications industry is at an inflection point for how networks are designed, deployed and operated. It calls for modernization to improve operational efficiencies that prepare for the explosion of mobile connectivity and data streaming demanded by end customers.
Network virtualization has emerged as a way to cope with these challenges and, currently, providers and operators are racing to find scalable, interoperable solutions that can be more easily deployed and managed across a network. LATAM is a region in which this expansion is expected, making it a target area to apply all these new technologies. Currently, Telefónica is testing a virtual Residential Customer Premise Equipment (vCPE) approach in Brazil where it operates under the Vivo brand.
Full Article
A network has a strong dependence on hardware. This characteristic makes it a very rigid and expensive infrastructure to operate. Making changes to it, to evolve or to enable the launch of new products and services often involves long periods of time.
The inclusion of new functionalities is often delayed by the need for economies of scale to manufacture the specific network hardware, as well as by the long standardization processes between the vendors and operators. What’s more, the very complexity of the network makes it hard to introduce changes from the perspectives of management, configuration and control.
In this situation, it becomes necessary to transform the network into a much more flexible, nimble and moldable infrastructure. Network must be able to adapt in a better way to the quick pace imposed by the evolution of the digital world. To that end, Network Virtualization is revealed as a significant lever for transformation.
An introduction to Network Virtualization
Virtualizing the network consists of separating software from hardware in network equipment so that network functionalities become independent of the physical equipment supporting them. As all functionalities are found in software and not in the physical machines, the same hardware can be used for several purposes depending on the software installed on it.
As functionalities reside exclusively in software, it is possible to package each network function in one or more virtual machines and to decide where to execute them. This is what is known as Network Function Virtualization, or NFV as shown in Figure 1.
Figure 1: Network Virtualization
This new concept is now feasible because a set of advances (related especially with the processor support for I/O operations – direct cache access for I/O transactions, direct PCIe connection to the processor without an I/O hub and large pages support for I/O transactions) in general-purpose servers that has managed to increase by an order of magnitude the performance software can achieve in the handling of data (see Figure 2). These advances have turned commercial off-the-shelf servers into the ideal candidate for Network Virtualization, owing to the significant economies of scale they represent. This is not a trivial process: hardware resources must be allocated with great care in order to achieve high and predictable performance so that virtualised network functions can work at the same speed as traditional network functions.
Figure 2: Outperforming today’s bare metal performance
At the same time, those network functions must be interconnected with each other in a coherent way in order to provide network services. In order to achieve this coherence among virtualized elements that can be deployed, moved or removed dynamically over the hardware infrastructure, the best approach is to manage the connectivity from a centralized control plane. The centralization of the control plane is what is known as Software Defined Networks, or SDN.
Advantages of Network Virtualization
• The hardware infrastructure will be versatile, allowing to install different network nodes on top of it. Thus, when there is a change in demand, instead of purchasing new and specific network nodes, we can allocate the software-based nodes on top of idle infrastructure.
• At the same time, from the point of view of management and investment, a uniform hardware infrastructure is more efficient than managing and investing in heterogeneous infrastructure.
• As this technology is gradually installed, the network will become a more moldable infrastructure, in which it will be possible to deploy virtual functions on demand quickly and efficiently, allowing to innovate in a faster way and reducing the risk of investing in technologies that may not succeed.
• The installation and configuration will be carried out in a more automatic way, making the process more economical, while also making the network more resilient thanks to the possibility of having self-healing procedures in case of failure of the infrastructure.
Making clear the differences between Cloud and Network Virtualization
Although Network Virtualization takes advantage of main cloud technologies already developed, and this has helped to create momentum for NFV, they are not exactly the same and there is still work to be done for NFV. One of main problems has been that vendors have just repositioned their cloud products as a valid base for implementing NFV and they have carried network functions without completely understanding further the implications of NFV.
It is true that there is a clear convergence trend between IT and Network infrastructure but today, hardware selection and configuration is different in both cases.
The main goal in cloud computing focuses on offering commodity computing through different service models (IaaS , PaaS , SaaS ) while in NFV the target is the deployment of E2E network services by composing network functions over a uniform and versatile infrastructure. For doing that, the main design requirement in cloud computing is efficiency afforded by maximizing usage of resources while in NFV it is the effective decoupling while ensuring carrier grade performance.
Workload associated with network functions have requirements for high and predictable performance that call for deterministic and dedicated allocation of resources (avoiding sharing) to be carrier grade . This is different in cloud computing practises where resources are shared and even with oversubscription for maximizing infrastructure utilization. Network Virtualization requires the careful selection of hardware components so a network virtualization environment will be configured differently from a cloud environment, which is just focused on the aggregated availability of computing power, memory and storage.
Regarding physical deployment, cloud computing is normally implemented by having a few, highly centralized and massive datacenters that act as end points and that terminate a relatively moderated amount of traffic, with a lot of resources involved, and with an associated massive cost per datacenter. Meanwhile, applying NFV to the network, which is necessarily distributed, mean to implement a lot of smaller and more distributed points of presence that act as transit points, traversed by a huge amount of traffic, and with very limited resources and associated costs per point of presence.
Finally, regarding the virtualized infrastructure management and orchestration, in cloud computing, the hardware infrastructure is managed just as an aggregated view of resources (CPU , memory, etc.) and virtual switching is used to simplify connectivity management, while in NFV, NUMA view and avoiding bottlenecks in the hypervisor and OS is critical to get proper and predictable performance for NFV, which implies avoiding the usage of virtual switching for the data plane virtualization and relying on an appropriate control of the physical switching. In cloud computing, the orchestration is focused in VMs lifecycle management while in NFV and NFV-O it is required to compose network scenarios (orchestrate coherent capacity allocation – VNFs and their physical connectivity needs) and to manage the network scenarios lifecycle.
In any event, the purchase of both technologies may follow the same process; the physical space can be shared but hardware and software are different. Hardware is based on x86 architecture, but mounted in a different way.
There is still work to be done
Industry has broadly embraced the concept of network virtualization but unfortunately, the industry is lagging behind. The virtual Residential CPE trial carried out in Telefónica Brazil has proved the technical viability of NFV not only in the control plane scenario but in the data plane one. Although it is still necessary to continue working with the industry, we know positively that technology is not a barrier.
The introduction of NFV involves organizational and cultural changes, not only for the operator but also for the industry as a whole, as it modifies the way in which the network is deployed and operated. In addition, the critical path in this kind of transformation projects is imposed by legacy and by the integration with the BSS/OSS (Business Support Systems and Operating Support Systems).
In addition, traditional and virtualized network functions will have to alongside each other during this process so the transformation can be carried out gradually to support the flexibility and scalability that the telecommunications infrastructure of the future requires.
The number of benefits of network virtualization increases in proportion to the growth of the NFV sphere of application. It is not easy to estimate the benefits, if any, of deploying a single virtualized function, but the benefits multiply when more than one use case is virtualized over the same NFV infrastructure. That is why the process of virtualization must be a continuous one that requires investment and that will gradually go through different phases.
NFV can be a transformation vehicle in its own right. Evolving the network towards NFV will improve operations and speed up innovation cycles, not only of the Telcos themselves but also of the manufacturers and start-ups, with all the above resulting in a healthier ecosystem based on the contribution of increased added value.
Conclusion
The network virtualisation race is a marathon not a sprint. And this race will take us through many phases. At the moment, the ecosystem is on the move. Many forums come and go and manufacturers present their approximations. The operators demand an offer that is open, multi-vendor and that adapts to their needs. Meanwhile, new players from the world of IT are circling the world of the network on the strength of their cloud computing experience. Newcomers and start-ups see possibilities of getting in on the act and forming part of a new value proposal. And in the background, virtualisation continues the natural process of any technological maturation cycle.
So far it has been shown that it is technically possible to virtualise network functions although work must continue to ensure that the technology goes on evolving and Telcos finally can capture all their benefits.
