China Telecom faces four major challenges in network reconstruction of SDN/NFV practice

On July 11, 2016, China Telecom released the "CTNet2025 Network Architecture White Paper", which uses SDN/NFV, cloud computing and other key technologies as support, and the infrastructure layer, network function layer, and collaborative orchestration layer as the three major sections of the architecture. It is committed to achieving 80% network function virtualization, on-demand network scale provision, and building a new generation of operating systems by 2025.

Zhao Huiling, executive director of the China Institute of Communications and director of China Telecom's cloud computing center, pointed out that network architecture transformation is actually the deep integration of traditional telecommunications networks, Internet technologies and IT technologies, making network architecture simpler and more open, network operations more intensive and automated, network deployment faster and at a lower cost, and network services on-demand and elastic. The maturity of SDN/NVF in different fields is different. SDN is first used in data centers and cloud resource pools for commercial use, while NFV is first deployed in VoLTE IMS.

[[178919]]

SDN/NFV is crucial in network reconstruction

Chen Yunqing, deputy director of China Telecom Beijing Research Institute, analyzed that at present, the IP network architecture is based on data centerization and is evolving in three levels: SDN, NFV and cloud. In the cloud, operators need to build a unified telecommunications management platform to achieve interoperability with the lower-layer network and achieve smooth docking with the upper-layer traditional BSS and OSS systems.

The second layer is the SDN controller layer, which must be managed in different domains (metropolitan PoP controller, DCI controller, WAN controller). During the advancement of the entire controller layer, it is necessary to maintain coordination with the orchestration layer.

The third layer is the NFV layer. NFV needs to find an entry point for virtualization based on the original network functions, and finally build an end-to-end service, from the user end through the edge DC layer, and then into the backbone network, traversing the entire IP metropolitan area network backbone network, to achieve a flexible customization process.

With the rise of industry cloud, users need a one-stop, customized, and fast-response end-to-end elastic IP network. This step needs to be achieved with the help of SDN/NFV and cloud technology. Enterprises will become the protagonists, and user needs will force operators to adapt to their changes. The agility of carrying suitable for cloud-network collaboration also requires the combination of SDN and NFV technologies.

In addition, the SD-WAN protocol in the cloud network adaptation protocol is crucial. In the wide area IP network, the centralized control architecture of SDN is used to create dynamic paths that are independent of the underlying network, and to achieve flexible service connection, end-to-end status visibility, and flexible path selection functions. In addition, in the networking process, SDN and Overlay technology can be used to achieve concentration, decouple resources from physical locations, solve the limitations of traditional networks, turn fixed lines into flexible lines, make resources more flexible, and provide differentiated services for network architecture.

SDN/NFV specific deployment faces many challenges

However, the current SDN/NFV development and deployment also faces many problems. Zhao Huiling pointed out that from the perspective of industrial development, SDN requires network intelligence and customization, but the current SDN solution is still mainly based on existing equipment. The mainstream equipment SDN north-south interface support capability is still insufficient, and there are interoperability issues across vendors. Therefore, many current solutions use orchestrators to solve interoperability issues. China Telecom has recently released the first version of Open-O, hoping to further promote this aspect.

There are four major issues in the specific deployment of NFV. The first is the decoupling issue, which involves hardware decoupling, software and hardware decoupling, and three-layer decoupling. Three-layer decoupling is the guarantee for sharing network resources, opening up network capabilities, realizing automated deployment, and network elasticity, and is in line with the development direction of NFV and industry trends. In principle, the three-layer decoupling mode is the first choice for deploying vIMS. When the conditions for three-layer decoupling are not met, the hardware decoupling mode is the second choice. However, three-layer deployment requires a coordinated monitoring system.

The second is the deployment issue, that is, the strategy of virtual network elements, virtualizing network elements according to the priority of the control plane, user data plane, and media forwarding plane. Zhao Huiling suggested starting with the control plane, and then promoting it to the media plane after the technology of media forwarding capabilities matures and is economically feasible. Network elements that communicate with traditional networks are not considered virtualized.

The third is the management orchestrator MANO. After the network is clouded to DC, it involves the operator's entire DC new architecture construction, resource pooling, and especially the deployment of network resource pools. These are all challenges faced by NFV virtualization.

Finally, in terms of operations, operators currently need to solve the problem of orchestration of network resources and business resources, and network resources involve the evolution of the entire OSS and whether the resources of the entire network can be quickly configured. At the business level, operators need to abstract business attributes and achieve goals such as end-to-end cross-domain orchestration.