Building DC-Centric Networks On Demand with Cloud-Network Convergence

 

 

Enterprise Access to the Cloud Has Become a Trend

As cloud computing matures after the first decade, enterprise access to the cloud has become a trend. China's economy is transitioning from internet-driven transformation to digital transformation driven by cloud computing and AI. Cloud computing is accelerating the integration with traditional industries such as finance, education, and transportation. The rigid demand of digital transformation for enterprises and the strategic support from the government have brought huge market space for cloud services. China's cloud computing market is developing at a high speed, and cloud is replacing traditional networks and gradually becoming the center of services.

Enterprise access to the cloud has several phases. As more and more enterprise systems migrate to the cloud, the quality of network connectivity has gradually become the focus of attention. In the face of competition, operators first need to tap into their network advantages. They provide flexible and diverse private lines for access to the cloud as well as secure and reliable multi-cloud connectivity services and introduce SDN to improve existing networks and achieve rapid service provisioning and network optimization. When combining cloud-network orchestration capabilities, operators can also build an overall cloud-network convergence system to provide end-to-end cloud access services for users.

Building a DC-Centric Network 

Cloud computing has become richer in its second decade. In addition to public clouds, private clouds, hybrid clouds and industry clouds are also developing rapidly and cloud services are gradually extending to the edge, bringing new market opportunities. The arrival of a more complex multi-cloud era is driving the transformation and upgrade of traditional networks, and clouds and networks are breaking through each other's boundaries and converging with each other. With the advance of all-cloud concept, a network needs to be built around the cloud and the network architecture is optimized with the concept of cloud computing, so that network resources can be dynamically and flexibly scheduled and allocated according to user needs and a quality network with agility, flexibility and intelligence can be built up.

At present, MAN devices use proprietary hardware, and a business model of optimal bandwidth raises high requirements for device capabilities. The network with tightly coupled software and hardware is not flexible enough, and it is difficult to introduce automation. The rapid innovation model led by internet thinking has brought uncertainty in network service and traffic, and the network architecture is difficult to meet the needs of service growth. Operators need to introduce the design concept of the IT industry. With SDN/NFV as the basic architecture, they can build a simple, automatic, and intelligent cloud network that supports unified transport (Fig. 1). The network can quickly and flexibly adjust resources for service innovation.

The construction of future MAN will be DC-centric and oriented to industries, 5G, HD video and MEC. The new architecture shall make better use of existing network resources and integrate with the cloud to meet the requirements of industry digitization for flexible deployment and ubiquitous access.

Future MAN tends to be a flexible, efficient, and scalable network built with simple-designed devices. Currently, being the core of MAN, CR shall meet the requirement of large-traffic forwarding while satisfying the service need of large routes, which results in high Capex. As traffic grows in MAN, an architecture for the separation of forwarding and service can be introduced to lower the requirements of intermediate forwarding devices. A single router or a large-capacity ASIC device can be used as metro core switching node to forward metro traffic and offload the traffic passing CR. ASIC devices are characterized by large bandwidth, low power consumption, and higher cost effectiveness. The EVPN protocol is introduced to build a Fabric matrix for rapid metro traffic forwarding, storage and computing resource pooling, flexible resource scheduling, and highly reliable resources based on physical distribution. The access layer provides large-bandwidth access through 10G PON and 5G, meeting large-bandwidth needs of future services such as HD video and VR. As the network extends to 5G, it is necessary to plan uniform transport of fixed-mobile and cloud-network services for better use of network resources.

The architecture for the separation of forwarding and service can better improve the efficiency of resource utilization, simplify devices, and allow flexible service deployment, which is the direction of network evolution. The cloud-based service layer is deployed on the physical network to achieve rapid service innovation and flexible expansion capabilities. The forwarding layer is simplified, reduces device requirements, and provides reliable transport and end-to-end automatic deployment to meet differentiated transport requirements. Network transport protocols are also gradually simplified and unified into SR/EVPN, and then evolve to SRv6. The transport protocols based on SR/SRv6 can realize end-to-end small-granularity scheduling, improving network scheduling capability by hundreds of times. SRv6 combines IPv6 and SR to realize the unification of common IP forwarding and tunnel forwarding without using MPLS any longer. This will greatly simplify devices and networks. The programmable characteristics introduced by SRv6 make the network more flexible and easy to implement automation and intelligence. SRv6 has the ability to deploy from the applications, so the network and applications will be deeply integrated, and the network will be more open and provide better support for services.

The key driver for building a DC-centric new MAN is the services in the metropolitan area. At present, most of MAN traffic comes from video and web, most of which is internet content and ends outside the metropolitan area. In the near future, there are three types of services that can be introduced into the metro DC: cloud-based network devices including vBRAS, vCPE and UPF, video services provided by operators and the internet, and local services provided by enterprises. With the development of 5G, the demand for edge computing and industry-oriented AI, especially VR bandwidth will be the main driving force of network re-architecture. VR will bring large-scale gains to the network after 2021. The new MAN can be gradually built according to business needs. When MAN east to west bound traffic accounts for a relatively high proportion, core switching nodes will be added to offload CR traffic and finally implement fixed-mobile converged service transport.

Building a New-Generation O&M System

Changes in network architecture need the support of an O&M system. The focus of improving cloud-network convergence experience is on the optimization of service processes. To deliver cloud-network converged products and services, operators need to build a new-generation O&M system.

Unified cloud-network orchestration is the key to the new-generation O&M system and also the basis for one-point service provisioning and end-to-end service guarantee. Operators should first focus on building the orchestration and collaboration layer, controlling the orchestration layer, and introducing atomic capability and service-based architecture to support rapid service innovation. Due to high complexity of multi-vendor and cross-domain management, it is suggested that controllers should be simplified as far as possible and be deployed in different domains, and the controllers in different domains should coordinate with each other to create connections. Therefore, converged management and control is the future development trend. Considering the compatibility with existing networks and the diversity of equipment functional interfaces, operators and vendors should have a reasonable division of labor, which will help to promote the rapid commercial use of SDN.

Since dynamic and flexible network architecture brings the O&M complexity, operators first need to have the network ability of closed-loop automation. Data collection is the basis of closed-loop automation. Traditional network management methods such as SNMP and FTP are inefficient, collect few types of data, and have poor scalability. To meet real-time collection requirements, it is necessary to introduce telemetry that can achieve millisecond sampling, improve the collection ability up to 100 times, and support real-time data reporting. Based on real-time collection, corresponding strategies can be introduced into equipment, controller and orchestration layers respectively to realize hierarchical closed-loop network automation.

The introduction of intelligence can further improve the cloud-network collaboration efficiency and give better user experience. Operators need to build an intelligent system and introduce hierarchical big data AI capabilities. Building the intelligent system is a gradual process, which can be applied first in network elements and individual function points to enable point-like network element optimization and intelligent fault analysis. Moreover, intelligent abilities of network elements are connected in series to form end-to-end intelligence capabilities to support end-to-end fault location and demarcation, recovery and self-healing. Finally, network intelligence is introduced throughout the network, entering the intelligent phase of the whole network. At the same time, intentional insights are superimposed on the basis of automation to achieve network autonomy.

New MAN oriented to future services and based on a flexible and scalable architecture can cope with the uncertainty of service development, coordinate operations and process transformation, and build network core capabilities. This will better assist operators in operational transformation and help them seize the opportunity of digital transformation.