Key Technology Analysis of the CR19000 Cluster Router

I. Introduction

Emerging technologies such as 5G, IoT and AR/VR have diversified the types of access terminals and increased the number of terminals exponentially. The boom of 4K and 8K videos, artificial intelligence, cloud computing, big data, blockchain and other emerging information industries has resulted in surging network traffic. The proliferation of terminals and service traffic imposes higher requirements for network bandwidth and latency.

Currently, the development trends of data services provided by domestic telecommunication operators include: the rapid increase of backbone network traffic with decreasing latency; the rapid increase in the traffic and proportion of mobile Internet traffic; the convergence of Internet traffic to IDCs; and a sharp increase in cloud-based interconnected services. These trends also drive the backbone network to cope with the data deluge with ultra-wide loading capacity, higher stability and reliability, on-demand deployment, and dynamic allocation of network resources.

II. A key component for backbone network – cluster router CR19000

Since the release of early enterprise routers with IPRs of China in 1997, the H3C router product line has become the most complete router product line in the industry after over two decades of R&D and technological innovation. In 2007, H3C launched an early self-developed high-end router, becoming one of the few manufacturers in the world with high-end router products and technologies. CR19000 cloud-based cluster routers were unveiled at the 2017 Navigate Summit.

The CR19000 cluster routers provide 100 Gb/s high-density Ethernet interfaces that support 12-chassis clusters and can scale up to 72-chassis clusters, standing out from the peers in the global market. The innovative hardware architectures such as cable tray architecture, OIO/CPO and intelligent fan micro module enable CR19000 to address future network application and development needs at ease.

Cluster technology—Infinite expansion and upgrade

Cluster routers provide the most effective solution for a telecom operator to expand its IP backbone network. Compared with the solution of increasing the capacity of a single device or changing the network architecture, cluster routers can easily meet the needs of rapid business growth, network performance improvement and capacity expansion without interrupting the business or changing the network architecture.

The CR19000 cluster routers support 3+12 standalone clusters and back-to-back clusters, forming the first Tb-level 3+12 cluster system in the industry. It can be expanded to up to 72 chassis in one cluster and enables tens of millions of family users to watch 4K videos online concurrently, standing out from its peers in the global market.

CR19000 supports multiple cluster modes and cluster specifications such as back-to-back, 2+6, and 3+12. In addition, CR19000 supports 12 cascaded chassis currently and up to 72 chassis in the future.

Back-to-back cluster

CR19000-20 back-to-back cluster system

The back-to-back system consists of two line-card chassis, respectively known as the master chassis and the slave chassis.

By default, the line-card chassis with the ID of 1 is the master chassis that is used to centrally manage all devices of the system. The active main processing unit of the master chassis is the active main processing unit of the back-to-back system, and all components (including boards, PSUs and fans) of the master and slave chassis are registered with the active main processing unit.

By default, the line-card chassis with the ID of 2 is the slave chassis with the minimum functions of its main processing. It serves as a management channel and also provides necessary debugging and diagnosis functions.

The active main processing unit of the back-to-back system and the standby main processing unit of the system are on the main chassis. The main chassis and the slave chassis can trigger the active/standby switchover through command lines or events. The functional partitioning of the master and slave chassis improves the reliability and manageability of the whole system.

Multi-chassis cluster

CR19000-20 2+6 cluster system

The N+M cluster is composed of N switching chassis and M line-card chassis.

A switching chassis is an independent hardware device without service interfaces. The switching chassis uses a special matrix card to complete matrix switching. When connecting the switching chassis and the line-card chassis, the control channel and the data channel are distinguished and the two channels are used to perform the cross-chassis connection of the control planes and the cross-chassis switching of data planes in the cluster system, respectively. As a common single-chassis device, the line-card chassis connects to the data channels of the switching chassis by adding switching network boards to the cluster. For cross-chassis traffic switching, the user traffic is divided into variable-size cells on the service board in the line-card chassis, and finally arrives at the destination service line card through the switching network board in the source line-card chassis, the switching network board in the switching chassis, and the switching network board in the destination line-card chassis. The service line card then reconstructs the cells into messages.

OIO/CPO—a dance on hardware

CR19000 has ultra-high forwarding performance, especially in a multi-chassis cluster that requires high performance of the cluster-cascaded fabric module. CR19000's cluster-cascaded networking cards can offer the forwarding capability of over 10 Tb/s, far outperforming the traditional way. To achieve this with the traditional method, we need to provide over 100 Ethernet interfaces at 100 Gb/s on each cluster cascaded networking board, which is a serious challenge for the planning, deployment and maintenance of devices.

CR19000 adopts the innovative OIO/CPO technology that integrates high speed silicon photodiodes into the chip, and achieves a great leap of electrical signals to optical signals. In the area of a single coin (diameter: 25 mm), it can provide the forwarding capability of up to 300 Gb/s, tripling the performance of mainstream routers. Compared with the mainstream cluster-cascaded interfaces, it reduces the power consumption by 60%, reduces the number of inter-chassis optical fiber connections in the cluster by 70%, and greatly simplifies the deployment and maintenance of devices while improving performance.

Cable backplane—Like a world hidden in a grain of rice, small connectors turn things around

CR19000 currently supports the 18-port 100 Gb/s board, which supports smooth upgrade to 8 Tb/s. This requires high scalability of the device. A traditional router uses PCBs, which cannot meet the requirements of high-speed signal transmission and signal crosstalk. The CR19000 routers use Cable technology, which can boot the quality of high-speed signals by seven times and double the overall capacity.

The Cable backplane of CR19000 combines high performance and high scalability. The Cable backplane has thousands of SerDes cables, which impose strict requirements on the connectors of the switching network board and service board on CR19000. CR19000 can transmit Tb-level traffic through its connectors in an area of less than 0.0006 square meters. It needs to connect hundreds of cables, and each cable is connected with the help of a high-power magnifier. This process is the most precise part of CR19000 and can be called the rice carving process on CR19000.

SR/SRv6 capabilities—Adapting to diversified business needs

CR19000 is an intelligent routing platform developed for the cloud era. It provides SR/SRv6 capabilities and helps users evolve smoothly from MPLS to SR/SRv6. Facing the diverse business requirements of the 5G and cloud era, CR19000 helps quick service application, end-to-end deployment and personalized bearing, improving user experience.

SDN-oriented design—Open network architecture for the flexible calling of resources

CR19000 adopts an SDN-oriented architecture, fully supports the SDN technology, and provides a variety of northbound protocol interfaces to connect with external systems. Therefore, devices can interact with SDN controllers in all directions to enable customers to accurately control network resources based on their application requirements, greatly improving the customers' network operating efficiency.

Network device layer:

It is mainly composed of routers to implement service access and high-speed forwarding. It bears services and provides northbound interfaces, through which the routers can be controlled by the control layer and the network and service traffic information is sent to the control layer for network monitoring and management.

Control layer:

SNA system is the core component of the solution architecture to realize the key capabilities, including network automation, intelligent traffic scheduling, and intelligent O&M. It is mainly composed of SNACenter, SeerEngine (controller), and SeerAnalyzer (analyzer).

SNACenter is the orchestration layer in the SNA architecture and is responsible for the orchestration of system management, including the unified interface orchestration of the controller and analyzer, hardware resource management, and interaction between the analysis policy and control policy. SNACenter provides unified northbound external interfaces and supports cross-scenario business orchestration.

SeerEngine (controller) provides the capabilities such as O&M, control, and visualization. It provides southbound interfaces for device control and basic O&M.

SeerAnalyzer (analyzer) provides intelligent analysis capabilities. By using big data, AI and other tools, it collects network-wide information (including network traffic, quality, associated event and alarm information), conducts data digging and mining, presents users with the core value factors of the network, helps users with IT decision-making, and facilitates business transformation. On the other hand, an intelligent network should be capable for flexible and simple intention execution, shield or abstract the network infrastructure as far as possible (such as access link and device form), and virtualize network elements and network services. SeerAnalyzer can provide a simple graphical operation interface and a portal according to the service and network operation status, simplifying user network O&M.

Orchestration layer:

It performs different functions depending on the customer and the business. Different operating scenarios include the scenario integrating a third-party supplier's orchestrator to H3C solutions and the scenario integrating the orchestrator developed by the user to H3C solutions.

Straight-through air duct—efficient heat dissipation, environmental-friendly and energy-saving

The traditional devices adopt Z-type, U-type, L-type and C-type air ducts. The air force decreases at every bend, resulting in a large air speed loss and low heat dissipation efficiency. Hence, the traditional air duct design cannot meet the heat dissipation requirements of high-performance devices.

CR19000 adopts a straight-through air duct and micro-module design to respond to the heat dissipation demand. The air duct adopts the straight-in and straight-out heat dissipation mode, which can minimize the air flow loss and ensure high heat dissipation efficiency, boosting the heat dissipation performance for the key components. CR19000 adopts 33 independent fan units to eliminate the blind areas in the air duct. The speed of each fan unit can be adjusted intelligently according to power consumption of the device, effectively reducing its energy consumption. Compared to traditional devices, CR19000 can reduce the device power consumption by 50%.

III. Successful cases of H3C CR19000

In 2019, H3C cluster router CR19000-20 passed the centralized procurement test of China Mobile, China Unicom and China Telecom, and won the bid for the core router section of China Telecom. After that, it was applied in the core key business scenarios of MAN by China Telecom and was deployed in Guangdong Province and other provinces. Besides, CR19000 has been adopted by Jiangsu Broadcasting Corporation, Guangdong Broadcasting and Television Network Co., Ltd., Zhejiang Watone Cloud Data Technology Co., Ltd., and Shandong Radio and Television. The successful application cases have proven H3C's leading expertise in the IP data communication industry, the carrier-grade reliability of its products, and its capabilities of participating in the construction of IP data communication infrastructure in the highest value sector of operators.

1. Successful case—IP backbone network of Jiangsu Broadcasting Corporation

The IP backbone network of Jiangsu Broadcasting Corporation carries all kinds of services such as interactive television, major customer private network, IP data and broadband services in Jiangsu Province. Core routers of the backbone network interconnected with edge routers of all cities should learn from the whole network routing. Therefore, the core router must provide powerful performance, large table entries, outstanding reliability and availability.

H3C has helped Jiangsu Broadcasting Corporation complete the two-plane construction of the core router for the IP backbone network. Supporting the cluster technology, CR19000 not only provides remarkable performance and reliability for the backbone network, but also provides extensive routing and QoS strategies, ensuring the quality of different services.

Successful case—Guangzhou Wired IP MAN

Guangzhou Wired IP MAN carries the services of 400,000 interactive videos, broadband users and a large number of government and enterprise VPNs of Guangzhou city with high traffic. The core router of the MAN connects with the core router and IDC of the uplink provincial network and connects with the converged router and BRAS devices of dozens of front-end equipment rooms across Guangzhou. It conducts the north-south forwarding of various types of service traffic, and implements logical isolation for various types of services through MPLS VPN.

For the capacity expansion of the MAN, two CR19000-20 routers are used to upgrade the original 10 Gigabit network to 100 Gb/s, meeting the high bandwidth requirements of new services such as 4K/VR/AR and 5G in the future. CR19000 supports SDN and SR/SRv6. Therefore, it can support the MAN and provincial network to realize SDN-centralized intelligent scheduling of services in the entire network in the future, and can support the deployment of SR/SRv6, simplifying and optimizing the current MPLS VPN services.

Successful case—Egress router of Shanghai Yovole Networks Co., Ltd.

Shanghai Yovole Networks Inc., founded in 2001, is an enterprise-class cloud computing operator providing extensive cloud computing data center services for medium- and large-sized customers in China. Egresses of a data center are important for IDC service quality. The IDC egresses of Yovole Networks feature various types of services, high traffic, multiple egresses, interconnection with several operators such as China Telecom, China Unicom, China Mobile, CERNET and GWBNS, and the introduction of network-wide routing. An egress device also needs to be scrubbed and monitored with the help of a traffic scrubbing device, imposing a high requirement on the performance of the egress router.

Currently, Yovole Networks has deployed a pair of H3C CR19000 devices at the egress as the egress routers. Each slot provides the forwarding capability of 1.2 Tb/s, meeting the high performance and scalability requirements of IDCs. It also provides ten-million-level routing table entries to handle scenarios with a heavy routing pressure. It supports extensive routing features and strategies and can redirect traffic using a traffic scrubbing device.

The CR19000 cluster router is widely used in the backbone nodes and MAN core nodes of operators, as well as in the interconnection nodes and core egresses of large data centers of Internet enterprises such as Baidu, government departments such as the e-government networks of Beijing, Nanjing and Zhejiang, and enterprises in healthcare, electric power, education and other industries.