Boosting Network Capacity With New Cluster Routers

Pushing the boundaries of router technology

Back in 1999, when the Internet was transitioning from 64 Kbit/s dial-up connections to 1 Mbit/s ADSL connections, the most advanced router in the industry was able to support a maximum bandwidth of 40 Gbit/s.

At the time, pretty much the entire industry felt that this was plenty for future network expansion. Huawei thought otherwise. We offered to develop a router with a capacity of 80 GB and a bandwidth of 160 Gbit/s, and proposed an architecture that would increase the router capacity linearly through stacking and cascading technologies.

By 2014, former giants in the cluster router domain had diminished due to a lack of forward-looking architecture. By contrast, Huawei's cluster routers supported the smooth expansion of network bandwidth and capacity for people the world over.

Owing to their advanced architecture, this new generation of routers have helped to address ongoing explosive growth in data traffic. The architecture we adopted for our cluster routers soon became mainstream in the industry, which has helped us achieve a roughly 40% share of the global router market.

First steps: supercharging circuit boards

In 1998, Huawei's engineers were in the process of researching and developing our mid-range NE08 routers. We were considering expanding the router's bandwidth by adding more slots. However, this particular router was only able to support a maximum of six slots, all of which would have to share a bandwidth of 1 Gbit/s. This was a serious bottleneck.

To solve this problem, Deng Chaojun, the project manager at that time, thought of replacing the shared bus architecture with a switching fabric, which could increase the number of slots and linearly increase the router's overall bandwidth.

To make this a reality, Mr. Deng's team had to overcome two massive technical hurdles:

• First, the team had to figure out how to transmit high-speed data signals on printed circuit boards (PCBs).
• Second, the team needed high-capacity chips that could support high-speed data switching.

The R&D project, named "Project 1011", marked the beginning of Huawei's journey into developing core routers that would change the industry.

Before this, signals transmitted over shared bus architecture traveled slowly on PCBs, with speeds barely exceeding 33 Mbit/s. This was because existing routers employed single-ended signaling, a design whereby an electrical signal is sent serially over a single wire. While cheaper and easier to design, PCBs using single-ended signaling were subject to noise and interference, which limited the speeds at which data could be accurately transferred.

At the time, our R&D team wondered whether a PCB could carry differential signals (i.e., sending a single signal through a pair of wires) to speed things up, although no one in the industry had ever verified this concept before.

Our team was the first to do so, and the lab results were surprisingly good – the transmission rate reached 1.25 Gbit/s. This "dark horse" technology was first used in another one of Huawei's switch products – the Radium 8750 – which would have otherwise failed owing to the same constraints created by the instability of traditional single-ended signal transmission.

The idea that emerged during research and development of the NE08 brought the Radium 8750 back to life. It later became the industry's first product developed with PCBs that could support lightning-fast data signal transmission.

Support for growing data traffic

By 2000, the development of data communications networks had entered a critical stage, and with an increasing number of Internet users, demands on traffic and bandwidth were growing exponentially.

Imagine half of China's 1.4 billion people suddenly going online. The demand on networks would be immense. Despite this growing understanding, no one in the industry would have thought that future routers would be able to support bandwidths of 100 gigabits, 400 gigabits, or even terabits per second.

At that time, one Cisco router on the market could support a maximum bandwidth of 40 Gbit/s, which was adequate for users at the time. Considering China's large population and broader industry trends, however, Huawei decided to go one step further by developing a high-speed switching chip with a capacity of 80 GB and a bandwidth of 160 Gbit/s, which had never been done before.

At the same time, we began researching a new router architecture that supported stacking and cascading to meet future demands for capacity expansion. Our management team told Project 1011 that it didn't matter whether or not the chip would ever see the light of day, they needed to start working on it.

After many twists and turns, our R&D team succeeded in developing a large-capacity and high-speed data switching chip in 2002, revolutionizing mainstream technology of the time and setting a new milestone in the development of Huawei's cluster routers.

First, the team established a new standard for high-capacity cluster routers – multi-chassis cascading – to overcome traditional constraints in bandwidth and capacity expansion. Our R&D team was acutely aware of what they had accomplished, and had the multi-chassis cascading technology for high-capacity cluster routers patented¹ in both China and the United States.

This technology was an industry first, and it laid a solid foundation for Huawei's core routers to support evolution toward greater bandwidth. As the industry continued to develop, core routers based on traditional single-chassis architecture gradually went by the wayside out of an inability to cope with surges in traffic and growing demands on bandwidth.

Second, Project 1011 established new hardware architecture for large-capacity core routers, replacing mainstream asymmetric uplink-downlink architecture for data flow processing with a higher-performing symmetric uplink-downlink architecture.

Symmetric architecture uses software to improve the quality of service (QoS) of data transmission, ultimately delivering better performance, lower power consumption, and greater scalability. With this innovation, Project 1011 had basically overcome all the major technical barriers to the development of next-generation core routers.

From ordinary to extraordinary

Huawei officially kicked off the development of our core routers in 2003. Three years later, we launched the industry's first back-toback core router – NE5000E 40G. As an industry first, this solution allowed operators to more efficiently expand and upgrade their networks, and also double network capacity by directly connecting two chassis with optical fiber.

In 2008, Huawei released the upgraded 2+8 (2 central chassis and 8 service chassis) core cluster router – the NE5000E 100G. In this new generation of routers, each service chassis would be connected to the central chassis through more than 90 fiber optic cables, which meant there were almost 800 cables for eight service chassis. To reduce the number of cables, we simplified the optical fiber connections between multiple chassis, which sped up installation and maintenance and greatly enhanced system reliability.

Faced with the daunting challenge of connecting China's population and supporting explosive growth in data traffic, Huawei designed a forward-looking product architecture that would support effortless and expedient network evolution. In 2013, the NE5000E core cluster router continued pushing the boundaries of speed as it expanded from 100 Gbit/s to 400 Gbit/s, and then again to 800 Gbit/s and 1.6 Tbit/s in 2019.

In addition to advanced architecture and powerful data processing capabilities, cluster routers on core backbone networks require powerful error correction, fault tolerance, and self-healing capabilities to prevent network-wide service interruption. Huawei's core cluster routers have a built-in, high-reliability defense system that accounts for 80% of the entire system's code.

When China Telecom first explored the use of these cluster routers, it took an entire month just to test the equipment. Service continued flawlessly up until the last fiber optic cable and switching board were removed. In addition, there was not a single packet lost during a 72-hour full-traffic test. No matter how the test was performed, there were no issues with the equipment. Our cluster routers withstood the most stringent tests, surprising the customer with their exceptional stability.

A long road to the present

Huawei began researching core router technologies in 1998 and spent 15 years building our products into the de facto industry standard for performance and reliability. Committed to long-term R&D investment, we continue to pursue innovation in all aspects of router technology – including system architecture, chips, reliability, and integrated systems – and have created many industry firsts:

• First to apply high-speed switching fabric to router design in 1999
• First to launch a back-to-back cluster solution in 2006
• First to launch 400 Gbit/s (2013), 800 Gbit/s (2016), and 1.6 Tbit/s (2019) core cluster routers

It was an iterative, step-by-step process of breaking through one bottleneck after another to maximize the potential of communications networks.