After decades of development, Ethernet technology — which is flexible and simple to deploy — has become the primary local area network technology. As networks continue to scale and expand and more applications emerge, Ethernet is becoming more complex, evolving from a bus topology to mesh, and from a small-scale single plane network to a large-scale hierarchical network.
The industry faces a great challenge. How does it ensure efficient packet forwarding while maintaining network simplicity? Various technologies and standards have been developed to achieve this goal; the Cluster Switch System (CSS) is one of these technologies. CSS virtualizes multiple switches into one logical switch to simplify the network architecture and improve network efficiency.
Conflict Between Simplicity and Efficiency
Inevitably, with the increasing popularity of the Internet, the Ethernet has become the primary protocol of choice because of its simplicity, flexibility and ease of expandability.
However, with the growth of Ethernet usage, many issues arise. The first issue is how to improve forwarding performance without making the network more complicated; with Ethernet, simplicity and efficiency can conflict.
- First, simple but inefficient: Initially, Ethernet was designed to use a bus network topology and carrier sense technology to control packet forwarding. Although this type of network construction is simple, its efficiency is low. Only one host at a time is permitted to send packets.
- Later, Highly efficient but complex: When a network has to scale, Ethernet’s low forwarding performance cannot meet the needs of increasing customer traffic. To address this problem, Ethernet switches have been developed, which have implemented full mesh topology to improve forwarding efficiency and to enhance network reliability.
But, although network performance has improved, a new issue emerges: Network loops, where network traffic cycles needlessly and create extra loading that could bring down a network. To prevent this problem, the Spanning Tree Protocol (STP) was developed in the 1990s. STP prevents network loops by blocking links, but this method degrades network performance. Another protocol, the Multiple Spanning Tree Protocol (MSTP), has been developed to overcome this defect. MSTP ensures high network performance, but is difficult to deploy and maintain. Additionally, when MSTP is deployed, Ethernet network topology is no longer simple.
Consequently, network designers face a dilemma: If the network is simple, its efficiency is low. If the network is highly efficient, it is complex. Is there a technology that can improve network-forwarding performance while keeping network design simple?
CSS: A Simple and Efficient Network Solution
CSS virtualizes multiple switches into a single, high-performing logical switch.
CSS offers the following features:
- Many-to-one virtualization: CSS virtualizes multiple switches into one logical switch that has a unified control plane and provides unified management.
- Unified forwarding plane: CSS uses a unified forwarding plane that shares and synchronizes forwarding information.
- Inter-chassis link aggregation: CSS aggregates links between member switches into one trunk link.
Figure 1: Physical and logical structure of a CSS and a CSS network
Figure 1 shows the physical and logical structure of a CSS and a CSS network. CSS virtualizes multiple switches into one logical device and supports inter-chassis link aggregation. CSS simplifies network topology and greatly improves network performance by offering the following features:
- Simplified operation and maintenance: A CSS functions as one logical switch, simplifying operation and maintenance, thus reducing OPEX.
- High reliability: When one switch in a CSS fails, another switch in the CSS takes over the control and forwarding of packets to prevent services from being impacted by single-point of failure.
- Loop-free network: CSS supports inter-chassis link aggregation to prevent loops. Therefore, the deployment of complicated protocols, such as MSTP, is unnecessary.
- Load balancing: CSS supports Equal-Cost Multiple Path (ECMP) across switches, making full use of network links and bandwidth.
CSS simplifies network architecture and improves forwarding performance while maintaining network functions. CSS has all the functions of physical switches in a cluster, but provides better performance, and has therefore gained wide customer recognition and is becoming the preferred solution for simple and efficient network deployment.
CSS has undergone two developmental stages.
In the early stages, switches — which were interconnected by dedicated cables — were set up as a CSS using dedicated line cards. Because the dedicated line cards did not occupy the slots for service line cards, they did not impact the system’s forwarding performance. Meanwhile, the interfaces on dedicated line cards had a higher transmission rate than the interfaces on service line cards, which improved the bandwidth for CSS interconnection.
In the later stages, physical switches, which are interconnected by standard cables, establish a CSS using service line cards. Because several service interfaces are used for this interconnection, the forwarding capability of the switches is greatly improved. The use of standard service interfaces and standard cables allows switches that are located a great distance from one another to form a cluster, making CSS deployment flexible. Moreover, the interconnected interfaces support link aggregation, which allows for the flexible expansion of interconnection bandwidth.
Huawei Next-Generation Switches Enabling High-Performance CSS
Huawei next-generation CE12800 switches support CSS technology and allow for flexible and efficient CSS deployment in the following ways:
- Flexible combination of models: Different CE12800 models can form a CSS. For example, a CE12812 and a CE12808 can form a CSS.
- Various available interface bandwidths: Each physical port in the CSS interconnection interface can be either a 10 GE, a 40 GE, or a 100 GE interface.
- High-speed interconnection: CSS allows a maximum of 16 interfaces to be bundled into a trunk interface for CSS interconnection, providing 640 Gbit/s of unidirectional forwarding bandwidth.
- Long-distance deployment: CSS uses service line cards for interconnection, and CSS switches are connected by standard optical fibers. These features allow switches that are located far from each other to form a cluster.
CE12800 switches provide superior networking capabilities because they have line cards with high-density interfaces and high forwarding performance. Figures 2 and 3 illustrate two possible configurations on networks with CE12800 switches.
Figure 2: High-density 10 GE access provided by CE12800 CSS
Figure 2 shows an application of CE12800 CSS. Two CE12812s form a CSS and connect to downstream CE6800s through 40 GE interfaces. The CE6800s are configured in stacks. The CE12800s and CE6800s establish a full mesh topology and are interconnected through trunk links. The entire network provides a maximum of 5,805 x 10 GE interfaces for downstream devices.
Figure 3: High-density GE access provided by CE12800 CSS
Figure 3 shows an application of high-density GE access provided by CE12800s and CE5800s. Two CE12812s form a CSS, connecting to downstream CE5800s through 10 GE interfaces. The CE5800s are configured in stacks, and the entire network provides a maximum of 25,344 GE interfaces for downstream devices. Because CE12800 switches offer high levels of forwarding performance, the networking capability of CSS formed by CE12800s is greatly improved, which allows for deployment on networks of different scales.
After years of development, CSS has reached a maturity level that is appreciated by many customers. CSS simplifies network architecture and improves network performance. Because CSS is so simple to deploy, customers are not required to learn new technologies, keeping their operating costs low. CSS will definitely offer wider applications as its development continues.