A detailed explanation of tiering and cascade differences in network switches

Cascade Extension

The Cascade extension mode is the most conventional and direct extension, and some of the older networks use hubs as cascaded devices. Because the hubs were already quite expensive, most enterprises could not choose a switch as a cascade device. That's because most team users are connected to a hub from a port level on the hub. In this way, the access ability is greatly improved, but due to some interference and human factors, so that the overall performance is very low, only to meet the needs of multiple ports, it is not enough to consider forwarding and switching functions. Now the cascaded extended mode takes into account the forwarding performance and the port properties of different switches, and through a certain topological structure design, it is easy to realize multi-user access.

Cascade mode is the most ideal way to build a large LAN, it can utilize various topology design technology and redundancy technology to realize hierarchical network structure, such as design redundancy through the topology of double return, and realize the bandwidth expansion of the redundant and up link through link aggregation technology, These technologies are now very mature and widely used in a variety of LAN and metropolitan area networks.

Cascading mode uses common Ethernet ports for hierarchical interconnection, such as 100M FE ports, GE ports, and emerging 10GE ports.

Cascade mode is the mainstream technology in Ethernet extended Port application. It realizes the unified management of the whole network equipment, such as topology management, fault management and so on, through the use of the unified network management platform. The cascade mode also faces the challenge, when the level of the number of layers is more, and there is a large convergence ratio between the layer and the layer, there will be some delay between the edge nodes because of the many exchanges and caches. The solution is to converge the upstream port to reduce the convergence ratio, improve the performance of the upper device or reduce the level of the cascade. In the Cascade mode, in order to ensure the efficiency of the network, it is generally recommended that the number of layers should not exceed four layers. If the network edge node exists through a broadcast Ethernet device such as hub expansion port, because it is a direct working mode, there is no exchange, not into the hierarchy, but it should be noted that the hub work in the CSMA/CD mechanism, The loopback generated by the conflict may result in a much larger network performance impact than switch-level connection.

Cascade mode is a necessary choice to build a structured network, the use of general-purpose cable (fiber), each component can be placed in any location, very conducive to integrated wiring.

Stacking Technology Extensions

Stacking technology is a kind of non-standard technology, which is another kind of technology which is used to extend the port on the Ethernet switch at present. Mixed stacks are not supported among vendors, and stacking patterns are developed for each vendor and do not support topologies. There are two main types of stacking models: Daisy-chain mode and star pattern. The biggest advantage of stacking technology is that it provides simplified local administration and manages a set of switches as an object.

Daisy-chain Stacking

Daisy chain Stacking is a cascade structure based on the hierarchical technology, there is no special requirements on the hardware of the switch, through the relatively high-speed port connection and software support, the final implementation of a multiple-switch cascade structure, through the loop, to some extent to achieve redundancy. However, in terms of exchange efficiency, the sibling-attached pattern is at the same level. Daisy-chained stacks typically have a mode of using a high speed port and two high-speed ports, and the structure of both is shown in Figure two. Using a high speed port (GE) mode, send and receive separate uplink and downlink on the same port, eventually forming a ring structure, the data exchange between any two member switches needs to revolve around the ring for a week, through all switch ports, the efficiency is low, especially in the stacking layer number, Stacking ports can become a serious system bottleneck. Using two high-speed ports to implement daisy chain stacking, you can choose to achieve ring redundancy due to the higher speed ports. Daisy chain Stacking mode and the Cascade mode, there is no topology management, generally can not be distributed layout, suitable for high-density port requirements of a single node mechanism, can be used at the edge of the network.

Because of the need to eliminate the radio storms caused by loops, the daisy chain structure under normal circumstances, at any point in the loop, one of the switches from the switch to the main switch can only be done through a high-speed port (that is, a high-speed port can not share the switch on the uplink data pressure), the need for all upstream switches to exchange.

Daisy chain stacking is a kind of simplified stacking technology, primarily an extended port technology that provides centralized management, the forwarding efficiency between multiple switches is not elevated (efficiency in one-port mode is much lower than that of cascaded mode), requiring more high-speed ports for hardware and software to achieve up link redundancy. The number of layers in a daisy-chained stack should not exceed four layers, requiring that all stacked group members be positioned close enough (typically on the same rack).

Star stacking technology is an advanced tiered technology that requires a single or integrated high-speed switching center (stacking center) for a switch, with all tiered hosts going up to the unified stacking Center via a dedicated (or general-purpose) high speed stack port. The stacking center is generally a hardware switching unit based on a special ASIC, in which the bandwidth is generally between 10-32g and the ASIC exchange capacity of the stack is limited by the number of layers.

Star-Stacked

Star stacking technology enables all tiered group member switches to reach the stacking center matrix to a level the forwarding efficiency of any two-end nodes needs to be exchanged only three times, and the communication structure of the edge nodes of the first-level continuous mode is the same, therefore, compared with the daisy-chain structure, It can significantly improve the data forwarding rate between stacked members, while providing a unified management model, a group of switches in the network management, can be a single node to appear.

Star stacking mode is suitable for single node LAN with high efficiency and high density port, star stacking mode overcomes the Gao Shiyan effect of daisy chain stacking mode multi-level forwarding, but it needs to provide high bandwidth matrix, high cost, and matrix interface is generally not universal, Tiered ports for either a tiered hub or a member switch cannot be used to connect to other network devices. Using a highly reliable, high-performance matrix chip is the key to star stacking. The average tiered cable bandwidth is 2g-2.5g (bidirectional), slightly higher than general GE. The higher part is typically used only for member management, so the effective data bandwidth is basically similar to that of GE. However, due to the special bus technology, the length of the cable can not be more than 2m, so, in star stacking mode, all the switches need to be confined within a rack.

Visible, the traditional stacking technology is a centralized management of the port extension technology, can not provide topology management, no international standards, and poor compatibility. However, in a single node LAN that requires a large number of ports, star Tiering can provide better forwarding performance and easy management features. Cascade is the foundation of the network, you can flexibly use a variety of topologies, redundancy technology, in too many levels, the need for careful design. For networks with few levels, the Cascade mode can provide optimal performance. For example, in a dot that needs to be expanded to a twice-fold port, using star stack edges needs to be exchanged three times, cascading mode and daisy chain stacking need to be exchanged two times, star stacking mode requires more investment, daisy-chain stacking mode needs to occupy more high-speed ports, common-level connection is the most economical and efficient way to build. In addition, it is possible to take advantage of previously existing switching devices without the need for repeated investment, but the two devices need to be managed separately.

Traditional stacking technology applications are often limited by geographical location, often need to be placed in the same rack, in the high density port application, will cause wiring difficulties. So the major manufacturers are actively seeking to support distributed stacking technology. At present, Huawei company Quidway S Series Ethernet switch products, Cisco series Ethernet Exchange products are providing cluster management mode. Quidway S Series Ethernet switches adopt Huawei's unified VRP operating system and a unified Imanager network management system. The network management system supports the Chinese interface, adopts standard protocol and open technology, and is fully compatible with the mainstream network management platform. Quidway S Series Ethernet switch is under the guidance of Huawei's two-layer switching full speed, three-layer switching line speed, service Exchange line speed and QoS service All-speed "four All-speed" design thought, under the direction of the product development, the system design, Expanding capabilities and providing rich business features to meet the needs of broadband metropolitan area networks and corporate networks can provide customers with more efficient, secure, and scalable customer solutions.

Taking Huawei Products (HGMP) as an example, through the support of cluster management mode, the centralized configuration and management can be realized in the network built by using the Quidway series switch through the common-level connection mode, a LAN can join as a group, for the network management system, a group can represent the same equipment, Managed with an IP address, which is equivalent to even better than the management effect of a previously stacked group. However, as a central manifestation of universality, the group membership switch enables topology design and distributed placement of members within a team, and a tiered port can be either a general-purpose port supported by a device or a convergence of ports, enabling users to gain flexibility in the ability to control the tiered bandwidth of switched networks to achieve higher flexibility requirements.

For different environments, the effect of choosing a different port expansion mode is inconsistent. In the current situation, the common cascading mode is the main application of the hierarchical network, the star stacking mode is a simple management mode to provide a single node port extension, and the distributed stack implemented through cluster management will be the primary way of the next generation stacking.