The development of Cisco switch hardware architecture

With the increase of Internet users and the expansion of bandwidth, the structure of the switch is also constantly developing, from the time of introduction, the Exchange architecture has experienced the bus-type and crossbar two stages. But since the development of Ethernet technology is thousands of miles, so the two architectures of the switch are currently active in the market. 1, the total line-type switching architecture based on the bus structure of the switch is generally divided into shared bus and shared memory bus two major categories. The first Ethernet exchange is built on the basis of a shared bus. The shared bus structure can provide limited exchange capacity, on the one hand because the shared bus is unavoidable internal conflict, on the other hand, the load effect of shared bus makes the design of high-speed bus more difficult. As users crave for "exclusive bandwidth", the structure of this shared bus quickly develops into a shared memory structure.

A shared-memory-structured switch uses a large amount of high speed RAM to store input data while relying on the central switching engine to provide a full port high-performance connection, with the core engine checking each input packet to determine the route. This kind of switch design is relatively easy to implement, but when the exchange capacity extended to a certain extent, memory operation will cause delays, in this design because the problem of bus interconnection increased redundant switching engine is relatively complex, so this switch if it provides dual-engine is very stable relatively difficult. So we can see that the early market in the network core switch is often a single engine, especially with the increase in the switch port, due to the need for more memory capacity, faster, the price of central memory has become very high. Switching engines can be a bottleneck for performance implementations. 2. crossbar+ shared memory architecture with the exchange capacity of network core switches from dozens of Gbps to today's hundreds of Gbps, a switching mode called crossbar is becoming the first choice for network core switches. Crossbar (that is, crosspoint) is called a crossover-switch matrix or a crossbar exchange matrix. It can make up for some deficiencies in the shared memory mode.

First, the crossbar implementation is relatively simple. The physical connection between the line card and the switching structure in the shared Exchange architecture is simplified to point-to-point connection, so it is more convenient to realize the stability of the large capacity switch.

More Wonderful content: http://www.bianceng.cnhttp://www.bianceng.cn/Network/jhjs/

Second, there is no obstruction inside the crossbar. A crossbar diagram, as shown in Figure 1, allows multiple different ports to transfer data at the same time, as long as multiple cross nodes (CrossPoint) are closed simultaneously. In this sense, we believe that all crossbar are internally non-blocking because it can support all ports at the same time exchanging data at the same speed. In addition, its simple implementation principle and non-blocking switching structure allow it to run at very high speeds. Semiconductor manufacturers have now been able to use traditional CMOS technology to create a 10gbit/s over the rate of point-to-point serial transceiver chip.

Basically the network core switch that appears after 2000 basically chooses the ASIC chip of crossbar structure as the core, but because of the cost of crossbar chip and so on many factors, at this time the core Exchange device almost all chooses the shared memory way to design the business board, thus reduces the whole machine cost therefore , "crossbar+ shared Memory" has become a more common Core exchange architecture. However, there is still a problem of crossbar interconnection between the Business board bus and the switching network board under this structure. Because the data on the Business Board bus is the standard Ethernet frame, the general crossbar adopts the mode of the exchange of the crossbar to embody the efficiency and performance of the system. Therefore, the structure of the shared bus used on the business board affects the efficiency of crossbar to a certain extent, and the performance of the whole machine is limited by the crossbar of the switching network board. 3, distributed crossbar architecture with the exchange capacity of the network core switch developed to hundreds of Gbps, while supporting multiple gigabit interface and the size of the backbone of the metropolitan Area Network and network Core, The Distributed crossbar architecture solves the challenge of high performance and flexibility faced by network core switches in the new application environment.

In other words, in addition to the switching network board using the crossbar architecture, in each business board also adopted the crossbar+ switching chip architecture. Switching chips on the business board can be a good solution to the problem of local exchange, the crossbar chip between the Business Board switching chip and the switching network board solves the business data of the business board and improves the exchange efficiency, and makes the data type of the business board and the Exchange Network Board's signal become two planes. That is, you can have a very rich business boards, such as firewalls, IDS systems, routers, content Exchange, IPV6, and other types of business integration into the core Exchange platform, thus greatly improving the network core switch business expansion capabilities. At the same time, the crossbar has the corresponding high-speed interface to connect to two main control board or switch network board, which greatly improves the speed of the dual master standby switch.

In the distributed crossbar design, the CPU also uses the distributed design. The main CPU on the main control Board of the equipment is responsible for the whole machine control scheduling, routing table learning and distribution; the business board is mainly responsible for local look-up table and Business Board state maintenance work. This realizes the distributed routing computation and the Distributed routing Table query, greatly alleviates the pressure of the main control board and improves the forwarding efficiency of the switch, which is also the important reason for the local forwarding of the business Board to improve the efficiency. The design concept of distributed crossbar and distributed exchange is the development direction of core network equipment design, which ensures that the current network core can support the demand of massive data exchange and flexible multi service support in the future.