Foreword applications are always the motivation for technology development. In the current increasingly complex business application, the technical architecture of network devices becomes a key factor applicable to the scope of equipment. Different application environments require different network devices. For example, NGN applications require a SoftSwitch technology architecture device; such as longitudinal networks, you need to pay attention to routers; park network focus is a switch, etc., which means that a technological architecture is inevitable to address some kind of application.
So what kind of architecture is the core switch takes to a current network of netware?
Evolution of the exchange architecture
With the increase in Internet users and the expansion of bandwidth, the structure of the switch is constantly developing. From the launch time, the exchange architecture mainly experiences two stages of bus type and crossbar. However, due to the development of Ethernet technology, the switches of these two architectures are currently active in the market.
1, bus type switching architecture
The bus structure-based switches are generally divided into two categories: shared bus and shared memory bus.
The first Ethernet exchange is to build on the shared bus. The exchange capacity of the shared bus can be limited, on the one hand, because the shared bus inevitably avoids internal conflicts; the load effect of the shared bus is relatively difficult to make the high-speed bus design. With the exclusive of "exclusive bandwidth", this shared bus has so far evolved into shared memory structure.
The switches of the shared memory structure use a large number of high-speed RAM to store input data, while relying on the central switching engine to provide a high-performance connection of the full port, and check each input package by the core engine to determine the route. Such switches are easier to implement, but the memory operation generates delay when the exchange capacity is extended to a certain extent, and in this design, the redundant exchange engine is relatively complicated due to the problem of bus interconnection, so this switch is If you provide a double engine, you have to be very stable and difficult. So we can see that the core switch launched in the early market is often a single engine, especially with the increase of the switch port, because the memory capacity is large, the speed is faster, the price of the central memory becomes very high. The switching engine will become a bottleneck that performs performance.
2, Crossbar Shared Memory Architecture
With the exchange capacity of the core switch from dozens of Gbps to hundreds of GBPs today, a swap model called Crossbar has gradually become the first choice for the core switch.
Crossbar (ie crosspoint) is called a cross-switch matrix or a vertical and cross-switch matrix. It makes it possible to make up for some shortcomings of shared memory mode.
Figure 1 CROSSBAR structure diagram
First, Crossbar is relatively simple. The line card in the shared swap architecture is simplified to the physical connection of the switching structure into a point-to-point connection, which is more convenient to achieve, making it easier to ensure the stability of large-capacity switches;
Second, Crossbar has no blocking. A Crossbar is shown in Figure 1, as long as multiple intersections (CrossPoint), multiple different ports can be transmitted simultaneously. In this sense, we believe that all Crossbar is free from inside because it can support all ports simultaneously switched data.
In addition, due to its simple implementation principle and unlafered switching structure, it can operate at very high rates. Semiconductor vendors have now used traditional CMOS technology to make a point-to-point serial transceiver chip with 10Gbit / s or more rates.
Basically 2000 core switches basically chose the ASIC chip of the Crossbar structure as the core, but due to the many factors such as Crossbar chip, the core switching equipment almost chose the shared memory mode to design the service board. Thereby, the cost of reducing the whole machine is therefore, "Crossbar Shared Memory" has become a relatively universal core switching architecture. But in this structure, there will still be a crossbar interconnection problem of the traffic board bus and the switching mesh. Since the data on the service board bus is a standard Ethernet frame, the general crossbar uses a modal switch to reflect the efficiency and performance of Crossbar. Therefore, the structure of the shared bus used on the service board affects the efficiency of Crossbar to a certain extent, and the performance of the whole machine is fully limited to the performance of the exchange of netboard Crossbar. 3, distributed Crossbar architecture
As the co-exchange capacity of the core switch has developed to hundreds of GBPS, the distributed Crossbar architecture is well solved in the new application when the multi-gigabond interface is supported and the scale is applied to the core of the metropolitan area and the center network. The high-performance and flexible challenges faced by the core switches under the environment.
That is, the architecture of Crossbar exchange chips is also used in each service board in addition to the Crossbar architecture in addition to the switched sheet. The exchange chip on the service board can solve the problem of local exchange, and the Crossbar chip between the service board exchange chip and the switching mesh is solved the service data of the service board to improve the exchange efficiency. And make the data type of the service board and the communication of the switching mesh are two planes, which means that there is a very rich service board, such as the business integration of the firewall, IDS system, router, content exchange, IPv6 and so on. On the core switching platform, it greatly improves the business expansion capability of the core switch. At the same time, this Crossbar has a corresponding high-speed interface to two main control boards or switched mesh, which greatly improves the speed of dual master host switching.
In distributed Crossbar design, the CPU also uses a distributed design. The main CPU on the equipment main control board is responsible for the control scheduling, routing table learning and delivery; the service board is mainly responsible for the local surfaction table, the service board status maintenance work. This achieves distributed routing calculations and distributed routing table queries, which greatly alleviate the pressure on the main control board, improves the transfer efficiency of the switch, which is also an important reason for the service board local forwarding to improve efficiency. This distributed Crossbar, the design concept of distributed exchange is the development direction of core network equipment design, ensuring that the current network core can support future massive data exchange and flexible multi-business support needs.
The switch architecture has completed the evolution process from "shared bus" to "Crossbar Shared Memory" to "All Distributed Crossbar", will continue to develop in the future, and the harbor network is one of the mainstream switch suppliers in the industry. It will definitely pay attention to the latest technology, leading the trend of exchange technology, providing users with a more perfect switching equipment.
