Describes the evolution of core switch architecture in detail

Applications are always the driving force behind technological development. In the face of increasingly complex business applications, the technical architecture of network devices has become a key factor in the scope of application of devices. Different application environments require different network devices. For example, the application of NGN needs devices based on the SoftSwitch technical architecture, for example, the vertical network needs to focus on routers, and the campus network focuses on switches, the emergence of a technical architecture is bound to serve a certain application.

So what kind of architecture does the core switch need under the current network convergence trend?

Evolution of exchange architecture

With the increase of Internet users and the expansion of bandwidth, the structure of the switch is also constantly evolving. From the launch time, the switching architecture has mainly gone through two phases: Bus and CrossBar. However, due to the increasing development of Ethernet technology, the switches of these two architectures are currently active in the market.

Bus-type switching architecture　

Vswitches Based on the bus structure are generally divided into two categories: shared bus and shared memory bus.

A vswitch with a shared memory structure uses a large amount of high-speed RAM to store input data, and relies on the central switching engine to provide high-performance connections across ports. The core engine checks each input packet to determine the route. This type of switch is easy to implement, but the memory operation will delay when the switching capacity is extended to a certain extent. In addition, in this design, redundant switching is added due to the problem of bus interconnection, the engine is relatively complex. Therefore, it is relatively difficult for such vswitches to be very stable if they provide Dual Engines. Therefore, we can see that the core switches launched in the market are often single engines, especially with the increase of switch ports, because the memory capacity is larger and the speed is faster, the cost of central memory becomes very high. The exchange engine will become a bottleneck in performance implementation.

CrossBar + shared memory architecture　

CrossBar is called CrossPoint. It can make up for some shortcomings in the shared memory mode.

First, CrossBar implementation is relatively simple. The physical connection from the shared switching architecture to the switching structure is simplified to a point-to-point connection, which makes it easier to achieve and guarantees the stability of large-capacity switches;

Second, CrossBar is not blocked. As long as multiple cross-node crosspoint is closed at the same time), multiple different ports can transmit data at the same time. In this sense, we think that all CrossBar is non-blocking internally, because it can support data exchange at the same speed on all ports.

In addition, because of its simple implementation principle and non-blocking switching structure, it can run at a very high rate, semiconductor manufacturers can now use traditional CMOS technology to create point-to-point serial transceiver chips with a speed above 10 Gbit/s.

However, this structure still has the CrossBar connection problem between the business board bus and the switching network board. Because the data on the Business Board bus is standard Ethernet frames, generally CrossBar uses the cell exchange mode to reflect the CrossBar efficiency and performance. Therefore, the architecture of the shared bus used on the Business Board affects the CrossBar efficiency to a certain extent. The performance of the entire machine is completely limited by the performance of the CrossBar on the Switching Network Board.

Distributed CrossBar Architecture

The switch capacity of the core switch has now grown to several hundred Gbps. It also supports multiple 10g interfaces and is applied to the backbone of the man and the core of the campus network. The distributed CrossBar architecture solves the challenges of high performance and flexibility faced by core switches in the new application environment.

That is to say, in addition to the CrossBar architecture, the switching network Board also uses the CrossBar + switching chip architecture on each business board. Adding an exchange chip to the Business Board can effectively solve the problem of Local switching, the CrossBar chip between the business board switching chip and the switching network board solves the problem of cell the business data of the Business Board, thus improving the switching efficiency, in addition, the data type of the Business Board and the cell of the switching network board are two planes, that is, there can be a very rich business board, for example, you can integrate services of the firewall, IDS system, router, content exchange, IPv6, and so on into the core exchange platform, which greatly improves the service expansion capability of the core switch.

At the same time, this CrossBar has corresponding high-speed interfaces, which are connected to two main control boards or switching network boards respectively, thus greatly improving the master-slave switchover speed of the dual master.

BigHammer is a network convergence Service　

At present, the development of the network shows a clear convergence trend, which includes business convergence, technology convergence, and network convergence. The BigHammer6800 core switch of the WAN network adopts the distributed CrossBar and distributed switching architecture to solve problems such as large-capacity switching and multi-service provision, therefore, it can help users' networks better achieve these convergence.

For the integration of data, voice, and video services, the CrossBar chip of the Business Board can simplify the large data streams of voice and video into a fixed-length switching mode, improving the switching efficiency. This is one of the reasons why BigHammer6800 can achieve "10 Gigabit wire speed switching.

The WAN network solution is also trustworthy in terms of integrated exchange routing and IPv4/v6 network convergence. BigHammer devices not only support a wide range of routing protocols, but also use MPLS technology to solve the biggest challenge facing the network industry-the transition from an IPv4 network to an IPv6 network. MPLS accelerates Data Packet Search and forwarding, and provides a more effective tunneling mechanism. Ipv6 integration on MPLS networks is currently recognized as the most reasonable IPv4 to IPv6 transition solution. MPLS is implemented through an NP-based Business Board with high performance and scalability.

The switch architecture has completed the evolution from "shared bus" to "CrossBar + Shared Memory" to "fully distributed CrossBar", and will continue to develop in the future.

1. Architecture Evolution of multi-layer exchange routers a>
2. TMX 880 MPLS core switch
3. Detailed process of evolution of optical network infrastructure