The system exchange capacity and processing power of high-speed routers are an important embodiment of their different router capabilities. The backplane switching capability of the high-speed router should reach 40Gbps, and the system must support the existing interface card and universal components even if the OC-192 / STM-64 interface is temporarily supplied. In terms of device processing capabilities, when the system is fully loaded, all interfaces should be able to process short packages, such as 40 bytes, 64 bytes, and at the same time, the exchange matrix of high-speed routers should be able to process all lines without blocking. The exchange of interfaces is independent of the type of traffic.
High-speed router overview
Router technology has developed to the development of 3 generations, and continues to evolve.
The first generation router is forwarded by the software set, and all IP packages must be forwarded by the center CPU, which is relatively low, and the forwarding capacity is about tens of thousands of packages.
The second generation router is software-based distributed forwarding, and each interface board has a CPU. The routing table generated by the main control panel is issued to each interface board to form a transfer table, each interface board. The forwarding work is performed independently according to the transfer form, and the forwarding capacity exceeds 1 million packets per second. The technical point of the second generation router is the refresh and synchronization technique of each interface board.
Due to the growth of CPUs, the traffic of the Internet is doubled every 6 months, so the router based on the software forward is an bottleneck in the network. In the late 1990s, IP businesses have exploded, and router technology has also achieved leap to third generations, and more than one degree of interface is expected to become a data network main technology. ATM. The high-speed interface of the ATM switch has only reached 2.5Gbit / s, and the highest port rate of the high-speed router has reached 10Gbit / s. At the same time, since the IP technology's QoS technology is constantly evolving, it is not an introduction of MPLS technology, QoS issues are gradually solving in the IP field, and the high-speed router re-occupies the core location of the IP network.
The third generation router is based on the forwarding of the IP packet, and the forwarding engine can be an ASIC (dedicated integrated circuit), or it is a network processor specifically designed for IP forwarding. Representative products include Huawei's NetEngine 50 high-speed router and Cisco's 12000 series routers. The speed of forwarding the packet has reached tens of millions of packs per second, and it is possible to make full use of a large amount of bandwidth provided by the transfer technology.
Although the basis of the package transfer technology has changed, the task that the router needs to be completed is the same: check the entry packet, compare its target address and the project in the routing table, then from The correct port is sent. In this process, the data package also accepts some additional processing tasks, such as QoS decisions, traffic engineering, traffic statistics, etc.
High-speed router composition
High speed router is usually composed of a main control board, a switchboard, a line interface board, which is connected to a high-speed backplane, inserted a chassis with a power, fan It makes up a complete router.
(1) Main control board: The main control board is the control center of the router, the CPU and the memory are in the main control board. The main control board is responsible for the governance and control of the entire router, and the IP routing protocol runs on the main control board. The main control board directly receives instructions from the network management center and issues to the interface board execution instructions, and each interface board transmits the running status and statistics to the main control board, and the necessary processing is performed by the main control board. NMS. The static route configured by the network governance and the dynamic routing generated by running the routing protocol is governed by the main control board, and the interface board is issued to each interface board can independently perform the forwarding of packets.
The role of the main control board is lifted, once it fails, the entire router will not work properly. For the core network device of the telecommunications network, the required availability is 99.999%, that is, 1 year of downtime cannot exceed 5 min. Therefore, the main control board is usually equipped with two, which is generally operated in the manner. The mother-standby board periodically exchanges handshake signals, once the spare board does not receive the handshake signal of the home panel, the switching process is started, and the playback process is replaced.
(2) Swap panel: The high-speed router is very large, and the way the early router's backplane-based shared bus delivery data cannot meet the needs of high-speed data transfer. First, the shared bus cannot avoid internal conflicts; second, the load effect of the shared bus makes the high-speed bus design difficulty. The introduction of the switching structure gradually overcomes the above disadvantages of the shared bus.
Crossbar switched network can be imaged as a switching network of a spending telephone switch, connecting the input port and the output port that needs to be communicated by point-to-point connections. The Crossbar structure can support high bandwidth. There are two mains: First, the line card to the switching structure is simplified to point to point connections, which allows the connection to run at very high rates. Semiconductor vendors have been able to make 4 ~ 10Gbit / s point-to-peer serial transceiver chips with traditional CMOS technology. The second reason is that its structure can support multiple connections while transmitting data at maximum rate, that is, multiple input / output ports can communicate with this switch network. This greatly increases throughput throughout the system. Simply closing multiple intersections, multiple different ports can transmit data at the same time. From this sense, we call all Crossbar is free from inside because it can support all ports while transmitting (or exchange) data at maximum rate.
The packet can be in the form of a fixed length unit (via the fixed length division of the packet), or may not be divided directly to conversion. The general high-performance Crossbar exchange structure uses a fixed length exchange, and it is used to divide it into a fixed length Cell before the packet enters CrossBar. These Cells are organized into the original growth bag (Packet) in accordance with the exchange structure. . The fixed length exchange method is more conducive to the control of the switch, the length of the group, and it is easy to determine the time of transmission and the departure. At the end of the time slot, the schedule checks the packet waiting for the transfer, determines which output is connected to which output is connected to, avoiding the idleness of the output or input, holding the high efficiency of the switch. Moreover, from the perspective of hardware design, the processing of fixed length packets is simpler and fast than processing different lengths. At the same time, fixed-length exchange can avoid the large length of the length of service flow to occupy the exchange network, affect the exchange of high priority services and real-time services.
Because the fault of the swap mesh can also cause the machine's paralysis, there is usually a mother-standby plate. In addition, the large-capacity switched network is typically multi-plane, consisting of a plurality of switched mesh, so that the host is usually N: 1.
(3) Line interface board: Line interface board provides a variety of line interfaces, common 10M, 100M, 1 000m, 10Gbit / s Ethernet port, 155m, 622m, 2.5g, 10Gbit / s POS interface, 155m, 622m, 2.5Gbit / s ATM interface, etc. The line interface board submits an IP packet from different physical layer and data layer information to a dedicated ASIC or network processor for processing. This process is no longer limited to simply forward the IP packet to the destination port.
High-speed router network application form
High-speed router can be used in Internet backbone network, high speed Internet access, and enterprise network. For the Internet, IP Data Business Designed backbone network, the I-POS based on the router-based router structure can be used.
(1) ISP Application
In the Internet Service Provider (ISP) application environment, Cisco 12000 resides in the core of the network, up to 44 OC-3 / STM-1 IPOS light link, receives the edge router Cisco 7500 Series platform for convergence services through the light link. The connection rate between the core router and the Internet backbone can be extended from OC-12 / STM-4 to OC-48 / STM-16. The use of IPOS will increase performance and link utilization.
(2) Works of communication company
When using a high-speed router with a communication company with SONET / SDH infrastructure, a business collection to the core router Cisco 12000 by the edge router of Cisco . Since Internet backbone is often far from the end user, the core router Cisco 12000 can be pulled near the end user through the communication company's existing SONET / SDH infrastructure (including SONET / SDH access ring and local loop). This structure can provide redundancy and protection in the whole network.
(3) Communication Company / ISP Application
Cisco 12000 is also available in the network including multi-service network and IP, including multi-service in communication company / ISP network The network, including the network of IP predominations. Use 12000
Series GSR to the ATM web flat phase interface, the multiplexed traffic of the digital subscriber line can be selected to the Internet backbone network. The IP traffic can be transferred directly between the two core routers at the rate of OC-12 / OC-48; the data can also span the Internet backbone network.
(4) Dedicated Network Application
You can take advantage of IPOS technology in the IP private network. In this application, the edge router collects the business to the Cisco 12000 at the rate of OC-3 / STM-1. Cisco 12000 interconnects at rates of OC-12 / STM-4 or higher, expanding the available bandwidth and introduction of scalability.
(5) Hybrid network application
IPOS and ATM can form a mixed network. On the one hand, the ATM platform provides a multi-service function of the network, and on the other hand, it controls the case of users, so that the core router specializes in the Internet line 3 selection and service classification. The IPOS link is used to enhance the domain connection between the core routers.
The most important components in the 2nd-generation Internet (Internet2) is some data switches or routers called Gigapop or more than or more rates. They are interconnect points in a region, connecting in-Ternet2 members in the area and connects to other GigaPops. The main function of Gigapop is to exchange In-Ternet2 business as specified, and the Cisco's 12000 will play an important role in the GigapopoPoPoPoPoPoPo of Internet2. 12000 Series GSR can also be used to form a campus network.