Routing and Exchange is two important concepts in the network world. Traditional exchange occurs in the second layer of the network, that is, the data link layer, while the route occurs in the third layer, the network layer. In a new network, the intelligence and exchange of routing is organically combined, three-layer switches and multi-layer switches are used in a large number of in the park network. This article will introduce some basic concepts of routing and exchange, divided into four parts: network hierarchy, exchange, route, and all-exchange park networks.
Network hierarchy
The definition of the network reference model gives a clear functional level division. The most commonly mentioned is the ISO OSI Reference Model and TCP / IP Protocol Cluster.
The International Standardization Organization defined OSI reference model divides computer networks as seven hierarchies, which is the seven-layer model or seven-layer structure we often say. The direct benefits of network functionality are these levels that each other can be used, and different levels of hardware and software devices developed by different manufacturers can be used. A level of device updates or software rewrites does not affect other levels. The reference model of each level and ISO in the TCP / IP protocol system have a general correspondence. As shown below:
The OSI Mid-layer, that is, the fourth layer performs a transfer function, which is responsible for providing reliable data transfer from one computer to another computer. Transport Layer is a layer of starting, and there are three layers below it, both of which are related to data transfer; there are three layers above, providing features related to web applications.
Layer 3 in OSI. Physical Layer is responsible for actually transmitting data signals, data link layers are responsible for frame transmission within the network, and network layer is responsible for computer addressing and data transmission between networks.
On the three layers on the OSI. Application Layer is the highest level, it is responsible for providing user operation interfaces, and email services commonly used in the Internet are provided by this layer. Representation Layer is responsible for data representation, such as encryption before transmitting data, decryption when receiving data, translation of the Chinese and English is the functionality provided by this layer. Session Layer is responsible for establishing and terminating data transmission of the network, and the computer name converted into an address is also completed.
The exchange of exchanges is the concept of the second layer. The function of the data link layer is to transmit frames within the network. The so-called "network inside" refers to the transmission of this layer does not involve equipment and network addressing between the network. Popular understanding, transmission of an Ethernet, transmissions on a wide area network line are responsible by the data link layer. The "frame" refers to the structure of the transmitted data, usually frame head and frame end, and there is a source Layer 2 address in the head, and the tracking information is usually included, the content between the head is user The data.
The data link layer covers a lot, so it is divided into two sub-layers, MAC (Media Access Control, Media Access Control) and LLC (Logical Link Control, Location Link Control). Common LAN and Urban Area Network's Layer 2 standards are an IEEE's 802 protocol. In the WAN, HDLC (High-Level Data Link Control, Advanced Link Control), PPP (Point-to-Point Protocol, Point Protocol) and Frame Relay (Frame Relay) are widely used.
The route is the concept of the third layer. The network layer is most important in the Internet. Its function is the transmission of end-to-end. The meaning of end-to-end is how far, how many networks are separated from the intermediate, and this layer guarantees them to communicate with each other. For example, our commonly used ping command is a network of a network layer, and ping is, means that the network layer is functioning properly. Typically, the network layer does not guarantee the reliability of communication, that is, although the data can reach the destination under normal conditions, the network layer does not make any correction and recovery. The protocols commonly used by the network have IP, IPX, AppleTalk, etc., where IP protocol is the cornerstone of the Internet. In the TCP / IP protocol system, the other auxiliary protocols of the third layer also include ARP (address parsing), RARP (reverse address analysis), ICMP (Internet Packet Control), and IGMP (Group Management Protocol), and more. Since the network interconnecting devices have path selection functions, we often discuss the RIP, OSPF, etc. The protocol is also discussed.
exchange
Talking about the exchange of exchanges, from a broad sense, the forwarding of any data can be called exchange. Of course, we now refer to a narrow exclusion, including only the forwarding of the data link layer. Most people who do networks are mostly starting from the switch. The circuit switches have been used in the communication network for decades, and the equipment exchanged equipment, especially the large-scale use of Ethernet switches is in recent years.
It is understood that the role of the Ethernet switch will be spoken from the principles of the bridge. Traditional Ethernet is a shared type. If there are four computers A, B, C and D on the network segment, and the A and B communications, C and D can only be passive listening. If the cable segment is separated (i.e., "on a paragraph, C, D is on another paragraph, then C and D can also communicate, such that the original 10M bandwidth is in theory It is 20m. At the same time, in order to ensure that these two network segments can communicate with each other, they need to connect them with bridges, and the bridge is a computer with two network cards, as shown below:
When the entire network is just started, the bridge does not know anything about the topology of the network. At this time, it is assumed that the A sent data to B, because the network is a broadcast, so the bridge has also been received, but the bridge does not know that B is on the left side or the right, it will forward the default forward, that is, send it on another network card This information. Although it makes a useless forwarding, through this process, the bridge learned the sender A of the data A at your left. When each computer on the network is sent over data, the bridge is intelligent, it understands which network segment every computer. When A is sent to B, the bridge does not perform data forwarding, at the same time, c can send data to D.
As can be seen from the example above, the bridge can reduce the chances of network conflicts, which is the main purpose of our use of the bridge, referred to as reducing conflict domains. However, the bridge does not block broadcast, and the broadcast information is isolated from three-layer connection equipment, the router.
According to the thoughts of trace, the more cable segments, the higher the bandwidth. The limit is that every computer is on a separate cable segment. If there are ten computers on the network, you need a ten-port bridge to connect them. However, it is not reality that such a bridge is not realistic, and the speed of the software is also can't keep up. Therefore, there is a switch, the switch is to achieve the above multi-port bridge hardware or firmware to achieve lower cost and higher performance.
An important function of the switch is to avoid switching loops, which involves STP (Spanning Tree Protocol, branch tree protocol). The function of the branch tree protocol is to avoid cyclic delivery in the network composed of the data frame. As shown in the figure below, if there is a redundant link in the network, the STP protocol is set out of the Route Bridge and then determines the path between each non-root switch to the root switch, and finally on this path. All links are set to forward, and the connection between the remaining switches is a redundant link, which is set to block state. Another important feature of the switch is VLAN (Virtual LAN, virtual LAN). There are three advantages of VLAN:
• Division of ports. Even on the same switch, the port in different VLANs is also unable to communicate. Such a physical switch can be used as a switch for multiple logic.
• The security of the network. Different VLANs cannot communicate directly to prevent the unsafe of broadcast information.
? Flexible management. Changing the network to which the user is not allowed to change the port and connect, only the software configuration is OK.
The VLAN can be divided by port or MAC address.
Sometimes we need to maintain the consistency of the configuration of the VLAN on the network configured in the switch. This requires the switch to communicate with VLAN information according to VTP (VLAN TRUNK Protocol, VLAN backbone protocol). The VTP protocol only runs on the port of the backbone port, ie the port between the switches.
routing
The router is a network connection device, one of its important work is the path selection. This feature is the core of the router intelligence, which is implemented by the administrator's configuration and a series of routing algorithms.
The routing algorithm has a quiet point, and the static route is a special route, which is manually set by the administrator. Hand-configuring all routes can make the network run normally, but it also brings some limitations. After the network topology changes, the static route does not change automatically, and there must be intervention of network administrators. The default route is one of the static routes, and is also set by the administrator. When the routing entry for the target network is not found, the router sends information to the default router (GATEWAY OF Last Resort). The dynamic algorithm, as the name implies, is the route automatically calculated by the router, often said RIP, OSPF, etc. are typical representatives of dynamic algorithms.
It is also possible to divide the routing algorithm is two kinds of DV and LS. DV (Distance, Distance Vector) Algorithm transmits routing information of the current router to adjacent routers, adjacent routers to add this information to their own routing tables. Ls (LINK State, Link State) Algorithm transmits link status information to all routers in the domain, and receives the router to build network topology maps, and use the shortest path priority algorithm in the chart to determine the route. In contrast, the distance vector algorithm is relatively simple, and the link status algorithm is more complicated, and more CPUs and memory occupied. However, since the link state algorithm is used as its own calculation result, it is not easy to generate a routing cycle. RIP is a typical representative of the DV class algorithm, and OSPF is a representative protocol of LS.
Four most common routing agreements are
RIP, IGRP, OSPF, and EIGRP.
RIP (Routing Information Protocols, Routing Information Protocol) is the most wide range of distance vector protocols, which is developed by Xerox in the 1970s. At that time, RIP was part of the XNS (Xerox Network Service, Xerox Network Services) protocol cluster. The TCP / IP version of RIP is an improvement version of the Xerox Agreement. The biggest feature of RIP is that regardless of the principle or configuration method, it is very simple. RIP is based on hop calculation routing and sends update messages to neighbor routers. IGRP is a proprietary protocol proprietary and implemented in a Cisco router. It also belongs to the distance vector type, so there is a common point in many places, such as broadcast updates, and more. Its and RIP's largest differences are aspects of metrics, load balancing. IGRP supports weight load balancing on multiple paths, so that the bandwidth of the network can be utilized more reasonable. In addition, with the RIP only uses the number of hops as the measurement, IGRP uses a variety of parameters, constitutes a composite metric, which may include: bandwidth, delay, load, reliability, and MTU (maximum transmission unit), etc. Wait.
The OSPF protocol was developed in the 1980s. In the 1990s, it became an industrial standard, which is a typical link state protocol. The main features of OSPF include: supporting VLSM (shielding long subnet), fast convergence, low bandwidth usage, and the like. The OSPF protocol exchanges link status information between neighbors so that the router establishes a link state database (LSD). After that, the router uses SPF (Shortest Path First, shortest path priority) algorithm based on the information in the database, and selects the route. The main basis is bandwidth.
EIGRP is the enhanced version of IGRP, which is also a Cisco proprietary routing protocol. EIGRP uses a Diffusion Update (DUAL) algorithm, to some extent, it is similar to the distance vector algorithm, but has a shorter convergence time and better operability. As an extension of IGRP, EIGRP supports a variety of roudably protocols such as IP, IPX, and AppleTalk, and more. When running in an IP environment, EIGRP can also be smoothly connected to IGRP because their metric is consistent.
The above four routing protocols are domain internal routing protocols, which are usually used inside the autonomous system. When the connection between the autonomous system is often used, a domain routing protocol such as BGP (Border Gateway Protocols, boundary routing protocol) and EGP (External Gateway Protocols, external routing protocol) is often employed. The domain routing protocol currently used on the Internet is the fourth edition of BGP.
Convergence is an important issue encountered when the routing algorithm is selected. The convergence time refers to the change from the topology of the network to all related routers on the network to know this change, and the time required to change accordingly. The shorter this time, the smaller the interference of the network changes to the whole network. The convergence time is too long, resulting in the emergence of routing cycles.
In the above-mentioned various domain, the convergence time of RIP and IGRP is relatively long. It is a minute order; OSPF is short, and it can converge in dozens of seconds; the minimum of EIGRP, the network topology changes, a few seconds The convergence state can be reached.
All-exchange park network
Traditional park networks are the structure of a router plus switch. As shown in the figure below, the switch is responsible for the transmission within the network, dividing the VLAN to ensure the security and flexibility of the second layer, and the router completes the addressing and data forwarding of the network.
Typically, the performance of the router is better than the switch, because the router is based on the software-based surfacket forwarding, and the switch can achieve hardware through-form Forwarding. But in the traditional park network, the router does not become the bottleneck of the network. Because 80% of the amount of data is communication inside the network, only 20% of data is to do remote access, that is, most exchange information does not pass through the router. This is the 80/20 flow model of the traditional network. In recent years, due to the rise of Internet / intranet computing model, the application is centrally managed, rather than being dispersed in the network, and the traffic model of the park network has changed greatly. A large number of network access is remote, that is, the router is to be router. This is called a new 20/80 flow model. Therefore, the router gradually became the bottleneck of the network.
In order to solve this problem from technology, network manufacturers have developed three-layer switches and is also called routing switches. It is a combination of the performance of traditional switches and the intelligence of the router. Routing is still completed by the router, but the result of the circuit is switched in its own routing cache. Thus, the first data package in a data stream passes through the router, and all the packets of the subsequent data packets are forwarded directly by the switch. Thanks to hardware forwarding, the three-layer switch can do line speed routing, as shown below.
Many manufacturers produced the three-layer switch itself is a combination of switches and routers, such as Cisco's 5000, 5500, 6500 series of switches, can optionally match the routing module, implement three-layer function.
As a result, the internal area of the park is the switches and three-layer switches. The full exchange of parks adapted to new traffic models, completely overcoming the traditional network of router bottlenecks, which greatly improves the efficiency of the network. At the same time, the router is not unemployed and is still used in remote connections, dial-up access.
In the first part of this article, the hierarchy of the network is reviewed. Next, the principle of traditional routing and exchange is discussed. Finally, how to combine the advantages of routing and exchange in the current park network, it is best to meet the user's traffic model . Transfer from: http://www.windsn.com/blog/blogview.asp? LogId = 124
