The standard topology of Ethernet is bus topology, but the current fast Ethernet (100BASE-T,100base-t standards) uses switching hubs to connect and organize networks in order to minimize collisions and maximize network speed and efficiency. In this way, the topology of Ethernet becomes a star, but logically, Ethernet still uses CSMA/CD (Carrier Sense Multiple Access/Collision Detection) bus topology and bus contention technology.
Directory [hidden]
1 history
2 overview
3 CSMA/ CD * * * Enjoy media Ethernet.
4 Ethernet repeaters and hubs
5 Bridging and switching
6 Ethernet type
6. 1 Early Ethernet
6.2 10Mbps Ethernet
6.3 100Mbps Ethernet (fast Ethernet)
6.4 1Gbps Ethernet
65000 Gigabit Ethernet
66 PCs 100 Gigabit Ethernet
7 references
8 See also
9 external link
[Edit] History
The initial development of Ethernet technology came from one of many pioneering technical projects of Xerox Palo Alto Research Center. People usually think that Ethernet was invented in 1973, when bob metcalfe wrote a memo to his PARC boss about the potential of Ethernet. But Metcalfe himself thinks that Ethernet didn't appear until several years later. 1976, Metcalfe and his assistant David Boggs published an article entitled "Ethernet: Distributed Packet Switching Technology in Regional Computer Networks".
network protocol
application layer
DHCP, DNS, FTP, Gopher, http, IMAP4, IRC, NNTP, XMPP, POP3, SIP, SMTP, SNMP, SSH, Telnet, RPC, RTCP, RTP, RTSP, SDP, SOAP, GTP Stun, NTP, SSDP, BGP, RIP and so on.
transport layer
TCP UDP TLS DCCP SCTP
Reply to PPTP·OSPF· Moore
Network layer
IP(IP v4 IPv6)ARP RARP ICMP ICMP V6 IGMP
Is IPSec more
data link layer
wi-Fi(IEEE 802. 1 1)WiMAX(IEEE 802. 16)
ATM DTM token ring Ethernet FDDI Frame Relay GPRS EVDO HspA HDLC PPP L2TP ISDN STP more.
physical layer
Ethernet, modem, power line communication (PLC), SONET/SDH G.709, optical fiber, coaxial cable, twisted pair and so on.
This template: View? Discuss? Editor? history
1979, Metcalfe left Xerox to set up 3Com in order to develop personal computers and local area networks. 3Com lobbied DEC, Intel and Xerox, hoping to realize the standardization and normalization of Ethernet with them. This universal Ethernet standard was released on September 30th, 1980 in/kloc-0. At that time, there were two popular non-public network standards, token ring network and ARCNET, which quickly shrank and were replaced under the impact of Ethernet tide. In the process, 3Com also became an international company.
Metcalfe once joked that Jerry Salz contributed to the success of 3Com. In an influential paper co-authored with others, Saltzer pointed out that token ring network is superior to Ethernet in theory. Influenced by this conclusion, many computer manufacturers either hesitate or decide not to use Ethernet interface as the standard configuration of their machines, so 3Com has the opportunity to make a fortune by selling Ethernet cards. This situation has also led to another saying that "Ethernet is not suitable for theoretical research, but only suitable for practical application". Maybe this is just a joke, but it illustrates a technical point: usually, the actual data flow characteristics in the network are different from those estimated before the popularization of LAN, and it is precisely because of the simple structure of Ethernet that LAN is popularized. Metcalfe and Saltzer used to work on the same floor of the Massachusetts Institute of Technology (MIT) MAC project, when he was doing his graduation thesis at Harvard University, during which he laid the theoretical foundation of Ethernet technology.
[Edit] Overview
1990s Ethernet card or NIC (network interface card). The card can support 10BASE2 (BNC connector, left) based on coaxial cable and 10baset (RJ-45, right) based on twisted pair.
Ethernet is based on the idea that multiple nodes of the radio system on the network send information, and each node must obtain a cable or channel to transmit information, sometimes called Ethernet. The name comes from/kloc-optical ether, an electromagnetic radiation medium assumed by physicists in the 9th century. Later research proved that optical ether did not exist. Each node has a unique 48-bit address in the world, that is, the MAC address assigned by the manufacturer to the network card, so as to ensure that all systems on the Ethernet can authenticate each other. Because Ethernet is very common, many manufacturers directly integrate Ethernet cards into computer motherboards.
It is found that Ethernet communication has autocorrelation characteristics, which is very important for telecom engineering.
[Edit ]CSMA/CD*** Enjoy Media Ethernet
Carrier Sense Multiple Access (CSMA/CD) technology with collision detection provides a way for multiple computers to share a channel. This technology first appeared in ALOHAnet developed by the University of Hawaii in1960s. It uses radio waves as the carrier. This method is simpler than token ring network or main control network. When a computer wants to send information, it must follow the following rules:
Start-If the line is idle, start transmission, otherwise go to step 4.
Send-If a collision is detected, continue to send data until the minimum message time is reached (to ensure that all other repeaters and terminals detect the collision), and then go to step 4.
Transmission success-reports the transmission success to the higher layer network protocol and exits the transmission mode.
Busy-Wait until the line is free.
The line enters the idle state-wait for a random time and go to step 1 unless the maximum number of attempts is exceeded.
Maximum transmission attempts exceeded-Report transmission failure to higher layer network protocol and exit transmission mode.
Just like in a forum without a moderator, all participants talk to each other through a common medium (air). Every participant politely waits for others to finish before speaking. If two guests start talking at the same time, they will stop and wait for a period of time at random before they start talking. At this time, if the two participants wait for different times, there will be no conflict. If the transmission fails more than once, the back-off time will be increased exponentially (the back-off time is realized by truncated binary exponential back-off algorithm).
The original Ethernet used coaxial cables to connect various devices. The computer is connected to the cable through a transceiver called the Attachment Unit Interface (AUI). Simple network cable is still very reliable for small networks. For large networks, the failure of a line or a connector will lead to instability of one or more segments of Ethernet.
Because all communication signals are transmitted on * * * lines, even if the information is sent to only one terminal (destination), the message sent by one computer will be received by all other computers. Under normal circumstances, the network interface card will filter out the information not sent to it, and only when the destination address is its own information will it send an interrupt request to the CPU, unless the network interface card is in Promiscuous mode mode. The characteristic of "one person says, everyone listens" is the security weakness of Ethernet, because a node on Ethernet can choose whether to monitor all the information transmitted on the line. * * * Shared cable means * * * shared bandwidth, so in some cases, the speed of Ethernet may be very slow, such as when all network terminals are restarted after power failure.
[Edit] Ethernet repeaters and hubs
In the development of Ethernet technology, the emergence of Ethernet hub makes the network more reliable and the wiring more convenient.
Due to the attenuation and delay of signals, there are distance restrictions according to different media Ethernet segments. For example, the longest distance of a 10BASE5 coaxial cable is 500 meters (1 640 feet). The maximum distance can be achieved by Ethernet repeater, which can amplify the signal in the cable and then transmit it to the next section. Repeaters can be connected to up to 5 network segments, but only 4 devices can be connected (that is, a network segment can be connected to up to 4 repeaters). This can alleviate the problem caused by cable breakage: when a coaxial cable is disconnected, all devices on this section can't communicate, and the repeater can ensure the normal operation of other network segments.
Similar to other high-speed buses, both ends of an Ethernet segment must be terminated with resistors. For coaxial cable, the terminals at both ends of the cable must be connected with a 50 ohm resistor and a radiator called "terminator" and connected to an M or BNC male connector. If you don't do this, it will be similar to the situation that the cable is broken: the AC signal on the bus will be reflected when it reaches the terminal, but it will not be dissipated. The reflected signal will be considered as a collision, so the communication cannot continue. Repeater can electrically isolate the two network segments connected to it to enhance and synchronize signals. Most repeaters have a function called "automatic isolation", which can isolate network segments with too many conflicts or too long conflicts, so that other network segments will not be affected by damaged parts. The repeater can restore the connection of the network segment after detecting that the conflict disappears.
With the expansion of application, people gradually found that star network topology is the most effective, so equipment manufacturers began to develop multi-port repeaters. Multiport repeaters are well-known hubs. Hubs can be connected to other hubs or coaxial networks.
The first hub is called a "multiport transceiver" or a "fan-out". The most famous example is DEC's DELNI, which enables many hosts with AUI connectors to use a transceiver. Hubs have also led to the emergence of small independent Ethernet segments without coaxial cables.
Network equipment manufacturers, such as DEC and SynOptics, have sold hubs for connecting many 10BASE-2 thin coaxial segments.
Unshielded twisted pair (UTP) was first used in star LAN, then used in 10BASE-T, and finally replaced coaxial cable as the standard of Ethernet. After this improvement, RJ45 telephone interface has replaced AUI as the standard interface of computers and hubs, and unshielded class 3 twisted pair/class 5 twisted pair has become the standard carrier. The application of hub makes the failure of a cable or a device not affect the whole network, which improves the reliability of Ethernet. Twisted-pair Ethernet connects all network segments point to point, so that the terminal can be made into a standard hardware, which solves the terminal problem of Ethernet.
Although Ethernet with hub network is physically a star structure, it is still a bus type in logic. Half-duplex communication adopts CSMA/CD collision detection method, and the hub has little effect on reducing packet collision. Every packet is sent to every port of the hub, so the bandwidth and security problems remain unresolved. The total throughput of the hub is limited by the single connection speed (10 or 100 Mbit/s), but the lowest cost of preamble, frame interval, header, trailer and encapsulation is still considered. When the network load is too heavy, collisions usually reduce the total throughput. In the worst case, when many hosts with long cables send many short frames and the network load is only 50%, the throughput of the hub will be reduced due to the collision. In order to increase the network load as much as possible before the collision seriously reduces the throughput, it is usually necessary to do some setup work.
[Edit] Bridging and Switching
Although the repeater isolates the Ethernet segment in some ways, the cable disconnection will not affect the whole network, but it will forward all data to all Ethernet devices. This severely limits the number of machines that can communicate with each other on the same Ethernet. In order to alleviate this problem, the bridge method is adopted, and the repeater works at the data link layer on the basis of working at the physical layer. When passing through a bridge, only well-formed packets can enter from one network segment to another; Conflicts and packet errors are isolated. By recording and analyzing the MAC addresses of devices on the network, the bridge can determine their locations, so that it will not pass packets to the network segment where non-target devices are located. Control mechanisms such as Spanning Tree Protocol can coordinate multiple switches to work together.
Early bridges had to detect every packet, so the data forwarding speed was slower than that of hubs (repeaters), especially when dealing with multiple ports at the same time. 1989 network company Kalpana invented the first EtherSwitch. Ethernet switch uses hardware to realize bridging function, which makes the forwarding data rate reach line speed.
Most modern Ethernet networks use Ethernet switches instead of hubs. Although the wiring is the same as hub Ethernet, switched Ethernet has many obvious advantages over * * * media Ethernet, such as wider bandwidth and better isolation of abnormal devices. The switching network usually uses star topology. Although the device works in half-duplex mode, it is still a multi-node network that enjoys media. 10BASE-T and later standards are full-duplex Ethernet, which is no longer a * * * media sharing system.
After the switch is powered on, it first works like a hub and forwards all data to all ports. Next, when it knows the address of each port, it only sends non-broadcast data to a specific destination port. In this way, any port pair can realize line-speed Ethernet exchange, and the communication between all port pairs does not interfere with each other.
Because packets are usually sent only to their destination ports, the traffic on switched Ethernet is slightly less than that on media Ethernet. Nevertheless, switched Ethernet is still an unsafe network technology, because it is easily paralyzed by ARP spoofing or MAC overflow, and network administrators can also use the monitoring function to grab network packets.
When only simple devices (devices other than hubs) access the switch ports, the whole network may work in full-duplex mode. If there are only two devices in a network segment, there is no need for collision detection, and the two devices can send and receive data at any time. The total bandwidth is twice that of the link (although the bandwidth in each direction is the same), but collision-free means that almost 100% of the link bandwidth is allowed.
Switch ports and connected devices must use the same duplex settings. Most 1000BASE-Tx and1000base-t devices support auto-negotiation function, that is, these devices coordinate the rate and duplex settings to be used through signals. However, if auto-negotiation is disabled or devices don't support auto-negotiation, you must set the duplex setting through automatic detection, or manually set the duplex setting on switch ports and devices to avoid duplex mismatch, which is a common cause of Ethernet problems (devices set to half-duplex will report delay conflicts, while devices set to full-duplex will report stunting). Many low-end switches do not have the ability to set the speed and duplex manually, so ports always try to negotiate automatically. When auto-negotiation is enabled but unsuccessful (for example, other devices do not support it), auto-negotiation sets the port to half-duplex. The rate can be automatically detected, so when the auto-negotiated 10/ 100 switch port is enabled, a half-duplex 10BASE-T connection can be successfully established. However, when a device configured for full-duplex 100Mb operation is connected to a switch port configured for auto-negotiation (or vice versa), it will result in duplex mismatch.
Even if both ends of the cable are set to automatic rate and duplex mode negotiation, there will often be false guesses and return to 10Mbps mode. Therefore, if the performance is worse than expected, you should check whether any computer is set to 10Mbps mode. If you know that the other end is configured as 100Mbit, you can manually force it to be set to the correct mode. .
When two nodes try to communicate at a data rate that exceeds the maximum data rate supported by the cable (for example, 1000Mbps on a Class 3 line or1000 Mbps on a Class 3/Class 5 line), a problem occurs. Unlike ADSL or traditional dial-up modems, they will detect the highest data rate supported by the link in detail, while Ethernet nodes only choose the highest data rate supported at both ends, regardless of the intermediate line. Therefore, if the cable is unreliable due to high speed, the link will fail. The only solution is to force the communication end to slow down to the speed supported by the cable.
[Edit] Ethernet type
In addition to the different frame types mentioned above, the difference between various types of Ethernet is only speed and wiring. Therefore, generally speaking, the same network protocol stack software can run on most Ethernet networks.
The following sections briefly summarize the different formal Ethernet types. In addition to these formal standards, many manufacturers have formulated some special standards for some special reasons, such as supporting longer-distance optical fiber transmission.
Many Ethernet cards and switching devices support multi-rate, and devices set the best connection speed and duplex mode through automatic negotiation. If the negotiation fails, the multi-rate device will detect the rate used by the other party, but it defaults to half-duplex mode. 10/ 100 Ethernet port support 10BASE-T and 100BASE-TX. 10/100/1000 supports 10 base-t, 100 base-tx and1000base-t.
[Edit] Early Ethernet
See: Gigabit Ethernet.
Xerox Ethernet (also called "Full Recording Ethernet") is the prototype of Ethernet. The original 2.94Mbit/s Ethernet was only used in the whole record company. 1982, Xerox formed the DIX alliance with DEC and Intel, and jointly issued the specification of Ethernet Version 2(EV2), which was put into the market of shopping malls and was widely used. The network of EV2 is currently recognized by IEEE as 10BASE5. [ 1]
This is out of date. Early standards supporting long-distance Ethernet. It runs on coaxial cable and uses broadband modulation technology similar to cable modem system.
1BASE5 ── also known as star LAN, the speed is 1mbit/s, and the business fails. Twisted pair is used here for the first time.
[Edit] 10Mbps Ethernet
10BASE-T cable
See: 10 Gigabit Ethernet
10BASE5 (also known as coarse Ethernet or yellow line)-the earliest 10 Mbit/s Ethernet. In the early IEEE standard, a single RG- 1 1 coaxial cable was used, and the maximum distance was 500 meters, which could connect the transceivers of 100 computers at most. A 50 ohm terminal resistor must be connected at both ends of the cable. The receiving end is inserted into the inner core and shielding layer of the cable through the so-called "plug-in tap". Use n-type connector at cable terminal. Although some systems are still in use due to early massive deployment, this standard has actually been abandoned and eliminated by 10BASE2.
10BASE2 (also known as thin Ethernet or analog network) ── The products after10base5 use RG-58 coaxial cable, with the longest transmission distance of about 200m (actually,185m), and only 30 computers can be connected. Computers use T-adapter to connect the network card of BNC connector. Although it is not as good as 10BASE5 in capacity and specification, it is widely used because of its thin wires, convenient wiring and low cost, and 10BASE5 has been eliminated. Because of the popularity of twisted pair, it has also been replaced by various twisted pair networks.
Starlan- Ethernet standard 10 Mbit/s is implemented on the first twisted pair. Later it developed into10base-t.
10BASE-T ── Use four wires (two pairs of twisted pairs) of Class III twisted pair, Class IV twisted pair and Class V twisted pair at100m. Ethernet hub or Ethernet switch is located in the middle, connecting all nodes.
Optical fiber repeater link. The original version of Fibre Ethernet.
10BASE-F ── 10Mbps Ethernet optical fiber standard, 2km. Only 10BASE-FL is widely used.
10base-fl ── upgrade of foirl standard.
10base-FB-The backbone network technology used to connect multiple hubs or switches has been abandoned.
10base-FP-A passive star network without relay has never been used.
[Edit] 100Mbps Ethernet (Fast Ethernet)
See: 100Mbps Ethernet
Fast Ethernet is a network standard published by IEEE in 1995, which can provide transmission speed as high as 100Mbps. [2]
100ba se-T-T- The following three twisted-pair standards are collectively called 100 Mbit/s, and the longest distance is 100 m..
100base-tx-10base-t has a similar star structure. Two pairs of cables are used, but Category 5 cables are needed to reach100 mbit/s. 。
100ba se-T4-T4- Use Category 3 cable, use all 4 pairs, half duplex. Due to the popularity of Category 5 line, it has been abandoned.
100 base-T2-T2- no product. Use Category 3 cables. Supports full duplex using 2 pairs of wires, which is equivalent to 100BASE-TX, but supports old cables.
100ba se-FX-FX- uses multimode fiber and supports up to 400m, half-duplex connection (to ensure collision detection) and 2km full duplex.
100 VG anylan-VG first appeared in the market only with HP's support. Four pairs of Class III cables are required. Some people suspect that VG is not Ethernet.
[Edit] 1Gbps Ethernet
See: 1Gbps Ethernet.
1000BASE-SX optical and electrical signal converter
1000base-t-1gbit/s medium super five-level twisted pair or six-level twisted pair.
1000base-sx-1gbit/s multimode fiber (less than 500m).
1000base-LX-1gbit/s multimode fiber (less than 2KM).
1000base-LX10-1gbit/s single-mode fiber (less than 10KM). Long-distance scheme
1000base-lhx-1gbit/s single-mode fiber (10KM to 40KM). Long-distance scheme
1000base-zx-1gbit/s single-mode fiber (40km to 70km). Long-distance scheme
1000ba se-CX-CX- short distance (less than 25m) scheme on copper cable 1Gbps. It was earlier than 1000BASE-T and has been abandoned.
[Edit] 10 Gigabit Ethernet
See: 10 Gigabit Ethernet.
The new 10 Gigabit Ethernet standard contains seven different control types, which are suitable for local area network, metropolitan area network and wide area network respectively. At present, the additional standard IEEE 802.3ae is used to illustrate that the IEEE 802.3 standard will be merged in the future.
10g base-CX4- Short-distance copper cable scheme is used for InfiniBand 4x connectors and CX4 cables, and the maximum length is 15m.
10g base-SR-SR- is used for short-distance multimode fiber, which can reach 26-82m depending on the cable type and 300m when using the new 2GHz multimode fiber.
10g base-LX4- Wavelength Division Multiplexing (WDM) is used to support multimode fiber with a length of 240-300m and single-mode fiber with a length of10km or more.
10GBASE-LR and 10g base-ER- support 10 km and 40 km respectively through single-mode optical fiber.
10GBASE-SW, 10GBASE-LW, 10GBASE-EW. Used for WAN PHY, OC- 192/STM-64 synchronous optical network /SDH equipment. The physical layer corresponds to 10Gbase-Sr, 10Gbase-LR and 10GBASE-ER respectively, so the distances supported by the same optical fiber are consistent. (There is no universal standard)
10g base-T-T- Use shielded or unshielded twisted pair, use 6A cat cable, and at least support 100m transmission. CAT-6 cable also supports short distance10gbase-t.
[Edit] 100 Gigabit Ethernet
See: 100G Ethernet.
The new 40G/ 100G Ethernet standard was formulated on 20 10, including several different types of control. At present, it is explained by the additional standard IEEE 802.3ba
40GBASE-KR4-backplane scheme, minimum distance1m.
40gbase-Cr4/100gbase-Cr10-short-distance copper cable scheme, the maximum length is about 7m.
40gbase-sr4/100gbase-sr10-is used for short-distance multimode fiber with a length of at least100 meters.
40gbase-lr4/100gbase-lr10-uses single-mode optical fiber, and the distance exceeds 10km.
Use 100 gbase-ER4- single-mode fiber, and the distance is more than 40km.
Personal deeds of advanced teachers' morality 2000 words 1
Some people say, "winning the soul with your heart is