Million Gigabit Ethernet
This entry is compiled and applied to the scientific entry of "Science China" encyclopedia. The Ethernet standard is an old and vibrant standard. Since the 1982 Ethernet Protocol was adopted by IEEE as a standard, it has experienced 20 years of ups and downs. In these 20 years, Ethernet technology, as the standard of LAN link layer, overcomes the technology of token bus and Token ring, and becomes the fact standard of LAN. Ethernet technology currently has a market share of over 90% in the LAN range.
Chinese name
10 Gigabit Ethernet
Web Standardis
an old and vibrant standard.
Web Technology
Wins the token bus as a local standard
After 20 years
the ups and downsDirectory
- 1 Introduction
- 2 Establishment Background
- 3 Technical Introduction
- 4 Technology Outlook
- 5 Technical Features
- 6 Technical Advantages
- 7 Development History
Introduction
in these 20 years, Ethernet was developed from the original 10BASE5 10M thick cable bus to 10base2 10M thin cable, followed by a short back: 1BASE5 1 Gigabit Ethernet, then Ethernet technology developed into a familiar star-shaped twisted pair 10BaseT. With the increase of the bandwidth requirement and the enhancement of the device capability, there is fast Ethernet: The 100BaseTX of the five kinds of line transmission, the 100baset4 of the three kinds of wire transmission and the 100BaseFX of fiber transmission. With the further increase of the bandwidth, the Gigabit Ethernet interface is staged: including the short wavelength optical transmission 1000base-sx, the long wavelength optical transmission 1000base-lx and the five kinds of line transmission 1000BaseT. July 18, 2002 IEEE passed 802.3ae:10gbit/s Ethernet, also known as Gigabit Ethernet.
Gigabit Ethernet does not simply increase the rate of Gigabit Ethernet to 10 times times, there are many technical problems to solve. [1]
in Ethernet technology, 100BaseT is a milestone that establishes the dominant position of Ethernet technology on the desktop. Gigabit Ethernet and the subsequent Gigabit Ethernet standards are two of the most important standards, with Ethernet technology extending from the desktop LAN technology to the campus network and the convergence and backbone of the metropolitan area network through these two standards..
The evolution of Ethernet from 10mb/s to 10gb/s proves that Ethernet is:
1) extensible (from 10mb/s to 10gb/s).
2) Flexible (multiple media, full/half duplex, shared/switched).
3) easy to install.
4) Good robustness. [1]
Setting up the background
Ethernet mainly in the LAN occupies an absolute advantage. But for a long time, it was widely believed that Ethernet could not be used in the metropolitan area Network,Related InformationEspecially the aggregation layer and the backbone layer. The main reason is that Ethernet is used as the backbone bandwidth of metropolitan area network is too low (10M and 100M Fast Ethernet ERA), transmission distance is too short. At that time, the most promising metropolitan area network technology was FDDI and DQDB. In the years that followed, ATM technology became a hotspot, and almost everyone thought that ATM would be the only technology for unifying local area networks, metropolitan areas, and WANs. But for a variety of reasons, at present, only the ATM technology in the three of the above technologies has become the urban area network convergence layer and the backbone of the option.
The most common Ethernet currently is 10M Ethernet and 100M Ethernet (Fast Ethernet). 100M Fast Ethernet is obviously not enough for the bandwidth of the metropolitan backbone network. Even with the use of multiple Fast Ethernet link bindings, the multimedia business remains powerless. With the standardization of Gigabit Ethernet and the extensive application in production practice, Ethernet technology is gradually extended to the aggregation layer of metropolitan area network. Gigabit Ethernet is typically used to bring community users to a metro pop point, or to connect a converged layer device to a backbone layer. However, in the current 10M Ethernet to the user environment, the Gigabit Ethernet link as a convergence is also reluctant, as the backbone of the force can not be. Although Ethernet multi-link aggregation technology has been standardized and multi-vendor interoperability is imminent, multiple gigabit links can be bundled. But considering fiber resources and wavelength resources, link bundles are generally used only in pop points or in short-range applications.
The transmission distance was once the obstacle that Ethernet could not be used as the backbone layer link technology of metropolitan Data network. Whether it is 10M, 100M or Gigabit Ethernet, the five line transmission distances are 100m due to signal-to-noise ratio, collision detection, and available bandwidth. The use of fiber-optic transmission distance limitation is restricted by the master-slave synchronization mechanism used by Ethernet. 802.3 Specifies the 1000BASE-SX interface uses the fiber core 62.5μm multimode fiber Maximum transmission distance 275m, uses the core 50μm multimode fiber Maximum transmission distance 550M;1000BASE-LX interface uses the fiber core 62.5μm multimode fiber Maximum transmission distance 550m, uses the fiber core 50μm the multimode light The longest transmission distance is 550m, the longest transmission distance is 5000m using a single mode fiber core of 10μm. The maximum transmission distance of 5km Gigabit Ethernet links in the metropolitan area is far from enough. Although the Gigabit interface based on the manufacturer has been able to achieve 80km transmission distance, and some vendors have completed interoperability testing, but after all, non-standard implementation, can not guarantee that all vendors of such interface interconnection.
To sum up, Ethernet technology is not suitable for the backbone/aggregation layer of metropolitan Area Network is the main reason is the bandwidth and transmission distance. With the advent of Gigabit Ethernet technology, the above two problems have been basically solved.
Technology Introduction
Ethernet adopts CSMA/CD mechanism, that is, carrier monitoring with collision detection multi-access. The Gigabit Ethernet interface is basically applied to point-to line and no bandwidth is shared. Collision detection, carrier monitoring, and multiple access are no longer important. ThousandThe biggest similarity between gigabit Ethernet and traditional Low speed Ethernet is the same Ethernet frame structure. Gigabit Ethernet technology is similar to gigabit Ethernet and still retains the Ethernet frame structure. The 10GBIT/S transmission speed is provided through different encoding methods or wavelength division multiplexing. So in its essence, 10G Ethernet is still a type of Ethernet.
10G Ethernet was passed in IEEE in July 2002. 10G Ethernet includes 10gbase-x, 10gbase-r, 10gbase-w and copper-based 10gbase-t, etc. (passed in 2006). The 10gbase-x uses a special compact package that contains 1 simpler WDM devices, 4 receivers and 4 lasers that work at approximately 25nm intervals around 1300nm wavelengths, with each pair of transmitters/receivers at 3.125gbit/s speed (data flow speed of 2.5gbit/ s) to work under. 10gbase-r is a serial interface that uses 64B/66B encoding (not 8b/10b used in Gigabit Ethernet), and the data stream is 10.000gbit/s, resulting in a clock rate of 10.3gbit/s. The 10gbase-w is a WAN interface, compatible with SONET OC-192, Gigabit Ethernet Its clock is 9.953gbit/s data stream is 9.585gbit/s.
10G Serial physical media layer the fiber Gigabit Ethernet specification
For LAN is: 10GBASE-SR, 10GBASE-LR and 10gbase-er.
1. The "SR" in 10GBASE-SR10GBASE-SR is the abbreviation for short range, which means it is only used for short distance connections. The specification supports short-wave (850nm wavelength) multimode fiber (MMF) encoded in 64b/66b with an effective transmission distance of 2 m to 300 m.
2. The "LR" in 10GBASE-LR10GBASE-LR is the abbreviation for "Long Range" (long distance), which means it is mainly used for long distance connections. The specification supports long-wave (1310nm) single-mode optical fiber (SMF) encoded in 64b/66b with an effective transmission distance of 2 m to 10 km.
4, "ER" in 10gbase-er10gbase-er is the abbreviation of "Extended Range" (ultra long distance), indicating that the connection distance can be very long. The specification supports an ultra long-wave (1550nm) single-mode optical fiber (SMF) encoded in 64b/66b, with an effective transmission distance of 2 m to 40 km.
The function of the PMD
sub-layer of PMD sub-layer is to support the exchange of serialized symbol code BITS between the PMA sub-layer and the media. The PMD sub-layer converts these electrical signals into a form suitable for transmission on a particular medium. PMD is the lowest sub-layer of the physical layer, which specifies that the physical layer is responsible for sending and receiving signals from the media.
The PMA sub-layer feature PMA
sub-layer provides a serialization service interface between the PCs and the PMD layer. The connection to the PCs sub-layer is called the PMA Service interface. In addition, the PMA sub-layer is also separated from the receive bit stream to connect the received data for the correct symbolic Alignment (delimitation) of the symbol timing clock.
The WIS sub-layer feature
WIS sub-layer is an optional physical sub-layer that can be used between PMA and PCs to produce a SONET STS-192C transmission format that is suitable for ANSI definition or an ITU-defined Ethernet data stream that defines the rate of SDH vc-4-64c containers. This rate data stream can be mapped directly to the transport layer without the need for high-level processing.
The PCs sub-layer function
PCs sub-layer is located between the coordination sub-layer (via gmii) and the physical media access Layer (PMA) sub-layer. The PCS sub-layer completes the function of mapping well-defined Ethernet MAC capabilities to existing coding and physical layer signaling systems. The interfaces of the PCs sublayer and upper Rs/mac are provided by XGMII and the PMA service interface is used with the downlevel PMA interface.
The function of the RS (coordination sub-layer) and the XGMII
coordination sub-layer is to map Xgmii's path data and associated control signals to the original PLS service Interface definition (MAC/PLS) interface. The Xgmii interface provides logical connections between the Mac and the physical layers of the 10gbit/sMouth. The Xgmii and coordination sub-layers enable Macs to connect to different types of physical media.
Since 10G Ethernet is essentially high-speed Ethernet, the traditional Ethernet-based frame format must be used to carry the business in order to be compatible with traditional Ethernet. In order to achieve the high rate of 10gbit/s can be transmitted in OC-192C frame format. This requires a mapping function from the Ethernet frame to the OC-192C frame format in the physical sub-layer. At the same time, because the original design of Ethernet is facing LAN, the network management function is weak, the transmission distance is short and the physical circuit has no protection measures. When Ethernet is used as a wide-area network for long-distance and high-speed transmission, the frequency and phase of the line signal will be greatly jitter, and the transmission of Ethernet is asynchronous, so it is difficult to realize signal synchronization at the receiving end. Therefore, if the Ethernet frame is to be transmitted over the WAN, the Ethernet frame format needs to be modified.
Ethernet typically uses special 10B (Byte) code in the physical layer to achieve frame delimitation. When the MAC layer has data to be sent, the PCS sub-layer 8b/10b the data, and when the frame head and frame end are found, the Special code group SFD (frame start delimiter) and EFD (frame ending delimiter) are automatically added, and when the PCS sub-layer receives the 10B encoded data from the underlying, It is easy to find the start and end of frames based on SFD and EFD to complete the frame delimitation. However, the Gigabit Ethernet frame delimitation in SDH is different from the standard Gigabit Ethernet frame delimitation because the multiplexed data has been restored to a 8B encoded code set, removing SFD and EFD. If the frame delimitation is performed only with the preamble (preamble) and Frame start delimiter (SFD) of the gigabit Ethernet, because the probability of the same code group appearing in the information data is the same as the leading and frame starting delimiters, the use of such a frame-bound policy may cause the receiver to always be unable to make the correct Ethernet frame delimitation. In order to avoid the above situation, 10G Ethernet adopts the HEC strategy.
The IEEE802.3 HSSG team proposed to modify the Gigabit Ethernet frame format, adding the length domain and the HEC domain to the Ethernet frame. Gigabit Ethernet In order to make it easy to find the next frame position during the frame process, and because the maximum frame length is 1518 bytes, it requires a minimum of 11 bits (=2048), so we replace the leading header two bytes with two bytes as the Length field in the process of re-connecting the Mac frame, and then the 8 bytes are CRC-16 verified. The last two bytes are inserted as HEC after the SFD.
The 10G WAN physical layer is not simply hosting Ethernet Mac frames with oc-192c. Although the oc-192c frame structure, pointers, mappings and layered overhead are used for reference, there is a lot of simplification on the basis of SDH frame structure, which makes the modified Ethernet not sensitive to jitter and the requirement of clock is not high. Specific performance in: reduced a lot of overhead bytes, only the use of frame positioning bytes A1 and A2, segment layer error monitoring B1, trace byte J0, synchronization status byte S1, protection switching bytes K1 and K2, and alternate byte Z0, the byte is not defined or not used to fill 00000000. It reduces unnecessary overhead, simplifies SDH frame structure, enhances the network management and maintenance of the physical layer compared with gigabit Ethernet, and can implement protection switching on the physical circuit. Secondly, the tedious synchronous multiplexing is avoided, and the signal is not multiplexed into high-speed stream from low rate, but directly mapped to oc-192c net load.
10G Ethernet LAN and 10G Ethernet wide area network (using OC-192C) physical layer of different rates, 10G Ethernet LAN Data rate is 10GBIT/S, and 10G Ethernet WAN Data rate is 9.58464gbit/s (SDH oc-192c, is the rate at which the PCS layer is not encoded), but the physical layer of the two rates is shared with a MAC layer, and the MAC layer has a working rate of 10gbit/s. The problem that 10G Ethernet WAN needs to be solved is to reduce the 10gbit/s transfer rate of 10GMII interface to match the transfer rate 9.58464gbit/s of the physical layer with the adjustment strategy. At present, the 10gbit/s rate is adapted to 9.58464gbit/s oc-192c adjustment strategy has 3 kinds:
1:at the Gmii interface to send hold signal, the MAC layer in a clock cycle stop sending;
2:using "Busy idle", the physical layer is sent to the MAC layer during IPG Busy Idle ", after the MAC layer is received, pause sending data. The physical layer sends "Normal idle" to the MAC layer during IPG, the MAC layer receives the data again;
3:uses the IPG extension mechanism: Mac frames each time a frame is passed, and the IPG interval is dynamically adjusted based on the average data rate.
Technology Outlook
Gigabit Ethernet takes into account the needs of the metropolitan backbone network at the beginning of the design. First, the bandwidth of 10G enough to meet the current stage and the next period of time in the inner-city backbone network bandwidth requirements (at present, most metropolitan backbone backbone bandwidth does not exceed 2.5G). Second Gigabit Ethernet maximum transmission distance up to 40 km, and can be used in conjunction with the 10G transmission channel, sufficient to meet the majority of urban metropolitan Area Network coverage. Using Gigabit Ethernet as the backbone of metropolitan area Network can omit the POS or ATM link of backbone network equipment. First, cost savings: The Ethernet port price is much lower than the corresponding POS port or ATM port. The second can make the end-to-end adoption of Ethernet frames possible: On the one hand can end-to-end use of the link Layer VLAN information and priority information, on the other hand can be omitted on the data device multiple link layer Encapsulation solution encapsulation and the possible existence of packet fragmentation, simplifying network equipment. Using Gigabit Ethernet link in the backbone layer of metropolitan area Network can improve the cost-performance of network and simplify the network.
We can clearly see that 10G Ethernet can be applied in campus network, metropolitan area Network, enterprise network and so on. However, since the current broadband service is not widely implemented, there is no urgent need for the bandwidth of single-port 10G backbone, so 10G Ethernet technology has no obvious advantage over other alternative link layer technologies such as 2.5G POS, bundled Gigabit Ethernet. Cisco and Juniper have introduced 10G Ethernet interfaces (implemented according to the 802.3ae draft), but there is little application in the country. At present, the problem of metropolitan area Network is not the lack of bandwidth, but the killer of bandwidth consumption Application is how to build the metropolitan area network into a manageable, operational and profitable network. So the application of 10G Ethernet technology will depend on the development of broadband service. Only a wide range of broadband services, such as video multicast, high-definition television and real-time games, can promote the wide application of 10G Ethernet technology and promote the healthy and orderly development of network.
Technology features
Gigabit Ethernet technology provides richer bandwidth and processing power, effectively saving users ' investment on the link, and maintaining Ethernet's consistent compatibility, ease-of-use, and easy-to-upgrade features. However, we also see that because Gigabit Ethernet is still in the early stages of development, there are still some problems and shortcomings: first of all, in terms of price, the current price of a 10GE port is about 100 times times the GE port, especially if the bandwidth is not fully utilized, it will cause a great waste of investment; Gigabit Ethernet inherits the characteristics of Ethernet's consistent weak QoS, how to carry out a guaranteed differentiated service bearer problem is still not resolved, RPR, MPLS and other characteristics of the support is not yet mature; again, 10GE requires the equipment to have a strong processing power, At present, some manufacturers in the industry to launch the 10GE port and do not reach the real line-speed processing, bandwidth advantages greatly discounted.
In view of the above problems and the current network bandwidth demand is not very urgent situation, it is recommended that network construction focus on business and cost-effective, the network core still use 2.5GPOS interface or GE trunk mode, when Gigabit Ethernet in technology and cost has made significant progress, and then smooth upgrade to million.
Technology Advantage
One of the advantages of Gigabit Ethernet is that it can use the same form factor as Gigabit Ethernet to get 10 times times the bandwidth. This applies to both the switch port and the host adapter port. For example, on a host server, if you need eight Gigabit Ethernet NIC interfaces, you can use two four-port network cards, four dual-port network cards, or eight single-socket NICs. This consumes two, four, or eight server I/O slots and causes eight cables to come out of the server. There are situations where you can do this, but not all servers have so many slots, and some server I/O slots need to be used for other purposes, such as RAID controllers, Fibre Channel host bus adapters (HBA cards), PCI Express (PCIe) SSDs, and so on. If you are using a Gigabit Ethernet port with a dual-port Gigabit Ethernet card, one port can provide bandwidth of up to eight Gigabit Ethernet interfaces, with one remaining for failover or other management purposes. This only takes up an I/O slot, frees the I/O slots, and reduces power consumption. And no more than two cables are needed. In addition, the same dual-port Gigabit Ethernet card can be installed on smaller servers, such as a 1U high server with only two slots, and still provide sufficient network bandwidth. [2]
Development process
in domestic network manufacturers, Huawei pioneered the high-end routers and switches that support gigabit Quidway S8500 (8505/8512), located at the core of carrier-grade operation core network aggregation layer, campus network and Enterprise network. Quidway S8500 million Gigabit core switches with large capacity, rich business interface features, protocol support complete features, backplane capacity of 1.2T, Exchange capacity 480Gbps, Ethernet interface to provide the largest 12个万 Gigabit Ethernet interface, and has a strong VPN support capabilities and perfect QoS capabilities. The Quidway NetEngine5000 series Gigabit Core router is a high-end network product for carrier-grade operation Core Network, which adopts the distributed architecture of three-dimensional switching network, each interface module comes with a distributed switching network, which can be easily stacked and extended, and the maximum 560 interface modules are provided. The machine provides 11.2Tbps switching capability, maximum port capacity of 5.6T, supports IP/MPLS line-rate forwarding for 10G POS, 10GE LAN, 10GE WAN interface, and supports smooth scaling to higher speed interfaces. With a three-dimensional architecture, the Quidway NetEngine 50 million Gigabit core routers have the advantage of scalability, load balancing capability, multipath Backup, and non-blocking, and have incremental scalability to increasemillion Exchange capacity, instead of a one-time high-configuration centralized switching network, to meet the future core network development characteristics and needs.
In addition, Huawei's fifth generation high-end core router Quidway NETENGINE80/40 also has the ability to smoothly upgrade to million. The launch of the Quidway series of gigabit routers and switches indicates that the design technology of our large capacity core routers and Ethernet switches has entered the international first-class level, which is not only a major breakthrough in the development of China's core network communication technology, but also a major breakthrough in the internationalization of China's data communication industry. And it will provide more impetus for the further development of informatization in China.Glossary
10GE---Ultra long-distance Gigabit Ethernet