The earthly thing people often unexpected, as the optical transmission medium "ether" has proven to be non-existent, but should be named "Ethernet" has been thrive. Ethernet from a variety of network transmission technology all the way through, not only into the mainstream, but the speed has made substantial progress.
Now watch the 1970s movie, which refers to the technology is unsatisfactory, but pregnant with a lot of technology initial. Ethernet therein. Although it was born only 3 Mbps, this rate now seems like snails pace, but at the time, although "data transfer speed than the computer to access the main memory is slower, but enough. Bob (Bob Metcalfe), Matt Corfe and other Xerox Palo Alto Research Center (Xerox) (Palo Alto Research Center, PARC) employees in the memories of the time.
From 3Mbps to the highest standards IEEE 802.3ba support 40/100G speed, room to grow to 10000. Let us ignore in this process, and take a look at todays sudden emergence of 40G network. Now, 40G began mass production the first users began to test the water of these products, but a lot of the deployment has not yet begun. 40G Ethernet now and in the near future, the user will have a profound significance. This schedule 10G Ethernet network card integrated into the server, unlimited eagerly pulled in front. From the device point of view, the core switches, high-density 40GE veneer (at least 20 * 40GE) is a general trend, especially in the field of data center switches, 40GE time has come, the Huawei Technologies experts said.
More and more interesting 40G Ethernet market
40G Ethernet from the beginning of the birth is quite interesting, Fast, Gigabit to 10 Gigabit Ethernet to 10-step upgrade, which is four times that of 10G Ethernet and 100G Ethernet standard introduced. This seems to be to network vendors and users a choice: the end is 40G or 100G? Which at the time is a problem, is still an issue in the next few years. From shortly before the online world do 100G markets and applications Investigation 100G applications is still in its infancy, just beginning to emerge in the field of operators, practical application in the enterprise network or a spark, yet instance . And this corresponds to the 40G network market burgeoned. From the current market situation, Extreme, and Dell can be said to account for the day, but other vendors release and no later. Cisco, Juniper Networks, Alcatel-Lucent, H3C, Rui Jie manufacturers are gradually released products, the emerging vendors Arista and Mellanox solutions are also distinctive features. Recent News Huawei published in the just-concluded Interop 2012 on the CloudEngine series data centers switches, dramatic performance parameters. Selected before 100G Ethernet as the main object of manufacturers have already launched their own 40G Ethernet products, and some have not yet released 40G manufacturers seems to have started brewing. From divergence to the recognition, 40G Ethernet development can be described as a rather convoluted.
Another interesting thing is the 40G Ethernet market data released at the end of last year, from the list view the list, is also quite interesting. 40G Ethernet market is not a market reshuffle new opportunities? I am afraid that a lot of peoples hearts in the drums. According to DellOro released in December last year, the first 40G market report shows that, in the third quarter of 2011, Extreme accounted for the largest share of the modular 40G Ethernet switch market. Suppliers, only Dell and Extreme in the third quarter of the income of the modular 40G products. Of course, sooner or later Date of the points, the market has not yet grown is very important reason for the formation of this data, it is insufficient to show the overall. But this figure makes a great space for reverie.
In addition, data from IDCs 2011 Global Ethernet switch revenue to reach $ 21 billion fourth-quarter earnings of $ 6.006 billion, an increase of 5.5%, growth of 1.2%. 10G Ethernet market sales are very brisk, year-on-year increase of 26.5%, while the same period in port shipments grew more than doubled. However, more rapid growth is 40G Ethernet port 12,000 in the fourth quarter of 2011. Therefore, 40G, although the actual number is not dominant, but the growth curve is very moving. It is envisioned that the data center switch vendors will increase in the near future, 40G Ethernet market development efforts.
40G Ethernet, which is a manufacturers do not want to miss the feast.
Inventory 40G Ethernet products
Stand this point in time, 40G switches longer fresh. That product inventory where to start? It is better to start from the most widely mentioned in the last year, DellOro report Extreme.
Extreme 40G Ethernet products, including core and aggregation layers. The products of the core layer is the BlackDiamond 8900 series and BlackDiamond X8 product, the product of the convergence layer Summit X670 series and Summit X650 series. The main event is a key recommendation in the Interop 2011 on the BlackDiamond X8 and Summit X670 series. The BlackDiamond X8 can support 192 wire-speed 40GbE ports or 768 wire-speed 10GbE ports. Can provide 1.28 Tbps per slot bandwidth per switch 20 + Tbps switching capacity. The Summit X670 series consists of two models: Summit the X670V and Summit X670. Summit X670V to configure optional VIM-40G4X by extension modules, can support an additional four 40G QSFP + interfaces. Use this expansion module, you can use up to four 40G ports on the connected server port connection 48.
Dell (Force10) win the the 40G market products mainly Z9000 and S4810. Z9000 products can support 32-port 40 GbE QSFP +, or 128-port 10 GbE SFP + with the branch cable. S4810 This product is a 48-port 10GbE products, while providing four 40GbE uplink ports. Well worth mentioning that the Z9000 40G port device maximum power is less than 800W. In addition S4810 10G port maximum power of 220W. Emphasis on energy efficiency, energy saving and emission reduction in the data center today, it is particularly meaningful.
Cisco as important manufacturers in the industry, released its 40G Ethernet is not high. In fact, until the first quarter of this year, Cisco was announced extended switch 40/100G function. Catalyst 6500 switch series new Catalyst 6900 Series 40G Ethernet interface module, the module allows the switch to support 44 40G ports. Nexus 7000 provides two modules: two-port 100G Ethernet card, as well as 32 non-blocking 100G Ethernet port. In addition, the two-port 40G Ethernet module can switch provides 96 non-blocking 40G Ethernet ports. Currently, these upgrades still be listed in the. Although the time-to-market is not high, but the Cisco program is very comprehensive, and the whole package includes not only switches, and cabling products, network card products. Cisco UCS server M3 version also includes a 40G card. In addition, the Nexus 3016 as a location for the super delay switch, it also has 16 40G ports.
Network powerhouse Alcatel-Lucent 40G release, although there is no upper hand, but good gains. "Now many of the worlds countries and industries have begun to try our 40G 40G market as planned at the end of 2011, completed the Interop data center acquired in 2011 winning innovative architecture 40G high-speed parallel interconnects POD blueprint to achieve a our commitment to the market and the industry. "Alcatel-Lucents data center and cloud network Asia Pacific market strategy and market development director Zhang Yufeng, put it this way:" Alcatel-Lucents products mainly include OS-HNI-U6 and OS-board, QNI-U3 two OS-HNI-U6 is a mixed board supports 2 x 40g + 4 x 10G or 1G (10G/1G, depending on the users to choose Tranciever type); OS-QNI-U3 is pure 40G sheet card, support 3 x 40G. "
As a network Rookie Arista in early 2011, also launched a 40G series: DCS 7050S-64, of DCS The 7050T-64 and DCS7050Q-16. DCS 7050S-64 has four 40GbE QSFP + ports and 48 10GbE SFP + ports can be used as the backbone of small and medium-sized data center (Spine) or access (Leaf) switch. DCS 7050T-64 has four 40GbE QSFP + ports and 48 1G/10GBase-T ports can be used as small and medium-sized the data center 1/10GbEBase-T server access (Leaf) switch. DCS7050Q-16 to 16 40GbE QSFP + ports and eight 10GbE SFP + ports can be used as the backbone of small and medium-sized data center (Spine) switches.
In addition, leaving 40G interface for future upgrade and expansion to allow enough space is the practice of many manufacturers. Currently, H3C S10500, S7500E, S5820V2 products support 40G interfaces. Follow-up to 2013, H3C product line of more support 40G, 100G port. "The H3C solutions Deputy Minister of Lu Guoqiang said.
Ruijie 40G released in January 2011, mainly RG-S12000 series and S6200 Series RG-S12000 10 slots and six slots as an option (S12010 and S12006). RG-S6200-48XS curing 32 Gigabit SFP + optical interface contains an expansion slot. An expansion slot can accommodate up to 16 SFP + 10 Gigabit module or 4 40G module. The machine can support 48 Gigabit or 32 Gigabit + 4-port 40G wire-speed.
Mellanox InfiniBand turn into the Ethernet. 40G problem, "Mellanox provides a complete solution, including chips, network cards, switches, cables and software products." Combo to the Mellanox Technical Director Song Qingchun package of solutions for the Mellanox very confident: "Mellanox has been deployed around the world a large number of high-speed InfiniBand large systems, the accumulation of a large number of deployment experience, 40GE Ethernet deployment for Mellanox will be more relaxed. "Mellanox has shipped NIC the single port PCIe3.0 40GE the NIC MCX313A-BCBT and dual-port PCIe3 .0 40GE NIC MCX314A-BCBT. Switch products provide 36-port 40GE switch SX1036 48-port 10GE and 12-port 40GE switch SX1024. The issue of why is there so many categories of products, Song Qingchun pointed out: "40GE four 10G channels to transfer data at the same time, how to avoid crosstalk of the four transmission channels to ensure the quality of the signal transmission from the card, cable, switches, and software level guarantee. "
Juniper Networks QFabric the time of launch, has caused uproar in the industry layer network designed to bring inspiration to many people. QFX3500 QFabric strategy released at the same time is a 48-port 10G ToR switches, 4 40G uplink ports.
Huawei CloudEngine series switches for release this month won a lot of praise in the Interop 2012 on. The CloudEngine series includes CE12800 core switch, CE6800 and CE5800 top-of-rack switch. On which the CloudEngine 12800 Series core switch machine support 48Tbps exchange capacity, support the 2 Tbit per slot bandwidth. Entire series support network virtualization, to support compute, storage, and network integration.
Outside the port upgrades and breakthroughs
Product comparison purposes, the port density is one of the most easily seen by comparison of items, some features are not obvious, but it is very important. With the development of the network, network integration and simplify become a trend. Compared to 10G Ethernet, 40G Ethernet in these areas is clearly a breakthrough.
"Compared with the traditional networking products, BlackDiamond X8 is not just in the port density and switching performance with a very big breakthrough in fact, the BlackDiamond X8 breakthrough innovation and optimization of the architecture, such as the BlackDiamond X8 is not back board, each line card to be exchanged directly and the matrix, which makes very fast, very low latency. "Extreme Networks network consultant Sue and say. In addition, the BlackDiamond X8 on a new technical support also includes storage fusion technology, such as Data Center Bridging (DCB), to achieve LAN-SAN fusion for iSCSI, NFS, CIFS, and supports FSB FCoE. The modular ExtremeXOS operating system to ensure the flexible and intelligent virtualization capabilities.
In fact, DCB, TRILL, SPB, and FCoE, 802.Qbg, and other features not a single patent, but some being drafted by network standards. Application requirements of the network is just one of a wide range of rate. Of course, it is a requirement that the most easily recognizable. Virtualization, cloud computing trend to simplification and integration of the network a higher demand, new network architectures are emerging, but open, large two-story, network integration is a basic consensus. I wrote a "networks crowded structure of the future", an inventory of the architecture of the mainstream manufacturers, not repeat them here.
These new network features and 40G are not necessarily linked, but it is an integral part of 40G and 100G product comparison options, because these characteristics are often related with cloud computing and data center.
Deployment of 40G: those things you have to consider
The 40G network devices has been presented, but the actual selection and deployment process, you want to build a network need to consider a lot. Example, in the face of 10G, 40G and 100G networks How should I choose? Another example, should also be considered in the actual deployment process supporting systems, wiring, power supply and cabinet with which to match.
What users will choose the 40G network? In fact, "The most important factor is the actual user network business needs, select the appropriate interface bandwidth according to the current traffic situation. Should give full consideration to the future of network upgrading and expansion of capacity." Lu Guoqiang answer from the present and future perspective explained. What means as defined criteria?
"Generally, we see the Gigabit port density, enhance the access layer Gigabit port density, the core layer 40G connection is the inevitable trend of development of the data center." The Ruijie Co., Ltd. Product Marketing data center product manager Jiang wave say. Arista international sales vice president Mark FOSS from several aspects gives a more detailed answer: "First of all, depending on the demand for network bandwidth; Second, we must consider the actual networking equipment interconnection distance less than 150 meters consider 40GbE/100GbE interconnect distance is more than 150 meters, it is still mainly consider adopting 10GbE; once again to consider the flexibility of the wiring for easy wiring, simple, we can consider the 10, 40 or 100GbE, we must also consider interconnect twisted-pair, dual coaxial cable or optical fiber. To connect to the server using existing twisted-pair cabling system, can consider adopting 10GbE first. Finally, port density, and the average investment in each port can not be ignored. . "
Mentioned in Marks summary wiring is not the same as the recommendations are given for different distances and flexibility requirements, which is what causes it? Wiring the field of professional manufacturers have the final say. "10G Ethernet fiber and copper transmission standard has been released, but 40G Ethernet transmission purposes only fiber standard copper standard." Simon Technical Manager Chen Yutong said. From the perspective of the development of the wiring, in general, the copper standard is always behind the application of standards and fiber standards. 10G history is the case. International Organization for wiring TIA and ISO have set up a Working Group on copper standard. The name of the TIA TIA 568-C.2-1, ISO name is ISO 11801-99-X, TIA standard mentioned is 100 ohms to support the next generation of cabling specifications, but does not explicitly 40G or 100G, and ISO clear in 40G transmission areas. Mentioned several times in the previous QSFP + is actually a short distance transmission. In general, it is used to connect the ToR switches, distance is no more than one cabinet. QSFP + and SFP + just to make up the twisted pair supports vacuum. Chen Yutong think so. Of course, some people think that the future QSFP + and SFP + will replace the existing twisted-pair wiring. QSFP + and SFP + shipping gradually decreased, and twisted pair gradual increase in 2013, the QSFP + and SFP + and 10GBase-T will be basically the same as, long ago, Intel released a forecast.
Since the wiring that is the case, the user should make a choice? Short distance can choose QSFP + and SFP +, can be considered long-distance optical fiber. Standard supports 100 m multimode fiber OM3, OM4 standard is 150 meters, you will not encounter single-mode fiber transmission distance. "Chen Yutong such a suggestion. However, for high-density, high-performance, long-distance, high-speed transmission requirements, MPO / MTP will is more reasonable choice.
Wiring problems can not be ignored, the more important the upgrade of physical infrastructure. Clearly, data center power and cooling system reliability is the normal operation of the premise of the entire IT system. "In a lot of low-density data center, a single cabinet power density is generally lower than 3KW, device applications in high-density rack power density has been increased to 5 to 8KW With the popularity of virtualization, power and even some of the data center room can reach 30 to 50KW. example: Supercomputer Center in Tianjin Milky Way One single cabinet power up to 50KW, rely on traditional raised floor air supply, physical infrastructure obviously can not meet the demand. "Schneider Electric strategic markets in Greater China Xie Weigang, director of digital simple and direct that the present problem. The level of air, close to the heat source cooling, hot aisle containment system for high-density deployment in low density in the engine room, to abandon the traditional raised floor is a practical solution. "This is from the data center physical infrastructure vendors pertinent suggestions. Efficient, green energy has become an important indicator of data center solutions.
Talking to go from a case
What customers start using 40G? Dell Force10 is currently in the global and the successful deployment of a large number of cases, mainly used in Web2.0, high-performance computing, cloud computing center operators in high-density data exchange and other industrial applications. "Senior marketing manager of Dell Greater China network Yaoxin Hao say so. In fact, Dell is not a special case, Web 2.0, high performance computing, cloud computing center operators in high-density data exchange these user groups appear in the user list of all manufacturers.
We might as well be talking about an actual case: in Nanjing, China Petrochemical Petroleum Geophysical Research Institute Network. In the new expansion of the network, it uses three Extremes BlackDiamond x8, bundled with two 40G link between these three devices, the Internet bandwidth 80Gbps. And other cluster bundled with multiple 10G, and relatively large cluster with eight 10G bundle. Therefore, the large link between nodes interconnected bandwidth 80Gbps. With the growing demand of users similar SGRI original network there are a variety of products of various manufacturers to do the expansion, the first need to face is how to integrate over the network, storage, and so to the same architecture up, which is often also a network optimization process. "In actual operation, we found three X8 is very similar to the structure of each of the above are calculated, stored and managed nodes. Thus, we consider the resource integration, calculation of the resources to be moved together to form a computing cloud. In addition to providing some of the integration of shared storage to multiple clusters together to form a storage cloud. Such benefits are reasonable allocation of resources, do not affect each other, a whole cloud solutions. "Su said.
In this case, there are several areas of concern. First of all, the original network expansion, interoperability between different brands need the whole program. Second, the expansion is not a simple extension of the linear, often accompanied by the optimization of the network structure. This part of the optimization function network is more important.
The development of network transmission can be described as the sum of thousands of miles, a uniform Heaven, with each passing day in the development. Behind its business needs evolve, it is the aspiration of the people more efficient life. 40G networks will be a milestone, as the 100M, 1000M and 10G. However, along with a milestone determine IT to bring to life but it is more and more big surprise.
