Cloud Architecture: Expand Cloud vision, landing large data

Dong Xin: Ladies and gentlemen, good afternoon, welcome to the Hello Cloud architecture of the sub-forum, I am sure that many of you have participated in the morning meeting this afternoon in the Cloud Architecture forum, where they sat together to discuss how to put the cloud's vision, how to put the big data to the final landing, how to make our calculation , our storage, our network, can be better integrated and played in the era of cloud application architecture change. So we are also very happy today to invite experts in the industry, corporate executives, including our customers and our partners, so here I also on behalf of the super cloud and marketing companies to express our heartfelt thanks.

The first speaker we invited today was from Wei, and we welcome you to the podium.

Wei: As one of the hosts, we welcome you to join our sub-forum. Almost two years ago, I once attended an internet conference, just met a few media friends interviewed, they asked such a question, "Now everyone is reflected, such as Google's data center, power consumption is very strong, maybe a Google, or the United States it manufacturers, using the United States 12% of electricity, Everyone also said that every time you click, it seems to cost 0.1 degrees of electricity so a probability, ask me about this point of view. I understand their starting point, we are to enlarge the power of the cloud, so that we feel that the cloud has brought us a lot of additional energy consumption, then I said So, If you google click once is really 0.1 degrees of electricity, but there is no Google this thing, you may have to drive the car to the library, you come back, might consume more energy, so I'm going to go straight to this example, with the advent of the cloud, the scale of the data center is getting bigger and larger, we have a big increase in the level of attention to energy than in the past, so we super cloud from the beginning of the venture to the present, such as Where our products are worth paying attention to, reduce our customer's comprehensive cost of ownership, whether from the product technology framework, and comprehensive cost-effective, we have been hoping to get the industry leading position in this respect.

So based on the cloud aspect, the hardware changes, from many aspects, from the performance, from the comprehensive energy consumption, the user's overall life cycle to consider our product design.

After more than two years of development, we gradually put this concept into our ongoing new product launch, from our first generation of products from the offline, to last year based on integrated solutions, including this year based on the launch of the Intel G9 platform products, including the introduction of high temperature integration machine, But I think through this series of launch, to achieve, how we give our users to reduce the overall cost of such a concept.

We have also made a simple calculation of what the high density and high temperature servers mean for a large data center. We can take a look, we can save 80% of the area, this is actually a considerable savings, if the conversion to money is very considerable. The general server needs to work in the 22 degree environment, if we launch 45 degrees of the machine, basically air-conditioning refrigeration power can save more than 70%.

Today there are many experts, in their respective fields will have a very detailed exposition, I simply made an opening statement, but also to publicize our Super Cloud company, also hope to spend a warm afternoon here. Thank you!

Dong Xin: Thank you, Wei, we're officially in the clouds today. Jointly released with Intel based on a new generation of hyper-cloud blade system, as well as products for high-temperature environment, the booth outside the physical, thank you very much in this hot weather to see our high temperature machine how 72 hours to run up.

When it comes to the entire server industry, including the cloud computing industry, Intel has been a leader and catalyst in the industry, and we are today invited to the Intel China Marketing general manager to share Intel's ideas and practices in the future cloud, thanks! Douglas.

Douglas: Good morning! If you don't mind, I can speak English. English）

Bin: I'm from Intel software division, Intel Worldwide he has a strong software department in the back, in fact, many people for Intel's impression is a hardware company, in fact, Intel in the world should have about 30,000 people to do the software part of the work, especially in the score said OpenStack accounted for a certain proportion, There are cloud computing areas, we have our own products, big data We also have our own products, so Intel is investing a lot of resources in this area.

Today I bring to you a share, we in the background of large data and cloud computing, how we build a scalable storage architecture, I may be from the software to introduce.

Storage and cloud services, in fact, we now see the cloud computing, from the concept of the words, we mainly divided into such a few major categories, one is Iaas,paas,saas, cloud computing itself is only a conceptual thing, and for this concept of classification, it is in fact according to different directions to the points, Divided into virtualized cloud computing, can also be divided into distributed systems of cloud computing framework, are different categories, today from these three directions to look at, in different contexts stored in the use of the scene.

The IaaS scenario is primarily virtualized cloud computing, with the main focus on your data, how to better store in a storage solution, many people are using centralized storage to exist your data, centralized storage allows different physical host to access to your current virtual machine, you can do some advanced features.

For PAAs, we are more focused on how to provide customers, the vast number of developers with a layer of storage API, to enable the vast number of developers to create their own platform based on the API.

The other is SaaS, there are very detailed user-related behavior, such as how to actively back up your data, how to store some scraps of paper, such as mail, video and so on.

For storage in this area, the requirement for persistence is exponential growth, 2020 data, and storage of the entire persistent data growth should be at 40ZB levels. On the other hand, its input is also constantly increasing. Intel solid state drives, which cost very much as early as two years ago, if you buy an SSD, the cost is very high, but from this year and these years trends, its price is gradually reduced, and if you look at TCO, it is actually more than your traditional hard drive to save costs, Because its performance is greatly improved.

In this context, how can we solve this dilemma, on the one hand, the data you're storing is exploding, and if you're storing your data in a traditional way, like using your traditional database, or traditional centralized storage, that's going to cause some problems. Although you can use some of the new technology, you can add a few of your network card, but the future certainly has a capacity to meet the requirements of the Tun, if you increase the amount of data, the above to do a query, the time spent will be very much.

There is no way to satisfy your business's need for storage resilience, as it is likely that there will be a burst of growth for the time being, and if you use centralized storage, you can think about how difficult this is. This is a very difficult thing to do.

In addition, for example, centralized storage, now that you have this RAID solution, but you will find it difficult, you have to spend a lot of effort to restore your data. From these points, we have a very strong demand, how to build a scalable storage, can improve our throughput, this is the industry's development trend.

How we do this and how our team supports it. What are the characteristics of distributed storage, compared to your centralized storage, you can see that on the right, for your data access, you can't see how many clusters you have under your service, and distributed storage can provide high reliability and scalability. When we talk to customers, the typical scene, the original traffic, the highway has a system, each time the car has how high, the flow of how much, if you use the traditional typical database analysis, you may only save one year of data in the inside, if you are trading and analysis in the same place to do, do not do, The only solution is to separate this type of transaction from the analytical one, which we can use distributed systems to deal with, the benefits of which can be extended, and the amount of data stored in it will be very high. The category has a distributed caching system, as well as distributed file systems, NOSQL, and finally distributed block devices, which is also a hot point. But all the scenarios, whichever you choose, end up in the business, how the specific business is done, and according to your business, what is the actual application scenario that you ultimately need.

Because each business has different requirements for its storage, such as the application of your Web tier, it is very focused on your page. What I've just mentioned is that it's important to have so much virtualization and so many deployments of virtual machines and how to find the right storage solution for your business.

Distributed storage applications, from the perspective of application, how do we put this storage solution to a division, for different business areas, or application areas, we have different solutions can be provided.

First-stream data processing, it contains a lot of OLTP business, it is very high real-time requirements. The single pick up to do analysis, for this kind of business, it is more important to analyze the ability and the whole framework can expand capacity. Finally, data persistence services, we can use the existing solution to create a centralized storage, a large number of data dispersed to different nodes to do storage.

For streaming data, it needs the first to be distributed, and secondly it needs the technology of memory caching, because it requires a high real-time performance. So we consider using this memory caching technology to speed up its processing power. Because its nodes are distributed, you can imagine that in the case of high-speed computing, the interaction between the nodes is very frequent, and it requires a very high rate, in order to ensure this rate, the network requirements are very high. In this regard, Intel also has a lot of hardware technology to help everyone, for example, for streaming, codec this side, we will have (Mike) chip. In addition to the network side, Intel has 10G network card, you can do based on gigabit network distributed system to accelerate your performance.

Data analysis is not much said, this side is mainly for the current hottest Hadoop, we have Hadoop products, in addition to Intel has a holistic solution.

Finally, the persistent service, the persistent service of the data, which basically says the background distributed system, or the distributed block device, how to provide a better, higher IO performance requirement, we can still use the 10G network, because he is a distributed network, 10G networks can be used to enhance throughput between distributed file systems, especially if you are using the process, not all of the data can be replaced to SSD, because considering the cost, we can replace the SSD with a part of Techi, increase your throughput. You will find that when you do the overall frame upgrade, there are some bottlenecks that you have not seen before.

There is actually a page that describes how our team can support just introduced so many classifications, so many solutions, in fact, our team's main task is to be able to put these solutions in our laboratory to do a validation, can be a customer, for partners to do some solutions, Can enjoy our LSB solution, we go to the market together. Thank you!

Dong Xin: Thanks to Intel's colleagues, who helped us share Intel's vision for the future cloud, and on the other hand shared the trend of computing, storage, and network convergence, and how Intel, as a traditional chip vendor, is moving forward with computing, storage, and network convergence. The next guest is the Qihoo 360 Tanxiaosheng from the Internet industry to share with us how Qihoo 360 embraces the cloud architecture.

Tanxiaosheng: That's what Intel two friends said, we almost all have comprehensive cooperation, from IDC, we also have the use of hyper-cloud production of servers, but also a large number of Intel SSD disk, Gigabit Network card this year used more than 100 pieces, so far the server is also Intel CTO. Over the cloud side, we use more servers, in addition to the new blade we have not used, other models, half we are useful.

Now, as a 360, it's safe to say, 360 of the idea is to make Internet users safe online, here, I do not know how many people have been so blackmailed, this is a typical QQ chat record of our customers, someone to say that your site was attacked by me, 200 dollars I will stop. who also jiabuzhu a week to blackmail many times, random alarm. For these situations, internet companies, or webmaster, think I am safe, he felt I have a framework, I application server are open source, these things are not safe, the actual security is just an illusion, in fact, there are a series of loopholes, Real Internet site you are living in a very insecure environment.

This is the case of network security incidents, most of the security attacks are not initiated by experts, but the script boy and so on, he downloaded what tools, using these tools to scan, do automated scanning, resulting in attacks. A small part of the second category is the commercial hacker, the level is higher, by attacking other people's website to profit. This is the most difficult, national security level, launched this year, the flames such a virus, in fact, behind the national power to do, and for certain political or military targets, to the apex of the attack, the general business site will not have such good luck. So for commercial websites, the main to prevent the front two.

What is the solution? Let's take a look, this is our 360 Web site scan swept out of the site tampering with the data, you can see, this is this year, from January to November, the site has been tampered with the situation gradually increased, your site if the tampering you will find it, in fact, the purpose of the site tampering like the virus spread, Its purpose is not to be found before he gets the benefit, he wants to get benefits such as hanging in the black chain, this year the government's website and education site is the most tampered with, tampered with in the inside hanging what the lottery and so on ads, hanging these black chain, only on the search engine has an impact, Search engine searched a look is the local government's website feel the Trust degree is higher, just point over, take to the fishing website, thereby profiting, the government to look at their own pages of things are normal, is such a situation.

We scan the security rating ratio to see that red is a high-risk 31%, 8%, Low-risk 37%, high-risk is 39%, this is China's current Internet site security overall situation, 31% is high-risk, do not think that a website, users see something on the Web is safe.

What is the solution, the traditional solution is to buy a bunch of boxes, buy a firewall, buy APS, buy IBS, hope that through these can work, in fact, the first is very expensive, a firewall is more than 100,000, because he sold you that box, update relatively slow, many updates in order to prevent problems, Often, updates are made more slowly by the administrator.

The real problem is not enough to buy a box, but to hire a bunch of security engineers. Last year, after the Csdn incident, the security workers were given a benefit and their wages soared, and I paid my safety division at least twice the year. Now that we have a security engineer, these engineers will have to keep an eye on the discovery of the leak, which is basically a new loophole in almost every week. If you have a security engineer, they have to assess how big the vulnerability is and do some action. For a company like 360, it's OK, I spent almost more than 10 people to maintain the safety of the whole station. But for example, to do the site, the scale is not very large, unlike the size of Jingdong, in fact, you have to maintain such a team, the burden is very large.

360 for such things, we give the webmaster of each website, or enterprises to provide a safe solution to the site, the first step is cloud monitoring, you have a station, the first to do a physical examination, we have a scanner, you can place time to find that your site is not a loophole, is not hung horse, is not been tampered with. Second, all of our scanners in China, the third party vulnerabilities, the most timely, in addition to our own mining loopholes, but also with the domestic collection of loopholes have cooperation, there are the latest loopholes we will update in time, we will spend money to buy these loopholes back, but users do not need to pay us. Our site security scan, registered Web site has millions of sites, daily use also has hundreds of thousands of. This service is completely free of charge.

What do you do when you have a leak? In fact, swept out to find that they have a loophole, for most sites do not actually have the ability to change, at this time we provide automated protection, 360 This product is based on the cloud mechanism of an application firewall, all user input in my here to filter again, After filtering the clean request to the website, even if your website has the flaw also may the relative security. The second is that the most difficult thing to deal with now is the Dos attack, which is very difficult to prevent, General website is difficult to prevent is, in addition to have more than technology is to spell, physical strength is you need to have enough loans, because Dos attack, for example, you opened a shop, people find a pile of men in your doorway, the real guests can not enter, To solve this problem, my shop is big enough, the door is big enough, you come to the people are not strict, customers can always come in. For this we have more than 200G of band, also rely on 360 itself we use a lot of loans, we can do the loan reuse. Also has the domain name resolution, we have done a domestic first-class formidable attack DNS, we can resist more than 10 million per second request the attack. This is against these attacks, we can prevent, this service is also free.

Here we can look at how it works, attackers will first hit my DNS, which is equivalent to your home street signs, if the attacker through DNS found this site's IP address, he came to play, I listed four, the actual current domestic there are more than 10, he will hit my protection cluster, At the moment I have encountered up to two nodes can hit me, if 10 node killed one to two is no problem. When visitors visit the site, also through this node, and finally back to the user site, if the access node killed, the user automatically wraps around the other node back.

This is the last week we met six times this week traffic sexual attack, this is a Acura, he to we carried on for one months of attack, generally play half an hour dozen not dead on the retreat, but this traffic is not big, also can not kill me, consecutive dozen one months.

This is number 10th, we went through a 3 G-flow attack, unharmed. Our specialty has the scanner, has the external firewall, has the attack dna,360 flow cleaning, I and the third party cooperation.

This picture is very interesting, this is what we started to do, this is an aspiration to attack, and through this analysis I can know where the attack originated and where it was played, because I am the equivalent of opening a peace hotel, and many of the stations behind me, his attackers may be the same attacker. I can do more analysis between the attacker and the attacked website, and the future will help to catch people, through this analysis and public security cooperation, caught the real launch of a Dos attack source.

Dong Xin: We've explored the entire cloud computing architecture, including the security architecture, and the next step is to shift the focus from the computing architecture to another topic, the data center, and the next speaker is Mr. Don Beaty. His company has designed all of Google's data centers in North America to help Google achieve its positioning in the cloud market, and the next guest is Don Beaty.

Don Beaty: (English).

Moderator: The next 35 minutes, we have a few guests we invited to do a round table, talk about the modular data center, I am the new CEO of the World Cloud box, we mainly do container data center this piece. Today, our forum discusses the time and prospect of a modular datacenter in the Chinese market, and the difficulties that will be encountered.

We will start this forum now, and we would like to ask each of you to give us a few minutes to describe the possibility of their development in the Chinese market and the obstacles they have encountered.

Cannan: Everybody good afternoon, today's weather is not too good, but we this room heat is very high, especially when I come up in the morning did not think, today Cloud base company to run the meeting so successful, we 9:30, basically is no position, are waiting outside. This afternoon I am very pleased to be invited to participate in our third forum discussion, the topic of this forum is and intensive, including our modular room related.

As we all know, at present, including China Unicom, several of our major operators, including our internet companies are speeding up the construction of data centers, mainly to meet the changes in the entire mobile internet, including the arrival of our big Data era. At present, the construction and operation of data center should be said to have undergone great changes, from the perspective of construction, scale, intensification is now a prominent development direction, including green energy conservation. So in the process of production of this base, modularity technology is very much used, including we are now in the development of high density data center construction standards, this model of construction is mostly based on modular solutions. And we're thinking, including just as our big customers, we are in the delivery of data center infrastructure for customers, we can not, as before, through what kind of cycle, we go to power matching, air-conditioning matching, and then integrated wiring, and then buy IT equipment, the entire cycle over, The link between the various links may not be very good. The second is that the delivery cycle is too long, customers may be directly out of tune, can not meet customer needs.

Some time ago we visited Europe, or some experience, I think the future can be in the modular construction will be applied to. The first is the foreign data center, its monomer design generally controlled between 10,000 square meters to 15,000 square meters. He will not blindly pursue a 100,000, plus production, the construction period is longer, the second combination of customer demand is not necessarily very close, or a little like the form of a car run, in the big city can be set up multiple points, in each point of land requisition area of almost 100 acres, he will plan two to three independent, 10,000 to 15,000 square meters of the building, One or two floors. One of the accessories density reached 12-kilowatt, such as the majority of the room will consider the use of modular mode, because they have to do customization for big customers, can not say that after the building, to provide direct to customers that impossible, small and medium-sized enterprises can, general, internet companies, but there are some higher requirements, Banks or large Internet company information centres may not be able to do so. So I think the foreign countries have taken the first step.

At present, we feel that telecommunications and our unicom are actively discussing the modular construction program, including distribution density, refrigeration mode, construction methods, maintenance methods, including delivery services, and so on, we next want further research, in our construction in the future more supporting cloud computing projects, Most will take into account the modular deployment approach, let me briefly introduce our overview.

Mr. Yao: Because I work in the data center for almost more than 20 years, in the domestic planning and design, I have three experience, do the data center now has three aspects, one is super large, 300.05 million square meters of data center. Also, we call it polarization, container-type, this is the European comparison of traditional methods, one is to take the container type, there is the module warehousing type. If from the warehousing, another is modular, I think all from the data center planning and design, must pay attention to a number of factors, one is how to achieve the reliability of the design, reliability is mainly rely on all distribution, HVAC, unit partition, as well as modular design, to reduce a point to all the damage. and the design of environmental protection, all power-saving can not be separated from Pue, the most effective way to reduce pue is to reduce air-conditioning, the northern region has only one of the best way, is natural cooling, or on the basis of air-conditioning, in the south in the air conditioning based on some efficient technology, such as we put 5 of the cup value, We use high voltage inverter compressors, energy efficiency can be to 9, magnetic levitation compressor can to 9.8 that High-tech, in the South region. There are now more popular, the north of the room in the production of so much heat, by some means of recycling back, as a bath water, heating water, can return about 59 degrees of water, a data center can take 30,000 square meters of office building all the heating and bathing, this is man-made pue down. There is no hydration, now a large data center, especially the cloud Computing data Center, a data center consumes almost 600 tons a day, this is only one day, in the north of the water shortage, to take into account the design of no hydration.

So I think in the future of China's data center changes, I personally think the future direction is the container-type or warehousing, I have participated in the two large warehousing, one is China Mobile in Harbin Warehouse type, Pue is currently the lowest in the domestic 1.35, last month did a domestic largest internet search company, Pue do 1.38, all do warehousing, which is today said modular concept. Container is easier, container in Europe is more popular, container inside can contain UPS, culvert almost generator, culvert air conditioner, also culvert cabinet, that is to say I take this box to the place, as long as there is oil to open the machine can be used, but this chassis almost eight cabinets, nine cabinets, but the density is very high. This is what I think may be the trend and direction of the data center in the future.

Tom Barton: (English).

Shangpeng: Good afternoon, everybody. People who do it may be unfamiliar to Schneider, the acquisition of the original in the IT field, etc and MGE, we set up an IT division, in fact, Schneider has always been the traditional data center physical facilities, the two years the entire market data center is very hot, As Mr. Kang and Yao always talked about, now the data center from the original layer of room, or thousands of square meters is a large data center, now whether the financial industry or telecommunications industry, or our internet companies, including IDC hosting company, we are now in a large number of thousands of square meters of data center. The original circle of colleagues said that the data center is a computer room, now we call a data center of the industrial park, now the construction of the difficulty and depth, completely with the previous is not the same. Especially now the construction of this kind of data center, it takes a long time like the 30,000 to 50,000 data center that Yao just said, usually the construction cycle needs to be 10-14 months, if combined with the prophase design and planning, including the project, the time is very long. The data center is the single most expensive building, it requires a lot of power and refrigeration stuff, and this engineering is so complex that something unprecedented before, of course, the current trend, including US Schneider, including ATC, we have been working on the product of the project, ATC originally had a call Infineon system, Integrate traditional ups, air conditioning including racks into one, reduced customer installation and commissioning time, now we are also vigorously developing modular physical facilities, not just as I mentioned earlier, it may be the entire power supply module, only a modular power supply, starting from the transformer, until the end of the PDU, and modular air-conditioning, so we reduce his construction cycle for customers, while the factory preset. This is the global trend, and such a method also greatly reduces the cost of customers, in this way, your data center on the civil requirements are not so high, the environmental requirements are not so high, it can completely put it modules, including air-conditioning modules, power modules, all in the factory ahead of time, to the scene is only installed assembly, And of course it combines a lot of cold passages, hot channel, so that the entire data center can be simplified, greatly reduce the cost of everyone, and because the data center so high input, we also want to be able to grow, while investment, if you make modular, you can according to your needs, and gradually grow, And just as you're running, you can also go to the factory to produce your next module, which is what we now Schneider has seen in the entire data center, including the future, and we're working on this trend to provide our customers with a comprehensive modular approach.

Moderator: Here we go to the stage of free discussion, I started, to mention a relatively focused problem, the benefits of modular data center, the industry has been discussed, most people are in favor of such a thing, just now Mr. Tom Barton said, in the United States over there module data center, the biggest benefit is the time, Save capital, but also according to the customer's new needs, trouble to meet demand, without the need to spend a lot of money. This is universally acknowledged. What are the biggest benefits of modular datacenter solutions in the Chinese market? Second, what is the biggest hurdle in the large-scale implementation of modular data centers in the Chinese market? I hope you will give your views.

Mr. Yao: I've just done a few projects, and I've probably calculated from a few aspects, one is from the investment point of view, I made two projects, one is 35 cabinets of the warehouse-type center, there is a 680, the cost is about 10,008 to 20,000, per kilowatt investment, one is more than 7 million, one is 4 million, This is the two projects that have just been completed, if the traditional do, stand-alone cabinets now do 140,000 or so, t3+ or t4-, distribution-type two-way, almost every cabinet spent 130,000 to 140,000, a cabinet of almost 4 kilowatts, 340,001 kw of investment, from the investment point of view can save the money. The second is the time cost, I do two projects customers give me 2.5 months, signed the contract after 2.5 months, the time is very short. Traditional data centers do it for almost 12-14 months, because you have a civil, land leveling, plus commissioning time, these two are a big change.

And from the operation, I have all the warehousing, because only to the cabinet-type, basically people walk not much, so, pue at present, according to European standards, the European requirements are 1.4, the highest can not break 1.4, the best to do 1.32. This situation is going to be lower than the traditional data center Pue, which is 1.6 to 1.8 better than today. The last one is operation, operation is human, warehouse-type words, all are factory, 10,000 square meters can become 10 warehouses, bad one, the entire storage is also surplus, a warehouse is broken I do not need to repair, since it is the cloud backstage I certainly do backup, so on the operation of the party's personnel requirements are relatively small. I think these are a difference.

Moderator: What are the main obstacles? If this is so good, why do we not use this plan now?

Cannan: I'll sort it out for you. Container Data Center This thing, we've been in touch since 09, and virtualization technology is moving forward together. But it has a natural barrier to what we call the traditional land utilization problem, and after we're done, if we're using a container, there's a shed and we can't do it without a shed. In the case of a shed, in principle two floors, in general, we from the land scale, we are now the data center of the modular, because there are many modules, there is power modularity, we say more from the overall data center room of the angle of the module, we are more respected, in our data center after the completion of the The modular deployment mode is used. For example, our current room, is generally a room of the volume, in principle, between 30,000 and 40,000, do not engage in the basement, because the cost of the basement is still very high, the use of a layer of power distribution, low pressure, high pressure these things, after finishing the first layer, we strive for another area to make a high density configuration, Our high voltage power supply is very efficient, not on the floor, but this area is not big. 234 floor In principle are the engine room, longitudinal can make a mouth shape, the middle is empty well, cut into four small, are independent power supply mode, deployment is also very convenient, we in the sales process, the operation is also very convenient.

The container can't be said to be optimistic, I think so, operator's point of view more, from the security, from the maintenance staff's point of view, after all, he is still a need for personnel to enter and leave, dust, temperature and humidity control, from the perspective of the cloud box, the cost of civil engineering, accounting for the entire data center investment package is not too high, We are also in the new cover of the two, the building, the entire decoration style or relatively simple and simple, in addition to steel concrete hard guys, the rest of the things we can simple simplicity, so we will not too much value civil. There is a just I said, you entered my Computer room, I can still be modular, in fact, all of those ways, but you add a skin outside, but removed the skin we took.

And we are now ready for modularity, the next step in the market competition is very fierce, we are the market competition, service implementation capacity of the promotion, delivery speed is very important, delivery quality is very important. When the client is on the previous project, it will not be like our state-owned enterprises, a planning department for two years, his last business decision chain will be very fast, to clock, at this time he asked IT support capacity, the speed of protection to be very fast, you can not say that this side of the decision is over, and then again to find data center, six months later to pay, The business opportunities are gone, especially the Internet from the mobile layer, a platform, an application system, soon this user was robbed, others want to enter, then this market user cost will be very high, and become not the leader, so this thing is a big factor, but this piece does not affect the modularity, Modularization with the birth of high-density, modular construction model must be changed. On demand, we are now mentioning the concept of We also need to deliver the user infrastructure on demand, can be divided into it, infrastructure, and so on, the cycle we are also compressed, our next IT technology company has a good communication, including Schneider, they are our major customers, we have to communicate next, Commitment to the customer delivery cycle is how much, what density, what standard. This is my opinion.

Hu Yin: I have a question to ask Schneider, Schneider more highly recommended modular data center standardized products, this aspect of what you think the future of this modular things are likely to be the most likely to be a global standardized products?

Shangpeng: The question that Hu asks is also the problem that the modular data center encounters, is the standard where, and in fact, many customers hesitate to use the module is also this point, I really how many cabinets is a module, in the end, how much electricity is a module, how much refrigeration is a module.

Now we can only say, like us Schneider in foreign countries may be a five-megawatt module, or a megawatt module, only in the mode of electricity to provide customers, because we mainly provide physical facilities, so we generally provide a megawatt of electricity, to the customer with air-conditioning modules. The future may be the entire industry, including the real end users, including our provider of solutions, we might want to plan together, of course I feel that in the future IT development, to cloud storage and cloud computing, the whole it is as a surface to provide, We do supporting the physical facilities of the manufacturers are in fact very easy according to the customer's final requirements to decide how big the module is. I think the future must be like this.

I think the last, most ideal modularity is it and these physical facilities are all in the factory, all the debugging, not like now, the electricity all done, put the cabinet to the scene, put it in accordance with his needs to plug in, I think the future trend, including our Schneider also think, The future must be we will closely with it vendors, it vendors to his entire number of surface, how much storage all done, and then do a whole modular solution.

Hu Yin: Do you have any questions or statements about modularity?

Guest: In China, first of all I am not very optimistic about the modular data center, culture is that you do not have two institutions, one is Cannan them, the other is the power company. For the power company, which is more difficult than the operator to deal with a place, he can give you the power to grant a good, or you pay, you have to pay him, the modular data center of the advantages of basically hedging. This advantage itself and the U.S. gap is not particularly big, the United States may be cheaper.

Hu Yin: Well, that's a very important point. One of the obstacles to the implementation of modularity in China may not be solved.

Then we end today's forum, thank you.

(Responsible editor: The good of the Legacy)