Peng: Perspective Cloud data Center hot issues

The 2nd Session of the "2012 China Data Center Industry Development Conference" in the 20,112-month 17th in Beijing, in the afternoon of cloud computing and flexible and efficient IT Infrastructure forum, China cloud computing expert committee cloud storage group leader Peng did "perspective Cloud data Center Hot issues" keynote speech. Peng that because of cloud computing's massive data processing capabilities, we can connect different clouds, the future of machine intelligence will be rapidly upgraded, machine intelligence will become an important symbol of social development. The challenge in the future is the challenge of human intelligence and machine intelligence differences, and people are too far away from machines. Peng that the hot issues of cloud computing mainly include several aspects, the following is the full text of the speech.

Peng: Well, first let's look at what the nature of cloud computing is. Cloud computing is able to develop rapidly, in fact it is behind the driving force, this drive is for the global data growth very fast, through the IDC statistics, the past more than 10 years, the total amount of data in the world every 18 months will be doubled, which is the weight of data, This means that all the data in human history will be greater than once in 18 months.

China cloud computing expert committee cloud storage group leader Peng

So fast, we can't solve it by Moore's law, but our data growth is going to be doubled by the cumulative sum of all the data 18 months ago. For example, Chinese mobile, China Unicom, Telecom, they are using minicomputer to deal with, but now do not need to deal with cloud computing is simply not possible, you use a minicomputer processing costs are too high to bear.

Then look at Taobao, if there is no cloud computing behind the support, we saw the double 11,2011 year November 11, in the day of the god Club, all netizens at 0 o'clock in the morning at the same time trading, trading volume is very amazing, if you do not have a good IDC facilities, you can not cope with this flow.

We know that 12306, our railway station booking site is not the use of cloud computing, what is the difference between cloud computing? Cloud computing is also the inevitable result of network development, we traditionally want to connect a large number of nodes through the network, constitute a supercomputer, we call cloud computing.

Google is called Google File System, the system will be able to reliably disperse the data stored in different nodes, each data at least three nodes have three different data, any two node loss does not matter, so that his room can be very soil, This machine is behind the ugly to avoid facing the corridor, because these machines are too easy to break, but a set of two is not OK, it will automatically restore data, make this appearance, that is, at any time the machine pulled down to calculate the new machine, at any time to manage.

This is the first set of cloud computing platforms, at first they didn't have the money to buy the machines, they bought a bunch of machines from the lab, they made the first Google platform, they used that technology, Google's cloud computing data centers were built in the wilderness, and Microsoft used the same approach, The construction model of a container with 200 servers and a cloud computing data center has been validated in the past few years and is considered possible.

Due to the rapid growth of the data center and the Yuncheng deplored network in 2010, electricity consumption soared, which had a great negative impact on the climate warming.

First of all, our little terminal becomes a supercomputer. For example, our GPS navigation is a fixed mode of navigation, if we put the city traffic and navigation results, through its navigation, it will not bring us to the road traffic jams. We think that there will be more functions of service, we all see a large number of free space for the network disk is growing, the server month will not be traffic jams, information readily available, this is the cloud to bring us changes.

The biggest change should be a huge increase in computational intelligence, due to the massive data processing capabilities of cloud computing, we can connect different clouds, the future of machine intelligence will be rapidly upgraded, representing the IBM supercomputer last February and people did the game, the result is the machine won, this is also a very striking case.

Of course, the future of this case will be more and more, machine intelligence will become an important symbol of social development.

Google has also done a number of services, it can do in 68 languages between the automatic translation, we previously thought that the machine translated things are unreadable, but now far beyond this level.

This is translated into English after the accuracy of more than the average person's translation level, which we have a cloud computing platform to support, so we can feel the great changes in this service.

So I think the real challenges we face in the future are not the challenges that we now see, but the challenges of the future are the challenges of human intelligence and machine intelligence differences, and people are too far away from machines.

I think the focus of cloud computing mainly includes several aspects, first we look at Cloud computing security, in the cloud computing environment, security problems and traditional security will be different, but most of the same. Because we traditionally have many databases of our banks that are open to the Internet. Now that we have the cloud, basic security should be possible, but the biggest difference between the cloud computing platform and the traditional platform is that the cloud computing system is a multi-tenant system, a lot of people come to your cloud computing platform, and we protect a data center very well so that no one else comes in, Now others are your tenant, he rented your cloud computing virtual machine, he did their things, at this time you can not be able to this tenant is credible or not credible.

So if he uses your platform to steal the secrets of other users, that's bad, how to prevent cloud computing data centers from being secure in a shared state, which is important in virtual machine isolation, is a security challenge that cloud computing data centers face.

Another problem with data center construction is that Facebook has made a huge contribution to the world by publishing a few papers, one in English, and a few papers that have changed the world to make cloud computing popular. Facebook wants to make a bigger contribution, it has completely disclosed his methods, the data center is the construction of the method is very amazing, it employs a lot of new technology, its pue energy efficiency to do the world's best, it can bring us a lot of things to reference, Its data center is built on a Gobi desert, a data center that uses a lot of new technology, such as spray, water mist, and non-standard cabinets, which include its motherboards, which are customized to improve its energy efficiency. We can learn a lot of things.

The problem with cloud computing energy is also a significant one, with Google acquiring a stake in the east coast of the United States to address its energy problems, a wind power plant in the east Coast Sea, to power the cloud computing data center. In the desert, it builds solar collectors and uses desert energy to supply cloud computing data centers. Microsoft is in the Chicago data Center using a large number of pipelines, the supply of cloud computing data center cooling, the scale has been built very large, the data center won the European Energy-saving award.

Google also has a number of new patents, such as the cloud data center into the sea, so that tidal power generation, as long as there is cable connected to power. The energy efficiency of Google's cloud computing data center after so many years of development is getting better, Google built in Belgium cloud data center is also built very good, raw materials do not have air-conditioning, Belgium this place in a year 7 days temperature will exceed the standard, this 7 days Google put this data center does not let it work, Global access will automatically go elsewhere, and Google's platform can support any data center to be destroyed and its data will not be lost. In the domestic 360 security guards also have this ability, we saw a room in Langfang, due to air-conditioning damage, the dangerous house temperature exceeded, 360 in Beijing to Langfang's computer room under when all the visits, the user simply do not feel my data transferred to another place.

Another is the cloud computing platform, we can see in the domestic open cloud computing platform is a grand, a Aliyun, we rented his platform can be used in cloud computing applications. We use Remote Desktop to connect to cloud computing, which can be used as my machine.

The application of cloud computing in this respect, we can see more and more applications are cloud computing applications, such as Baidu, Baidu also built some cloud computing platform, the largest cluster is 3000 machines, in 360, Jinshan, rising have done a cloud security.

We can see this application on the Internet, a variety of notebook search, are based on cloud computing to do, this multi-scale search using traditional methods to do efficiency is not high. We see Sina Weibo, and once you log in, there are 5 people in his care who are the same people you care about, do you want to add him as a friend, everyone knows this function, this function is a great person, because everyone will pay attention to a group of people, Sina users have a lot of people, so you think about how much the calculation, And always. So we are able to make this application, relying entirely on the unprecedented computing power that cloud computing offers us.

Taobao also offers some cloud computing services, for example, Taobao can give some of the better sellers to provide services, we see Taobao inside the crown sellers, they do not necessarily because his things sell very good, mainly he did the data analysis is particularly good, which for us to improve sales is very good, I am a what all sell the enterprise, I through Taobao inside purchase data Analysis Service, I to choose the commodity, what price to pricing, this is very good. If you're always selling things that are hot issues, prices are competitive, plus you're crown users, it's easy to get a few crowns. Taobao sells things, or people, are doing data analysis to do good talent is the last winner.

Ma also said that Taobao's biggest wealth is its data, through Taobao data analysis, Ma Yun mastered a lot of things, such as which province of the customer is the shortest height, and so on, similar to these analyses are Taobao data analysis, which province customers wear the longest shoes, he can do data mining.

Now the international organizations are doing cloud computing standards, these standards of cloud computing in our cloud has flourished after everyone discussed the cloud computing standards, at the very beginning there is no standard, at this time how you do this standard is difficult to do. We all have more standards to discuss how to achieve interconnection between different purchases, so that it can exchange resources, exchange information, and have a security facility, that's what we do with cloud computing, and the most important thing is the standards between clouds and cloud, not within the cloud.

So we can see now that the standard of cloud is hot, most of them are in between, even if we say billing mechanism, is how to charge between different clouds.

Let's talk about the nature of cloud computing, which is the service of rotten machines. Here's a look at a typical cloud computing platform. This is the Cstor cloud storage System, this platform its machine, a pair of management nodes can manage thousands of storage nodes, this is two sets of machines, after all the access to the first is to manage the node to ask me where the data should be, management node tell it where you should put it, The program at the end of the user will automatically access the specific storage node, there will be a bunch of storage nodes for the user service, different users will be assigned to different storage nodes, so its IO ability is very strong, so that the system has an unprecedented throughput capacity.

Now the cloud storage is a kind of architecture, so that all nodes are both storage nodes, but also external service nodes, all the nodes can be external, so its service capacity becomes very strong, any loss of the node does not matter. This architecture is done using a cloud architecture.

In addition, it can be the different poles of the cloud storage virtual together, making it an unprecedented large storage, the total amount of data in the world so far is about 2000 EB, but this system can support 10 million EB capacity. Performance is also as the node grows, it can increase its performance, because each node is an external service node, so its number of nodes is sufficient enough performance is strong enough, this and network-related, the main bottleneck in the network.

The hardware is a new cloud storage control hardware, this hardware makes the power consumption greatly reduced, this board with 16 hard drives, this is 100,000 energy consumption, this system can be very high density, Such a rack support of 1.1P, is more than 1000 TB storage capacity, any knot is not good, this section will change the red light, you do not have to tube, after a few months you change a hard disk inserted into the can, all the maintenance do not need to use personnel to do. It is highly reliable, it can lose half of the node, the data will not be lost, but its solubility density is 1:1, this method is very extreme.

All nodes can be external services, power consumption is the traditional data center of the cloud storage power consumption of one-fifth, the entire system energy consumption is very small, integration now NMC provides 360T of cloud storage, this is 1.1P, it is 3 times times. In terms of use, it can do without learning, it hangs on the cloud storage, your machine will be a more than a hard drive, do not need to do any transformation and need, it is also very low cost, cloud storage price of the equivalent of the traditional price of one or 10 per cent of the price of 5.

We all know that a very important case in Nanjing, January 6, the Ministry of Public Security in Nanjing, a case, shooting, all the process of solving the data are on this system. The Nanjing Government Data Center also uses this system.

Cloud processing this platform, the bottom uses the (English) system, this platform it is not a commercial system, in its performance and reliability of some limitations, such as a management node to be paralyzed, this time the system to stop, at least one hours to recover. In addition to its performance, it is suitable for statistical applications, so it is not always applied, so it is in the data management, can not always make feedback.

For example, we have to deal with the data moved by China, the effect is very good, basically, for example, Guangdong Mobile data is very large, can be in a second product structure to give out, and other systems of the same kind have to be a few minutes to have results, the difference is very obvious.

At the same time our technology is also used to deal with traffic data, the city has a large number of camera equipment, as soon as the car passes it will take a picture, the record is few when we do not see, just like in Nanjing such a city, a year is 200 billion, if we put this relationship in the relational database, general 1 billion is over, There is no way to deal with it, its query time is two minutes.

Suppose I have every car in Nanjing Pass I have to check the car is not a set of cars, we will check all the cards have the license plate, which means that the car and the same car license plate number appears in a different position, which means the car is a set of cars. You can never do this with a relational database. So with any car on the cloud platform, you can compare it to all the data. Now that we have done this platform in Jiangsu Province, we can make this data can be quickly feedback.

In addition, the cloud transmission system, can be able to (English) to the speed of remote data transmission greatly increased, this is our standard protocol to transfer, is this. Because (English) already have 30几 years of history. When the transmission, the two sides are repeatedly to shake hands, so it is inefficient, if you lose the packet it's speed will be reduced by one times. So the protocol is poorly transmitted over a remote and low quality network. We have made a new set of protocols called (English), which can improve performance to such a high amplitude, there is a significant increase. From Beijing Miyun to Peking, optical fiber transmission can be to 70 trillion per second, is a different transmission system, and now is also a number of listed companies, such as Shenzhen, the days of the media cooperation, he is to the television program to the country, the day is to send the hard drive, now has this system, He doesn't need to send a hard drive, this program to the system, on the Internet through the television program passed, to the national television stations, has done 6 provinces, this year to add 10 provinces, can quickly through the system to pass it to all parts of the country, thus greatly saving the cost of our operation.

We have two books, a book is "Cloud computing", this is the largest domestic sales of cloud computing book, we recommend reading this book, this is all the representative school of Cloud computing is how to achieve the description. In addition, there are many examples of Hadoop, there are some practical cases. Thank you.

(Responsible editor: The good of the Legacy)