IBM Shu Jinghua: Software definition Infrastructure

December 12, the world's first large-scale conference to explore the industry Internet, 2014 CVW. The industry Internet conference was held in Beijing and was synchronized through the big screen of New York Times Square. The conference was made by the Asia Letter Group, the cloud base and the Chuang-Zhuang economic and Technological Development zone jointly hosted more than 5,000 global it and traditional industry leaders and elites who are concerned about the development of Internet and traditional industries, and explored the evolution of "Internet access to traditional industries", "traditional industry internet" and the technological model and business innovation of industrial Internet.

In the afternoon cloud Platform @ internet forum, Mr. Chiang Kai-shek, from IBM, brought a keynote speech on software definition infrastructure, which introduced IBM's software definition infrastructure products and application cases.

The following is the full text of his speech:

Shu Jinghua: Thanks to the invitation of the organizers, I came here today with a jam, so I moved to the end. The topic of my speech is called Software definition infrastructure, which has two main verbs, one is acceleration time, reduce cost, in fact, I think this is the nature of cloud computing.

Cloud computing has been going on for many years now, fried is very hot, but in the circle of people should know that the profit of the company is very few, or today by the cloud, whether it is this model, or the concept of it, eventually the rapid development is not particularly much. Even like Google and Amazon, I don't think there's much to gain. Of course, this is not the subject of my speech today, my speech today is called software definition infrastructure, whether it is cloud computing or large data, or any of the concepts discussed today, in fact, the software should be a flagship.

Today, we use the Internet so high, but you do not know where you use the service is in Beijing, in Shanghai, or in the United States, you do not care, but you still use a great. Back to the mobile Internet is also the case, this is a trend we are facing today. Take a look at the details of my report here.

First, from the trend, our organization or enterprise, today's theme is called Industrial Internet, our industry is actually facing many new opportunities. This new opportunity is really about cloud computing, or big data, and today's mobile internet app. Although we don't earn too much money in the cloud, it creates a new economic model, our consumption pattern. Your consumption pattern of computing resources has completely changed. In the past 2000 years, everyone will run to Zhongguancun 撺 a machine, one is playing games at home, the second is the internet is a new thing, we also want to get some news on the Internet, but your carrier or rely on your machine, and the higher the machine configuration is better. But today is not the case today, the game can be online, you do not need to buy too good computer, this good background by the basic service provider to provide. And you can get the resources you need on demand. For example, cloud storage, on the internet to provide a lot of online cloud space, including 360, Baidu provides more than 10 T. This was unthinkable before, buying a 1000 g hard drive in the past is unthinkable, and now everything can be put online. This is a change in consumption patterns.

The second is big data. Big data can help us get the answers you want faster, but these data are made by many of our mobile terminals here. This data itself is worthless, the data only to do analysis to be valuable, today's domestic bat, and such as foreign Google, the company has become a data company. Of course, we are different from the previous Internet companies are we do is enterprise-class large data, whether it is ERP or CRM has a lot of data, these data is not ordinary, how to use these data to assist enterprises to make decisions, this is the big data to do things. This is big data sharing.

Third, social media. The most important change in social media and mobile interconnection is interaction, we can now communicate with friends at home and abroad, and do not need to spend money. Today, telecommunications companies, mobile or linkage, are worried about what to do in the future, micro-letter has been free to call, free video, text is gone, the telecommunications companies rely on what to live, so this is the question they want to think about.

IBM proposes a software definition infrastructure that is the basis for future businesses to either do their own private cloud or a hybrid cloud.

From three levels to consider, one is optimization. Optimization is from the application layer, just come up to discuss whether it is from the analysis of large data to optimize some of my data, or from the mobile end of the data generated to optimize, so that my application cycle can be more effective, more effective to support the development of my business.

The second is automation. Today's infrastructure is actually more and more huge, technology upgrading is also very fast, we saw the CPU immediately 18 nuclear single processor will come out, 18 cores in the past is 18 machines, how to let your application in this 10 machines running. And our cloud computing environment may have 100 or thousands of 18-core machines, how to make it run like a machine, it is also rely on good standardized software to do this thing.

Third, how to better extract your resources, so that the interaction of your resources more convenient. We may have a lot of servers, a lot of storage, even a different application, what kind of resources to what kind of people to use, what kind of dimension. For example, I have 100 machines, is for 1 million of people or 100 people use, this is the strength of your resource division, is defined by the software.

The infrastructure has three ideas, the first is the formation of pools, how can I better integrate resource pools for different infrastructures purchased in different time periods. How to make more applications, no matter what the application, whether you are a database or unstructured video or pictures or text text, and so on, how to interconnect these different platforms. In addition to management, this management is dynamic. How to use the better management tools or software to connect the previous resource pool and these applications, this is our software definition infrastructure, but also we do for the enterprise cloud or analysis of the design of a framework. Based on such a software definition infrastructure, you can get up to 4 to 100 times times more than the optimization of your infrastructure. At the same time reduce your IT costs, because electricity, bandwidth costs are very high.

This is our application based on scalable applications, how to operate in the data center, which is today a status of applications, High-performance computing, high-performance analysis. For example, the stock market in the financial industry, the index changes very quickly, this is the use of typical high-performance analysis software. The second is based on Hadoop large data distributed computing framework. Three is the past has been tall, called high-performance Computing, our weather forecast, automotive, aircraft manufacturing are derived from high-performance computing simulation. High-performance computing has undergone many stages, the earliest time high-performance computing is a large host, the machine is very large, and later developed into distributed computing, to today more extensive, but it can carry a lot of distributed applications. Four is the distributed computing framework, many Internet startups are using different databases, these future on large data platform should be the main direction.

Like the smartphone we're taking today, your smartphone, no matter which company the apps is developing, will have an iOS platform to host your application and be well connected. Our traditional enterprise data center or also want to do, how to do? Because you have to do this, there are two key points, one is the infrastructure, the infrastructure is the server, such as each of the different features of the application, I put what application on the server. X86 inside also I am distributed to run, or in a large memory of the machine running, there is a need for a good scheduling. Speed up the operation of the infrastructure through scheduling.

The second is also to have some workflow engine. Workflow engine is equivalent to a middleware, after the hardware can be unified action, I need to host some of the above application middleware, it to start, I put an application to the architecture, of course, it with the background of the resources management software to have good communication.

The first one on the left is the high performance analysis low latency middleware developed by IBM based on 15 years of experience, which can support many of the calculations of stock trading in the financial sector. Banks spend a lot of money on deposits, loans and loans, and more than 10 years ago the Chinese government spared trillions of of the bad assets, and why foreign banks rarely have such bad assets, and the key is that they often do this kind of liquidity risk analysis. For example, I expect to have 30 billion loans back tomorrow, I will lend 40 billion to go out first, of course not. If you do not do this liquidity risk analysis or the cycle is particularly long, we have a domestic customers, one months ago, this cycle is obviously too long, and later adopted our framework, now almost one day, the key is to improve the utilization rate of funds. I know how many assets I can lend, and I can operate well and reduce bad debts. We have this middleware to support these applications. We keep a good interface, and what you do on open source Hadoop can be very compatible.

The third is the High-performance computing engine. I think this is also an ancestor of distributed computing. Many of our country's large computing centers, such as Shenzhen, Beijing, Shanghai, they are public services, in fact, is similar to our internet companies, but he provides services to compute resources. For example, pharmaceutical companies need to do drug screening, the middle of a large scale calculation, I can rent equipment in the super Center, rent 100 units for two months. The center has 1000 machines, every day to apply for a lot of people, how to coordinate, how to distribute. It can be done with an enemy of 10.

We have a mobile phone chip customers, to do the test, Terminal Server equipment from the 2002 200 to 2012 6000, up 30 times times, but also need to rise, because people constantly find them. This can easily cause waste of resources, in fact, our software definition infrastructure is the core of how to improve the use of resources. In response to this situation in 2012 he used our workflow engine, the number of tasks processed daily from less than 1 million in the past day to 3 million, they did not buy any server, you can make full use of 6000 units. Through workflow engine, combined with resource sharing middleware, it is very good to improve the utilization rate of the following resources. This is the reason why we do the cloud or do distributed computing, the purpose of the sharing of resources to improve efficiency.

Four is for the new application now, we apply more and more complex, open source technology, open source platform is also endless, distributed in fact, database and so on, how to define these applications in my resources priority, how to control it. I can not always rely on virtual machines, and some applications do not need to do virtualization, do virtualization will reduce its use efficiency. This new application framework is the same, how to improve the efficiency of resource sharing, we made an application service controller. In addition to being able to guarantee or control the installation order and boot sequence of the application, but also to do the migration, once I found that you have 10 of which 2 are broken, I can find a new 2 to migrate past.

After we have solved the previous calculation, what about the data that comes next? Share resources, no matter who you are, or what the architecture of the device, no matter what your operating system, my middleware can be very good to dispatch and distribution.

After the server level, the workflow in front of the computing resource is consolidated, the data management issues are then generated. Before the application of the resource needs of physical machines, there are virtual machines, many units are also doing their own cloud, there is a private cloud, but also some data on the public cloud. How these different architectures to integrate, through the bottom of the three tools, one is to do the management of physical machines, the second is the integration of the middleware mentioned just now, like high-performance computing to provide large data Analysis Services, we can directly deploy the application. If you want to make a private cloud, how to use software? We have a openstack based software that allows you to build a hybrid cloud with the resource middleware mentioned above, including the core strategy of scheduling. Put your private cloud and public cloud in your home, whether it's IBM or Amazon, and make it a pool of resources that you can share. We did a PAAs on top of the IaaS layer to help you do different kinds of clouds.

After the cloud is done, go back to the data management issue that you just talked about. Background storage and data relationships we do a flexible storage, you can use the calculation file to partition storage resources. Just said that Baidu provided so much space, in fact, he knew that these spaces must be used, but once someone broke, they will take certain measures, is actually flexible storage. This is the level of data management or storage management.

We've done the infrastructure, and after that the server layer is done, store it, and put it on a personal application or business application.

This is our case for a century in which Citi has nearly 50,000 CPUs running a specific business, including venture capital. The management and sharing of his global resources is done with one component of IBM, whose system has been upgraded 100 times times, and the efficiency of operations and the utilization of resources have improved.

There are other cases, the first is the Infiniti Red Bull Racing, three consecutive world championships, he used our high-performance computing middleware, and later used our large data analysis to analyze the running of the car in the process of collecting some data, so as to better optimize the tires and other automotive design.

The second is a mobile phone chip company, using our flexible storage technology, can improve the usage rate of 10 times times. Improve the efficiency of his entire storage space usage.

The third is a cancer-treatment company in Texas, USA, which provides decision support and diagnostics to more than 7,000 scientists around the world, using our flexible storage platform. They treat cancer in a different way, using a molecular atlas.

The following is a telecoms operator, and telecoms operators are doing a lot of things that internet companies are doing today, and they are also changing. Use our distributed computing technology to help him do some good scaling.

The main purpose of the software definition infrastructure is to balance the needs of different users at different levels within the enterprise, and our IT manager wants to reduce costs and not buy so many servers. Of course, end users or various business units also have their own considerations, I hope that when I need to use the resources can be quickly, I hope that I developed the chip can be two months earlier than others listed. In general, the use of software to improve your resource utilization, the infrastructure of the sharing of work well, management work well, so as to meet the needs of different levels. This is the core of our software definition infrastructure.

Thank you!

(Responsible editor: Mengyishan)