Microsoft STB: The private cloud application of the research and development team

STB China, which started in early 2005, established its vision of becoming the world's leading development, innovation and business hub, helping the server and Development tools Division succeed in the Global and Chinese worlds. As the core Department of Microsoft Asia Pacific Research and Development Group, the China Research and development team of the Server and Development Tools Division currently has nearly 400 developers in both Shanghai and Beijing, actively participating in SQL Server, Visual Studio, System Center, Windows Sgt Global development and innovation of Microsoft core platform products such as Business server, Windows HPC server, etc.

Microsoft Server and Development Tools division China Research and Development Engineering Laboratory (Shanghai)

3 Operational engineers behind 1500 servers

From the beginning of 2005 to 2011, with the expansion of the business and the scope of growth, the number of servers in STB to increase to more than thousands of units, however, in the number of business growth and server management tasks increased, its responsibility and the huge maintenance work is still composed of three engineers of the Research and Development Engineering Laboratory team It also needs to build and maintain the product testing environment for more than 10 product groups. Like all fast-growing organizations, STB is suffering from a lack of strategic planning for IT infrastructure.

Liu Yin demonstration of private cloud practice in research and development Engineering labs

"At present, STBC China's operation and maintenance team in Shanghai, together with me, there are three people, one of whom is responsible for the hardware warranty, one person responsible for the network, two for mutual backup, the Microsoft Server and Tool Development Division in Shanghai, Research and Development Engineering Laboratory management engineer Liu Yin, this pressure is undoubtedly enormous. 2007 joined the STBC Liu Yin, is currently responsible for the Microsoft Server and Tool Development Division in Shanghai Research and Development Engineering Laboratory management, the main content of Microsoft's development, testing and product management team to provide lab services, including testing and building environment, as well as product construction and product design.

"STBC has nearly 1500 servers, the platform is based on system Center, a total of virtualization Services management system, server management, management client, backup management a total of four systems", Liu Yin said, cloud computing has become the only way to solve the current operational bottlenecks.

It is understood that in StB, Liu Qing spent nearly a year with 412 servers to build a private cloud, a total of more than 5,000 virtual machines running. And on the entire management platform, the team's resources are grouped by the way, there are more than 13 to 15 product groups, they will be assigned to different groups, each user will be within their own group to allocate virtual machine resources.

"In a few years, the number of servers has increased 2.5 times times, but we have not increased the number of team members", when it comes to the expansion of the room in recent years, Liu Yin that a strong management platform coupled with it process automation management tools Opalis is the key to achieve.

"Research and development is characterized by the frequent need for servers to work offsite, at least two people working on the same platform, Beijing or some data center servers in the United States need to add to Shanghai, so we need to integrate it together." ”

"But we haven't actually reached the internal targets of Microsoft, because Microsoft's internal metrics are that the data center is 1000 servers equipped with an engineer." ”

Cloud Computing in the research and development team

Liu Yin that for the research team, the benefit of the private cloud is to make computing resources easier to use, like electricity and water.

"The templates for thousands of virtual machines are ready in advance in our library, and any development engineer can make a choice based on the automated mechanism." ”

In the field of virtualization, there are two modes of work, one is the resource first, a call performance first, popular, that is, less of the resources to spend more time and less time to spend more resources. For users, the best one is more flexible, able to deal with a variety of emergency needs, very agile. "We've been pursuing the maximization of resource utilization, where each team used to have their resources in a lab, without a good shared platform, and with this virtualization management, he could easily share team resources." Just like this project A, he has his busy time, busy stage. While another team was busy, he had no more machines when he bought new machines. Now there is a choice to transfer a portion of the resources from another team, because all machines are virtualized and he doesn't interfere with each other. ”

STB Virtual Machine Management efficiency

It is understood that, on a specific implementation, VMM's SDK has an automated script based on the power shell. As an operational engineer, you need to first understand the requirements, abstract the common requirements as a service pack, and automate the device management through VMM's automated scripts. This allows users to use no duplication of automation work, which also solves the problem of repetitive labor between teams and teams.

"The original automation was hardware-based, and now that we've become virtualized, in fact, automation still needs to be done, but we write an abstract description of these, the user does not need to use the level of duplication, for him is what he wants, and we do more work, combined with some automation script", " We can imagine that developers need a test environment, first you need two CPU how much memory, and then need a few windows, and then need this Windows is turned into ad, that windows into Exchange, you choose these, click, Behind us automation will be the virtual machine ready, and then call you selected good. In this way, the developer will be able to focus on the technology he is focusing on by fully automating the operation. ”

Developing a private cloud requires understanding the bottleneck first.

Liu Yin that, for research and development, we need to first understand the bottleneck of the problem. Since taking over the job last year, Liu has spent six months researching internal processes, resources, and internal engineers ' work habits. So when designing a private cloud, the goal is to solve the top three issues.

"We need to assess how the private cloud improves production utilization", Liu Yin said, "for research and development, especially between team and team have a lot of duplication of work, such as testing, you need to first deploy Windows to an environment, and then build a network environment, run has not officially listed products." In fact, in the process of building the environment, different teams often need to do the same thing. So this is a problem and we need resource sharing. ”

In addition, engineers are also very interested in the time it takes to deploy the machine. For example, it takes at least 20 minutes to install the system, or at least 40 minutes to install all the latest patches, so that the complete time is at least one hours. "If the physics machine is doing this test, you need to do an ad, and through the server, not 2.5 hours is not completed." Eight hours a day, engineers can do only four sets a day. And now virtualization technology can be deployed in parallel, each state of the virtual device, from the beginning of deployment to the end of only 20 minutes or so. Reduced from 2.5 hours to 30 minutes, this is a qualitative change. ”

Azure represents the future

In fact, Microsoft's platform, Service Cloud (PaaS), Windows Azure, has been open to businesses for a year, but it has not been widely used domestically for well-known reasons.

In response, Liu Yin said, Microsoft's cloud computing is currently two-legged walk, one for the computing center of the distributed computation, the other is Windows Azure, Microsoft internal called the OS for Datacenter. Instead of a separate hardware, Windows Azure supports the design of the entire datacenter, covering storage, networking, and so on.

"Microsoft has no baggage, because in Microsoft's view, the ultimate goal of cloud computing is also the public cloud, Microsoft has been a rational approach to", "and the current adoption of the public cloud + private cloud mode, is largely due to the enterprise's own IT infrastructure, at least for a very long time. ”

"Microsoft's next-generation VMM will combine the two to allow users to manage both public and private cloud resources to migrate applications more freely," Liu Yin said. "I believe everyone likes more choices, and no one likes only one option." ”

(Responsible editor: admin)