Video Interview and Editing: the Way of the Microsoft R & D team's private cloud applications (2)

3. The number of physical servers is increased, but the number of maintenance personnel is not increased.

 

Original video address: http://v.csdn.hudong.com/s/article.html? Arcid = 1, 302330

 

Tan Mao:What are the changes in the maintenance staff of these 1,500 servers, coupled with thousands of VMS above?

Liu Yun:There was no change in the number of maintenance personnel. At the earliest time, we were actually three in Shanghai, and our business in Beijing was not started yet. In fact, we have added a new team in Beijing since, we have added one engineer in Beijing. As a result, the number of people increases by 2.5 times from the server, but the number of people does not increase.

 

Tan Mao:How many people do these 1,500 servers need to be managed according to industry standards?

Liu Yun:This is not the same in all regions. I will give you an internal indicator of Microsoft in the United States. In fact, we have not yet met Microsoft's internal indicators. Microsoft's internal indicator is that the data center has 1,000 servers with one engineer. This is the standard for Microsoft data centers.

 

4. seamless migration of traditional physical servers

 

Original video address: http://v.csdn.hudong.com/s/article.html? Arcid = 1, 302331

 

Tan Mao:We also want to talk back to the technology. In fact, we also want to know that many customers also want to build private clouds, but they are quite embarrassed, how can they seamlessly migrate traditional physical servers to the private cloud?

Liu Yun:In fact, we have encountered the same problems as traditional industries. When we perform migration, the first problem we want to solve is the one you just mentioned, this is how we migrate existing applications to virtualization. In Microsoft's system center product, a p2v function is set up to migrate some early services and applications from physical machines to virtual machines, with this function, you can transfer a windows, Linux, and application running on the hardware layer to a virtual machine in about 10 to 20 minutes. All settings can be retained, all team relationships can be retained, and all application settings, including storage and database.

In fact, we also used a method, that is to say, after we migrate the physical machine resources to the virtual machine, we re-install the physical machine, that is, we call it reprepare, put it back to the resource library to make it a virtualized and available resource. It did one thing, but now it does three things. In this way, we can maximize.

 

Tan Mao:Another point is that we know that some of Microsoft's technologies are used in the cloud, and some products are used in the cloud. In fact, some management tools andProgramIn the future, especially platform-based servers, management may be a big headache. Microsoft is managing it because Microsoft also has its own products. This one?

Liu Yun:This is actually the most common problem we have encountered by all our engineers. We started to look at the opinions of various vendors, including HP, Dell, Lenovo, and inspur servers. Each has its own management tools, such as HP's Insight Manager and Dell's openmanage, at the same time, we also need to manage the switches and storage in the lab. At this time, as you just mentioned, we have read a lot of open-source management tools and commercial management tools on how to manage so many systems. Finally, we can see that Microsoft's system center is the most convenient.

We can be integrated with Cisco devices, Dell's openmanage, HP's ingisht manager, and Lenovo's management tools, we can see the CPU load in Cisco switches on a platform, and the power consumption, fan speed, and CPU memory of Dell servers, you can view all the information on a platform. In addition, the health status of each device is reported by email in real time, so that we can take the initiative to take appropriate actions in the first place.

 

Tan Mao:For this kind of management, it is done by the developers themselves according to Cisco standards?

Liu Yun:This is not because all vendors follow common standards. For example, Cisco complies with SNMP network management standards, and Dell has its own specially designed suite that complies with Microsoft COM standards. You can use these packages as a software package to describe its hardware. They are all downloaded for free. With this, you can monitor these servers until the hardware layer. For example, if an error occurs in the memory slot and the hard disk may be damaged immediately, the first hard disk and the first memory can be reported in real time.

 

 

5. Security and performance of Microsoft Virtualization

Original video address: http://v.csdn.hudong.com/s/article.html? Arcid = 1, 302332

 

Tan Mao:In fact, virtualization may not be easy to accept now. They still have doubts about some of their security. In virtualization security?

Liu Yun:Indeed, many users have many questions in addition to security-based virtualization. We include actual testing and Microsoft's internal, in fact, Microsoft's virtualization product goals, because every product we design has a product goal, for example, what problems should this product solve. So we can see from our product design goals and our implementation goals that the conversion efficiency is basically above 98%. In this way, we can basically say that the conversion efficiency is, you can understand that the performance of a virtual machine is exactly the same as that of a physical machine, in terms of CPU conversion. There must be so much memory, and in terms of disk efficiency, it is based on running virtual machines. As long as you reach the maximum I/O, it is safe. You mentioned security just now. In fact, there are some standards in Virtualization in the industry. How do you implement virtual machine segmentation.

 

Tan Mao:98% of the conversion efficiency is based on the Windows platform?

Liu Yun:In fact, the CPU conversion efficiency of all things is as follows: whether it is windows or Linux. It actually acts as a fake CPU, so Virtualization has a fake CPU, and your virtual machine is on this fake CPU. The conversion efficiency is to pass the virtual machine.

 

Tan Mao:Are you still optimizing and optimizing the entire system?

Liu Yun:In our daily work, the first one is to keep adding new machines, because we have many projects that will come in, and we will also eliminate some resources. In fact, our daily management work is maintained on this platform.

Another piece is that we add new things on the self-service portal of this platform. We are doing, for example, my level-1 Management, and my management is in Shanghai, however, we now have servers in some data centers in the United States and Beijing added to Shanghai. In this way, I can manage Microsoft on a platform. Because our project requires us to work in different regions at the same time, but at least two people need to work on the same platform, and we need to integrate it together.

 

Tan Mao:Another concern is that virtualization-based automation and service-based management are the core tasks of virtualization. In fact, you can also talk about this. Microsoft has also done a lot of work in automated management. Let's talk about it in detail.

Liu Yun:Virtualization actually influences productivity and makes virtualization technology more convenient to use like electricity and water. You must definitely need some supporting facilities, for example, you must have strong and weak current wiring of the building before it can be delivered to users. In fact, virtualization technology has brought about computing power in this form. First of all, let's look at what our engineers need. He needs several virtual machines, how can we quickly use the service warehouse as a simple method and then choose. Prepare thousands of Virtual Machine templates in advance in our database. If you need them, you need to choose a Chinese or Western style template, just like ordering food, hunan food is still Cantonese food. We have prepared everything we need to choose during the test. This is the food. He can directly choose from the menu and use our automatic mechanism. Then he can choose his dish and directly turn it into what he wants.

Therefore, we will select users, integrate virtualization technology and automation to serve internal customers.

 

Tan Mao:What do you implement at present?

Liu Yun:Through the vmm SDK, it will have an automatic script based on power shell. First, we will understand the user's needs. He may need three machines. We abstract this common requirement because we have shared requirements with many teams, abstract these common requirements as a service package, and then use the automated scripts of vmm to automate many things. In this way, the user does not need to repeat the automated work, which solves one of the problems I mentioned earlier, that is, there will be repeated tasks between the team and the team. Now we can abstract this repeat into a hierarchy. We call it the virtualization infrastructure layer. Previously, automation was based on hardware. After we become virtualized, automation still needs to be done, but we have written an abstract description, so users do not need to repeat the layer, for him, it is what he wants, and we have done more work, combined with some automated scripts.

in his original automated script, we developed a concept similar to the script library, that is, to directly transfer the SQL Server and office in his script library to that place. In this case, you can imagine that you need a test environment. First, you need two CPUs of memory, a few windows, and then the windows is changed to AD, and the windows is changed to exchange, after you have selected these items, click here and we will automatically prepare the virtual machine and then call the selected ones. In fact, the script is in our library. In this way, users focus on their own professional fields.