Seven principles of data center efficiency from Microsoft
Aerial view of the Microsoft data center in San Antonio, Texas (DC 43660 m2)
Paul Slater is one of the most involved Microsoft telecommunications workers. He is one of the employees of the corporation who are responsible for the development strategy of corporate data centers. At the same time, Paul believes that the automation of data centers, with the maximum replacement of human workers by robots, is the business of the next 10 years.
In addition, Paul Slater sees some other areas of data center development that he recently presented to his audience at the Data Center World conference , which was held in Las Vegas. These directions, or principles of the effectiveness of data centers are set forth in the sequel.
It is worth noting that Microsoft has very good data centers, which many experts call the most effective data centers in the world.
1. Designing a data center in accordance with local conditions
The answer to the question “what makes a data center effective” is quite complex and complex. This is one of the reasons that suppliers of solutions for data centers offer different equipment, which, often, differs significantly from its analogues on one principle or another.
According to Microsoft, one of the main factors that should be taken into account when creating a data center is local conditions. For example, a data center can be built in a place where land and electricity are very cheap. In this case, it is possible to build large data centers with a high level of energy consumption without any special costs. If the company needs a data center in the middle of Manhattan, this is a completely different story, every square meter and watts of energy matter here.
All this may seem obvious, but many companies are building data centers according to standard solutions, not at all taking into account local conditions. In this case, you should not expect high returns at low operating costs.
2. Standardization and unification
The most effective will be a data center, the space of which is used at 100%. The corporation is trying to fill the space of data centers belonging to it, completely, as soon as possible. In addition, the equipment that the company uses is standardized.
In this case, the operational cost of servicing equipment (which may suddenly fail) is lower than in the case of using any special equipment and elements.
The company tries to manage its data centers using tools such as, for example, Data Center Infrastructure Management.
3. Designing the possibility of change
Over time, the data center and the requirements for it will change. Therefore, when creating a data center, it is worth preplanning the possibility of changes. After all, technologies are changing so quickly that few people have time to adapt to them right away. Technological solutions used today may need to be changed in six months. At the same time, now you can’t even imagine what exactly will undergo changes.
An example is the replacement of conventional disks with SSDs - this has led to some changes in the thermodynamics of data centers. Imagine immediately it was very difficult, if at all possible.
Process automation is most effective in the case of a homogeneous, homogeneous environment. If everything is unified, as recommended above, then introducing automation processes will be a simple task (everything is relative, of course).
Two years ago, Microsoft tried to standardize all of its equipment and processes in data centers. As a result, later it was possible to quickly deploy some automation solutions for a number of processes. At the same time, Microsoft is trying to actively participate in the Open Compute Project, using its best practices and partners' best practices.
Openness is another important factor.
5. Integrated Environment
If you want to get an effective data center, Slater says, you should plan from top to bottom. First, you need to consider the capabilities of the processes and applications that the data center equipment should support, and after that, select the equipment itself.
If everything went right, the design will end with the creation of a highly integrated system where each element is used in the work, supporting a particular application or service.
6. Reliable, flexible software
Good, reliable software pays for itself quickly enough. At the same time, using software, you can “squeeze” out the maximum possible work efficiency out of equipment, even if the equipment is somewhat outdated.
Microsoft uses cloud solutions, and the company also began offering cloud solutions to its customers instead of individual software licenses. Therefore, now the company can scale its work, to a very large extent.
One of the most striking examples of the implementation of cloud software is the Microsoft Office 365 service.
7. The data center must respond quickly to changes
The sooner a new service is put into operation, the better. Moreover, the service should be supported by the equipment of the data center itself. Accordingly, initially the data center project should include the possibility of such changes and support for new equipment. Thus, Slater believes, one can "saddle Moore's law ."
At the same time, he emphasizes the need for the standardization of equipment and services described above. This allows you to most effectively use the resources of your DC.