The implementation of successful virtualization system and its operational dimension

If the virtualization system is compared to a house, good design makes the building frame diagram, implementation will brick the house up. Lack of work, the consequences of delay period is self-evident. And Yun-dimensional is the maintenance of the House, alteration and maintenance.

In this article, we talked about the importance of design for a successful virtualization system, which is just a good start. The hundred-Mile Road is half nine, and the implementation and operation of the Victoria will have to be done.

One implementation

Architects, if you really care about your well-designed plan, then you should not relax in the implementation process. I have also been a project manager, and this experience has helped me a lot in architecture design. The architect is the project manager's strategist, and you need to help design efficient and reasonable project plans and anticipate and control risks. For example, in our common physical machine virtualization project, consider a phased process. Put the important key servers behind the stage, using the experience accumulated earlier can reduce the risk.

Common problems in implementation are:

Personnel waste: Some people busy death, some people idle. Some people are unable to start a task because of the unreasonable arrangement of the project task or the dependency condition.

Recommendations: The architect of the level of personnel and the skills required to make a reasonable arrangement of personnel, timely advice to the project manager to increase or reduce personnel. Pay close attention to assumptions and dependencies, and respond to them in a timely manner.

Extra work: Even if Zhuge Liang is not available, then an experienced architect cannot anticipate the special circumstances that might be encountered in the implementation that would result in a workload exceeding the budget.

Proposal: can start from two aspects: one in estimating the workload is to have contingency measures and reserve funds, in order to prepare for a rainy occasion. What are the factors contributing to the increase in workload? It is due to the original assumption that the conditions are not tenable, or the new demand, or technical problems have not been taken into account and so on. If there is a clear explanation of the assumptions and requirements in the contract, it will not be passive.

Imperfect testing or excessive testing:

Testing is important, but it is often easy to ignore. Sometimes some important tests are skipped because of the time pressure, and sometimes it takes too much manpower and time to test because of the excessive patterning.

Recommendation: Testers Enter the state ahead of time and develop a streamlined and adequate test plan. As far as possible to separate large test tasks into small test tasks, shorten the test and feedback cycle, can be corrected in time to detect errors. Avoid late detection of major problems, resulting in delays in time and waste of early work. Although agile is applied to software development, it is also useful to consider which tests can be used to draw on the agile concept in the virtualization project.

Implementation is not in conformity with design

There may be a lot of people involved in the implementation, there may be a lack of cooperation, there are limitations on the experience and ability of the implementation staff, not a good understanding of the design intent.

Recommendation: A complete and executable installation document can avoid this problem. The documentation specifically indicates the portions that are tailored to the requirements, as well as the configuration details, parameters, and so on.

Two-operation-dimensional

Is it bob after the project is implemented? After a celebratory drink, you'll have to figure out how to transfer the project smoothly and help the support department to manage the system well.

A good system, if there is no perfect management, not long before it will be porous. The operational dimension guidance should standardize the key aspects of operational dimension:

Clear responsibilities (role & Responsibilities)

Do not give admin permission, do not think this is to trust colleagues, you may do bad things, authority means responsibility. The person who does not have the knowledge ability obtains the excessive authority, but also has the opportunity to make the mistake, against it to suffer.

Who has what authority, should be clear, and as far as possible process, standardization. The authority needs to be changed in a timely manner when the person leaves the job.

It also reduces the risk caused by human error.

Capacity management (Capacity Management)

People often have a misunderstanding of the virtualization platform, that there is no need to purchase additional hardware, there is a need to build a virtual machine bai. Often see a large number of virtual machines can not be very good use, excessive waste of server resources, resulting in the performance of the entire platform down.

Recommendation: Standardize the configuration of virtual machines, such as setting Primary (1vCPU 4G), intermediate (1vCPU 8G), and Advanced (2vCP, 16G) for user selection. The default 1vCPU, users require more VCPU specific problems specific analysis. Whether the application supports multiple CPUs, and whether the server is effectively supported. In CPU scheduling, there are situations where multiple CPU VMS are unable to quickly obtain the requested resources, while the latency is prolonged and performance is worse.

The vcenter Operation Suite can be a good help for capacity management.

Problem Management (Incident/problem Management)

Minor problems will become scourge if they are not remedied in time. Sometimes under time pressure, we will take some measures to temporarily solve the problem of appearances, but can not analyze the root cause. Then you'll be a fireman and be too busy every day to be ungrateful. There is a VM performance problem, vmotion to another host temporary mitigation. It could take a while to drag down the mainframe.

Change management (changing Management)

Change management is essential for a virtualized platform that runs critical business because of the catastrophic events that have occurred in the event of a change failure.

The operational staff is to fully understand the original design of the structure. Otherwise, the usual changes may conflict with the original design. For example, the administrator adds a backup port group and shares an uplink physical network card with Port group of the working network. When the backup task is running, the application on the virtual machine is found to be extremely slow. Because the backup preempted the network bandwidth.

When the storage administrator changes the storage device, the virtualization administrator evaluates the impact on the platform.