Overview
Server virtualization technology has been very mature, in a virtualized environment, multiple vm running on a single server, so you can more effectively share resources, help cut costs. However, if the resource is shared by excessive competition occurs, it may cause performance degradation. Therefore, the balance of resource efficiency and performance is very important.
What is server virtualization
Server virtualization refers to running multiple virtual servers on a single physical server, that is, we often say that the virtual machine vm. Vm respectively by running the respective os, you can run multiple applications on a single physical server.
In a virtualized environment, cpu and memory and other resources on the physical server will be shared by multiple vm. Each os will think he is running on a separate physical server, so vm is completely independent and does not affect other vm.
Overload using virtualization technology
The biggest advantage of virtualization environment is the effective use of resources. Cpu and memory allocation in the physical machine to vm can afford. If the total amount to create multiple vm on physical servers, making the allocation of memory and cpu carrying amount exceeds the physical server itself, it can achieve higher resource utilization. This is called overload use.
Virtual environment performance tests of knowledge
Performance analysis tools
And the normal methods used in performance tests, note there is no difference matters not to say
CPU performance management
From a physical point of view the server
First, we need to confirm cpu utilization of physical servers in the end Gaobu Gao. If you engage in, you need to confirm the situation of individual vm.
Cpu performance issues, the need for physical servers to confirm cpu is competitive, in other words, whether the logical cpu competition. Two common state waiting cpu assigned (refer to virtual cpu vm's proposed allocation request to the logical cpu, but in fact due among other vm is in use, and therefore can not be assigned, in a wait state) and wait for multiple cpu synchronization (that is allocated to a plurality of virtual cpu vm to sync the requested allocation logic cpu, but since a part of the logic being used by another cpu vm, and therefore can not be assigned to enter a wait state)
If the trend wait longer exists, there may be performance issues vm occurred. On the subject of physical servers, if the waiting time per unit time accounted for about 10% to 20%, it could be a problem vm occurred. Note that this ratio is just reference, because of the different characteristics of different applications also have differences.
From the viewpoint of vm
In the physical server cpu usage is not high, the performance issues that may be considered vm itself. vm itself problematic situation may take into account two reasons: the problem of virtualization overhead and inside the vm. Generally through the use of hardware-assisted virtualization capabilities can reduce overhead.
Vm refers to the internal problems due to lack of cpu resources, design problems or imperfections settings can not lead to full performance state. In this case, as in the physical environment, the need to consider increasing resources for applications or transfer of Excellence. If only certain vm there is a problem, probably because of cpu vm setup has been restricted. The lifting of restrictions can be.
Memory Performance Management
And situations like cpu, memory performance factors can be said to compete for resources with virtualization overhead and overload caused by the use of
From a physical point of view the server
Like, also make sure the physical server memory usage is too high. If it is not high on the individual vm memory usage for confirmation. If the physical server's memory usage is too high, then confirm the memory overload usage, it is likely to result in performance degradation.
From the viewpoint of vm
If the physical server's memory usage is relatively low, there is no overload used, can determine the initial performance problems are caused by vm itself. Similarly, if it is to consider two kinds of reasons: virtualization overhead (through hardware-assisted virtualization capabilities can reduce this extra cost) issues and internal vm (with cpu, so the explanation will not be repeated).
Reference: "graphic performance optimization."