Shulou(Shulou.com)06/03 Report--

Summary

Server virtualization technology has been very mature, in a virtualized environment, multiple vm run on a single server, so resources can be shared more effectively, which is conducive to cost reduction. However, if resources are overshared, competition will occur, which may lead to performance degradation. Therefore, it is important to balance resource efficiency and performance.

What is server virtualization

Server virtualization refers to multiple virtual servers running on a physical server, which is what we often call virtual machine vm. By running os separately in each vm, you can run multiple applications on a single physical server.

In a virtualized environment, resources such as cpu and memory on the physical server will be shared by multiple vm. Each os thinks it is running on a separate physical server, so the vm runs completely independently and does not affect other vm.

Overload use in Virtualization Technology

The biggest advantage of a virtualized environment is the efficient use of resources. Allocate cpu and memory to vm as much as the physical machine can handle. If you create multiple vm on the physical server so that the total amount of cpu and memory allocated exceeds the physical server itself, you can achieve higher resource utilization. This is called overload use.

Knowledge and performance Analysis tools for performance testing in Virtualized Environment

There is no difference between the methods and matters needing attention in the normal performance test.

Performance Management of CPU

From a physical server point of view

First, you need to confirm whether the cpu usage of the physical server is high. If you don't do it, you need to confirm the situation of individual vm.

In the performance problem of cpu, what needs to be confirmed for the physical server is the competition of cpu, in other words, whether the logical cpu is being contended or not. The two common states are waiting for the assigned cpu (the virtual cpu of the vm puts forward an allocation request to the logical cpu, but in fact, because other vm is in use, it cannot be assigned and is in the waiting state) and waiting for multiple cpu synchronous processing (the vm assigned multiple virtual cpu requests the allocation of the logical cpu for synchronous processing, but because part of the logical cpu is being used by other vm Therefore, it cannot be assigned to enter the waiting state)

If there is a tendency to wait longer, it is possible that there is a performance problem with vm. On the object physical server, if the waiting time per unit time accounts for about 10% / 20%, there may be a problem with vm. Note that this ratio is for reference only, because different applications have different characteristics.

From vm's point of view

When the cpu usage of the physical server is not high, you can assume that the performance problem lies with the vm itself. There are two reasons why there are problems with vm itself: the cost of virtualization and problems within vm. Overhead can generally be reduced by using hardware-assisted virtualization capabilities.

The internal problem of vm is that due to the lack of resources in cpu, the design problems or imperfect settings of the application can not give full play to the performance state. In this case, as in the physical environment, additional resources or tuning of the application need to be considered. If only a specific vm has a problem, it may be because cpu is restricted in the settings of vm. Just lift the restrictions.

Performance management of memory

As in the case of cpu, the factors that affect memory performance can be said to be the extra overhead caused by virtualization and the resource competition caused by overload usage.

From a physical server point of view

Similarly, first confirm whether the memory usage of the physical server is too high. If not, confirm the individual vm memory usage. If the memory usage of the physical server is too high, then confirm the memory overload usage, which may cause the performance degradation.

From vm's point of view

If the memory utilization of the physical server is low and there is no overload, it can be preliminarily determined that the performance problem is caused by vm itself. Similarly, if there are two reasons to consider: the cost of virtualization (which can be reduced by hardware-assisted virtualization) and problems within vm (same as cpu, so I won't repeat it).

Reference: "graphical performance Optimization"

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.