This paper analyzes the meaning of virtualization from the personal point of view
1. The impact of virtualization on costs
A. Saving procurement costs:
Improve PC CPU, memory, hard disk and network efficiency while reducing power consumption
B. Savings in maintenance costs:
I. Less PC, less maintenance work
II. centralized management to relieve the pressure of maintenance workers
Iii. Flexible deployment of resources
2. Analysis of resource usage before virtualization
The following scenario is the most intuitive resource idle in order to ensure the continuity of service, many applications require a master, all the standby is idle test clothing and development machine, only testing needs or compile time will be used. Many IT company's servers are based on the early budget early procurement, if the project progress or user needs than expected, will produce a large number of machine resources waste test client, only testing time to test the development machine, only compile or debug when using a test server, testing time only useful
The following scenario is a low utilization access machine, high network occupancy, CPU, memory, low disk working PC, when high time low, especially those who need to shift the business, a group of people, the machine is idle (or not, but shared PCs can easily lead to work scene or file loss) cache server, IO low application server, general memory And IO are relatively low
3. How to deploy a virtualized host
3.1 Horizontal Division
For large companies, the cloud is too big, will lead to management confusion, so we have to do a certain level of division.
There are many ways to divide the level, such as the sensitivity of the business, or the department, the openness of the business (internal or external), the severity of separation. If the division is still too large, you can continue to level 2, 3 level division.
3.2 Maximum utilization of resources
When we are done, we can deploy our virtual cluster.
For the analysis in section 2, we can categorize and deploy our scenarios
|
Scene |
Cpu |
MEN |
NET |
Io |
Business Features |
Deployment |
A |
Main Standby function |
M |
M |
M |
M |
Require primary standby quarantine |
Different applications of the main standby mix, you can put A process and the back of the B process on the same machine on the different virtual machines |
B |
UI Testing and Working PCs |
M |
M |
L |
L |
Resource requirements are not high, most of the time is idle |
Pre-sell as many units of work as possible through XEN bottom |
C |
Front-End Network access machine |
L |
L |
H |
L |
Network internal and external network resources occupy high, other low |
It can be mixed with some networks that don't require much, like D. |
D |
Application 1 |
L |
L |
H |
H |
Storage applications are typically only high on IO and net requirements |
Mix and Match |
E |
Application 2 |
H |
H |
L |
L |
Background run applications, such as the generation of static reports, etc. |
|
F |
Application 3 |
H |
H |
H |
L |
such as caching or host-controlled applications |
Relatively independent, you can deploy those inactive backup logs for mashup |
G |
Application 4 |
L |
L |
L |
H |
Backup storage Applications |
|
H |
are high |
H |
H |
H |
H |
such as DB, |
Not suitable for virtualization |
Through virtualized resource pools, we can maximize the use of resources for the scenarios on the table above
The following is the use of the view
The figure above shows that virtualization can at least help us save more than 80% of our PC resources, and we will save more if we sell more, or if the application is less demanding on resources.
4. The difference between virtualized hosts and multiple processes
Many times, when doing distributed applications, there is a certain similarity between multiple virtual machines and multiple process deployments.
In fact, the difference is very large: virtual machine isolation is better, application problems caused by the extension machine or resource leakage will not affect other applications virtual machine deployment more flexible, users do not need to know exactly which physical machine, according to IP on it. Multiple processes must specify IP and port numbers to be more flexible to the user, and users can deploy crontab applications on their computing units without fear of impacting other users