CN106201700A - The dispatching method that a kind of virtual machine migrates online - Google Patents

Abstract

The invention discloses a scheduling method for virtual machine online migration, which includes: step 1, recording the CPU utilization rate of each host once per second, and the CPU utilization rate of each VM in each host; step 2, predicting each Whether the host is overloaded; step 3, select the VM with the largest coefficient of increase in the CPU utilization of the host on the host that is determined to be overloaded and migrate it; step 4, find the low-load host according to the minimum CPU utilization method based on the threshold; step 5 , reallocate the VM queue to be migrated according to the energy-aware optimal descending method. The method of the invention can significantly reduce the energy consumption of the cloud computing center, and has a lower SLA violation rate than the previous algorithm, that is, better service quality.

Description

A Scheduling Method for Online Migration of Virtual Machines

technical field

The invention belongs to the field of computer system memory system structure, and in particular relates to a scheduling method for virtual machine online migration.

Background technique

With the rapid rise of the "pay what you need" model of cloud computing, the workload and scale of cloud data centers are also expanding, and their power consumption is also increasing. According to ICTresearch statistics, in 2012 my country's data center energy consumption was as high as 66.45 billion kWh, accounting for 1.8% of the country's industrial electricity consumption in that year. By 2015, my country's data center energy consumption was as high as 100 billion kWh. In addition, according to the data analysis of the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE), 75% of the operating cost of the entire data center comes from the energy consumption of the infrastructure. The high energy consumption of data centers will inevitably reduce the profit margin of cloud service providers and increase carbon emissions. High energy consumption is bound to become an important factor restricting the development of cloud data centers in the future, so how to reduce the energy consumption of cloud data centers is an urgent problem to be solved in the sustainable development of cloud computing systems.

In the research, while reducing the power consumption of the cloud data center, it may cause the decline of service quality, and then violate the service level agreement (Service Level Assignment, SLA) signed with the user, which is completely unacceptable in practical applications . Therefore, the research needs to start from two aspects of reducing power consumption and ensuring service quality.

In data centers that provide IaaS services, infrastructure resources are generally provided to users in the form of virtual machines. Users can complete access to virtual machines and resource requests through the web interface. These virtual machines are servers deployed in the data center. superior. The survival time of the virtual machine is limited. When the user no longer needs these resources, the corresponding virtual machine can be uninstalled. As the user deploys the virtual machine in the data center for a long time, the data center will open many servers, correspondingly with the virtual machine If the virtual machine is offloaded, the utilization rate of data center hardware resources will drop very low, but by using the virtual machine online migration technology, the resource utilization rate can be greatly improved and IT costs can be reduced. When the virtual machines are over-concentrated, the response time of the data center and the guarantee of service availability may not be guaranteed, thus violating the service level agreement (SLA) signed with the user, so the service quality of the data center can be guaranteed in the future, or After the virtual machine is frequently uninstalled, the resource utilization rate will decrease. At this time, it is necessary to detect the resource usage of the data center and migrate the virtual machine if necessary.

In general, live virtual machine migration is divided into the following four steps:

1) Overload detection. Detect whether a host is overloaded, and if so, migrate a virtual machine on the host.

2) Light load detection. Detect whether there is a host under light load, if so, migrate all the virtual machines on the host and switch the host to the standby state to reduce energy consumption

3) The virtual machine migration option. Determine which virtual machine on the host to migrate out after the host is overloaded.

4) Virtual machine reallocation. Allocate all the virtual machines to be migrated to other hosts.

A complete virtual machine online migration strategy will be optimized in the above four aspects.

At present, some research has been devoted to optimizing the online migration strategy and reducing the power consumption of cloud data centers. Beloglazov et al. proposed a minimum CPU utilization strategy. When overloaded, one or more virtual machines need to be migrated out, so it is necessary to make a choice among many virtual machines running on the host. The minimum CPU utilization policy (Minimum CPU Utilization, MCU) is applied to this selection process. The core idea of this policy is to migrate out the virtual machine with the smallest CPU utilization to alleviate the overload. In recent years, many scholars have applied this theory to do research on virtual machine online migration. However, it is still biased to make a decision only based on the one with the smallest CPU utilization. From the perspective of the experimental results, the effect is not good. to the overload state. Abawajy et al. proposed that the MMT (Min Migration Time) strategy is to first select the virtual machine with the smallest migration time among all virtual machines. The point of this strategy is to fully consider the migration efficiency and minimize the impact of migration on performance. . However, the disadvantages are also obvious. Because the virtual machine that has the greatest relationship with the increase in host resource usage has not been migrated, it cannot be guaranteed that the host's overload problem can be solved after migration according to this strategy. After the migration, the host may still maintain In overload condition or on the verge of overload condition. Anton et al. proposed the Power Aware Best Fit Decreasing (PABFD) algorithm, which is the generalization of the BFD algorithm on the virtual machine placement problem. BFD is an algorithm to solve the box-packing problem. When solving the box-packing problem, its main idea is: first sort the "items" (virtual machines) in descending order, and then check all non-empty "boxes" (hosts) to find the most suitable one. The "box" of the "object" and put the object into the "box", if no such "box" is found, open the "empty box". This algorithm is specially researched for the low power consumption problem of cloud data center, which is consistent with the research object of this topic, and plays an important guiding role in the research of this topic.

Contents of the invention

The technical problem to be solved by the present invention is to provide a scheduling method for virtual machine online migration, which adds an improved algorithm to the four steps of virtual machine online migration: overload detection, light load detection, virtual machine migration selection, and virtual machine reallocation , because the virtual machine online migration itself will cause a certain amount of energy consumption and reduce the quality of service. Migration” to reduce energy consumption while maintaining quality of service. The virtual machine online migration process is optimized through the prediction-based host overload detection strategy, the dynamic and limited migration detection light load algorithm, the more optimized virtual machine migration selection strategy, and the reallocation algorithm.

A scheduling method for virtual machine online migration includes the following steps:

Step 1. Record the CPU utilization rate of each host and the CPU utilization rate of each VM in each host once per second;

Step 2. Predict whether each host is overloaded at the next moment;

Step 3. On the host that is determined to be overloaded, select the VM that has the largest correlation coefficient with the CPU utilization of the host to migrate out;

Step 4, find out the low-load host according to the threshold-based minimum CPU utilization method;

Step 5, reallocate the VM queue to be migrated according to the energy-aware optimal descending method.

As preferably, step 2 comprises the following steps:

Step 2.1, calculate the weighted regression curve at the current moment according to the historical CPU utilization of the host;

Step 2.2, calculate the CPU utilization forecast value of the host at the next moment according to the weighted regression curve;

Step 2.3, judge whether the host will be overloaded at the next moment according to the predicted value calculated in 2.2, the specific judgment method is as follows:

(1) If the predicted value is greater than or equal to 0.9, it is determined that the host is overloaded at the next moment;

(2) If the predicted value is less than 0.9, it is determined that the host is not overloaded at the next moment.

As preferably, step 3 comprises the following steps:

In step 3.1, the CPU utilization of each VM on the host computer predicted to be overloaded is formed into a matrix to calculate the complex correlation coefficient;

In step 3.2, according to the multiple correlation coefficients of each VM obtained in 3.1, select the VM with the largest value and add it to the queue of VMs to be migrated.

As preferably, step 4 comprises the following steps:

Step 4.1, find out the host with the lowest CPU utilization in the cloud computing center;

In step 4.2, compare the CPU utilization of the host with the optimal threshold of 0.45 obtained through a large number of experiments, and judge whether the host is under light load. The specific judgment method is as follows:

(1) If the host CPU utilization rate is greater than or equal to 0.45, it is determined that it is not in a light load state;

(2) If the host CPU utilization rate is less than 0.45, it is determined that it is in a light load state.

In step 4.3, hosts determined to be lightly loaded are added to the queue to be migrated.

Compared with the prior art, the present invention has the following advantages:

While maximizing the reduction of energy consumption brought about by the online migration of virtual machines, it is possible to avoid the "over-migration" phenomenon that occurs due to no restrictions in the light-load determination process. Therefore, compared with previous research results, this paper The inventive migration strategy can further reduce energy consumption, and is far superior to existing migration strategies in terms of service quality. The method of the invention can significantly reduce the energy consumption of the cloud computing center, and has a lower SLA violation rate than the previous algorithm, that is, better service quality.

Description of drawings

FIG. 1 is a flowchart of a method for online migration of an entire virtual machine.

Fig. 2 is a schematic diagram of the comparison experiment results of the method power consumption of the present invention with no algorithm and the best algorithm studied by predecessors;

Fig. 3 is a schematic diagram of the comparison experiment results of the service quality standard of the method of the present invention: SLA violation rate and the previous best algorithm.

detailed description

In order to make the purpose, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings.

The present invention relates to an energy-efficient migration strategy for virtual machine migration. Taking a cloud data center with 800 physical nodes as an example, all hosts are simulated as Huawei Fusion Server RH2288H. The running simulation program is CloudSim 3.0.2, and the simulation data used are ten days of data randomly collected from the PlanetLab project, that is, a total of ten sets of workloads. Specific steps are as follows:

Step 1. Record the CPU utilization rate of each host and the CPU utilization rate of each VM in each host once per second;

Step 2, using the local weighted regression method to predict whether each host is overloaded at the next moment;

Step 2.1, calculate the local weighted regression curve at the current moment according to the historical CPU utilization of the host Among them, a and b are obtained by the method of least squares (1)

$\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; bx</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

Where n is the number of recorded historical CPU utilizations, x _i is the i-th time, and y _i is the CPU utilization of the host at the i-th time. The weight function w _i (x) is obtained from (2)

${\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; (</span>\frac{{\mathrm{<span\; class="notranslate">\; \&Delta;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; \&Delta;</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>{<span\; class="notranslate">\; (</span>\frac{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 3</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 3</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

Where x _k is the time at point k, x _i is the time at point i, Δ _i (x _k ) is the time difference from point k to point i, and Δ ₁ (x _k ) is the time difference from point k to the initial recording time.

The core T in the weight function is obtained from (3)

$\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>\left\{\begin{array}{cc}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>{<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 3</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 3</span>}}<span\; class="notranslate">\; ,</span>& <span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; 1</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; ,</span>& \mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; h</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; r</span>}\mathrm{<span\; class="notranslate">\; w</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; e</span>}\end{array}\right.<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$

where x is time.

Step 2.2, according to the weighted regression curve obtained in step 2.1 Calculate the CPU utilization forecast value of the host at the next moment

Step 2.3, judge whether the host will be overloaded at the next moment according to the predicted value calculated in 2.2, the specific judgment method is as follows:

(1) If the predicted value is greater than or equal to 0.9, it is determined that the host is overloaded at the next moment;

(2) If the predicted value is less than 0.9, it is determined that the host is not overloaded at the next moment;

Step 3, through step 2, select the VM with the largest complex relationship coefficient with the CPU utilization of the host on the host that is determined to be overloaded;

In step 3.1, the CPU utilization of each VM on the host computer that is predicted to be overloaded is formed into a matrix to calculate the complex correlation coefficient. Taking a VM as an example, the CPU utilization history composition matrix of other VMs on the same host is expressed as X1, X2,...Xn, and the CPU utilization history of this VM is expressed as Y, as shown in (4).

In the X matrix, take any element x _{a, b} as an example, where a is the a-th virtual machine, and b is the b-th moment. Then x _{a, b} is the CPU utilization rate of the a-th virtual machine at time b. y _n is the CPU utilization rate of the currently evaluated virtual machine at the nth moment.

Then the complex correlation coefficient R ² is shown in (5).

${\mathrm{<span\; class="notranslate">\; R</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; \&lsqb;</span>{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&rsqb;</span>}{{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; (</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; (</span>{\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$

Among them, y _i is the CPU utilization rate of the virtual machine currently evaluated at time i, The average value of the CPU utilization of the currently evaluated virtual machine.

in The value is shown in (6)

$\begin{array}{cc}\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; x</span>}\mathrm{<span\; class="notranslate">\; b</span>}& \mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; =</span>{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}\mathrm{<span\; class="notranslate">\; the\; y</span>}\end{array}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ;</span>$

Step 3.2, according to the complex correlation coefficient of each VM obtained in 3.1, select the VM with the largest value and add it to the VM queue to be migrated;

Step 4, find out the low-load host according to the threshold-based minimum CPU utilization method;

Step 4.1, find out the host with the lowest CPU utilization in the cloud computing center;

In step 4.2, compare the CPU utilization of the host with the optimal threshold of 0.45 obtained through a large number of experiments, and judge whether the host is under light load. The specific judgment method is as follows:

(1) If the host CPU utilization rate is greater than or equal to 0.45, it is determined that it is not in a light load state;

(2) If the host CPU utilization rate is less than 0.45, it is determined that it is in a light load state;

In step 4.3, hosts determined to be lightly loaded are added to the queue to be migrated.

Step 5, reallocate the VM queue to be migrated according to the energy-aware optimal descending algorithm. The algorithm process is as follows:

The following is a detailed analysis based on the experimental results:

The main purpose of the virtual machine online migration strategy of the present invention is to reduce power consumption and ensure service quality as much as possible, wherein the service quality is measured by SLA (Service-Level Agreement, Service Level Agreement) violation rate. As shown in Figure 2: dvfs is a low-power strategy for cloud computing centers without online migration of virtual machines, and lr_mmt_mu is the best strategy based on online migration of virtual machines in previous studies. As can be seen from the figure, In all ten groups of Workloads, the power consumption of the present invention is far lower than that of dvfs, and also slightly lower than lr_mmt_mu. The impact of the virtual machine online migration strategy of the present invention on the quality of service is shown in Figure 3: the SLA violation rates of all ten groups of Workloads are far lower than lr_mmt_mu, which means that the strategy of the present invention is also far from the quality of service superior to previous research.

The above embodiments are only exemplary embodiments of the present invention, and are not intended to limit the present invention, and the protection scope of the present invention is defined by the claims. Those skilled in the art can make various modifications or equivalent replacements to the present invention within the spirit and protection scope of the present invention, and such modifications or equivalent replacements should also be deemed to fall within the protection scope of the present invention.

Claims (4)

1. A scheduling method for online migration of a virtual machine, characterized in that, comprising the steps: Step 1. Record the CPU utilization rate of each host and the CPU utilization rate of each VM in each host once per second; Step 2. Predict whether each host is overloaded at the next moment; Step 3. On the host that is determined to be overloaded, select the VM that has the largest correlation coefficient with the CPU utilization of the host to migrate out; Step 4. Find the low-load host according to the threshold-based minimum CPU utilization method; Step 5. Redistribute the VM queue to be migrated according to the energy-aware optimal descending method. 2. The scheduling method of virtual machine online migration as claimed in claim 1, wherein step 2 comprises the steps of: Step 2.1, calculate the weighted regression curve at the current moment according to the historical CPU utilization of the host; Step 2.2, calculate the CPU utilization forecast value of the host at the next moment according to the weighted regression curve; Step 2.3, judge whether the host will be overloaded at the next moment according to the predicted value calculated in 2.2, the specific judgment method is as follows: (1) If the predicted value is greater than or equal to 0.9, it is determined that the host is overloaded at the next moment; (2) If the predicted value is less than 0.9, it is determined that the host is not overloaded at the next moment. 3. The scheduling method of virtual machine online migration as claimed in claim 1, wherein step 3 comprises the steps of: Step 3.1, the CPU utilization of each VM on the host computer that is predicted to be overloaded is respectively formed into a matrix to calculate the complex correlation coefficient; Step 3.2, according to the multiple correlation coefficients of each VM obtained in 3.1, select the VM with the largest value and add it to the queue of VMs to be migrated. 4. The scheduling method of online migration of virtual machines as claimed in claim 1, wherein step 4 comprises the steps of: Step 4.1, find out the host computer with the lowest CPU utilization rate in the cloud computing center; Step 4.2. Compare the CPU utilization of the host with the optimal threshold of 0.45 obtained through a large number of experiments, and judge whether the host is under light load. The specific judgment method is as follows: (1) If the host CPU utilization rate is greater than or equal to 0.45, it is determined that it is not in a light load state; (2) If the host CPU utilization rate is less than 0.45, it is determined that it is in a light load state; Step 4.3, adding the hosts determined to be lightly loaded into the queue to be migrated.