CN105955795A - Virtual machine migration method facing green cloud computing - Google Patents

Abstract

The invention discloses a virtual machine migration method facing green cloud computing; the method combines with physics gravitation correlation concepts according to cloud data center server disposition and temperature control mode real conditions, and defines concepts like logic gravitation, correction factor and correction gravitation between physics nodes and the virtual machines; the logic gravitation gathers virtual machines in sequence onto certain data node or data nodes of adjacent areas, thus closing idle nodes; the correction factor can realize virtual machine diffusion on locally heated data nodes, thus stabilizing the system, and preventing node system from collapsing or being damaged; the correction gravitation can preferably gather the virtual machines onto the data node with good thermal diffusivity according to the heat distribution condition and thermal diffusion performance priority strategy in the cloud data center, thus realizing heat distribution balance, and preventing overheating of a frame portion with poor thermal diffusivity; the method can obviously reduce data center energy consumption, can gather the virtual machines, can close partial computing nodes and related refrigeration nodes, thus maximumly saving energy consumption.

Description

A kind of virtual machine migration method of Oriented Green cloud computing

Technical field

The present invention relates to the virtual machine migration method of a kind of Oriented Green cloud computing, be specifically related to a kind of towards cloud computation data center The energy conservation optimizing method of Virtual Machine Manager, belongs to cloud computing applied technical field.

Background technology

Cloud computing be one utilize the Internet realize whenever and wherever possible, on-demand, easily access share resource pool (as calculate facility, deposit Storage equipment, application program etc.) computation schema, it makes full use of virtualization mechanism, allows client is on-demand obtains the resource needed, by This can reduce medium-sized and small enterprises' hardware maintenance cost etc..Current China's Various types of data center total amount about 430,000, can accommodate service Device about 5,000,000, the five-year China will increase by 7～10 times to the demand of data centralized traffic disposal ability.But, cloud number According to the continuous expansion of center scale, thousands of calculating node also brings the problem of high energy consumption, with an average merit of station server Consuming 400 watts of calculating, year power consumption total amount is about 127,600,000,000 kilowatt hours, exceedes the pipe of reinforced concrete at Three Gorges Power Station generating total amount of a year already (2014 It is 98,800,000,000 kilowatt hours).

Cause this high energy consumption phenomenon Producing reason not just merely because needed for hardware resource energy consumption high, be more because these The poor efficiency of resource uses.Cloud computation data center server utilization is seldom close to 10%, and works as server utilization less than 50% Time, extra cost will be produced because of transition supply.Thus it is guaranteed that server Efficient Operation is necessary.Existing grind Study carefully proposition, the server of low-power mode can be closed by migrating virtual machine, reach energy-saving effect.

In order to effectively promote the service performance of cloud computing system, reducing service cost and energy resource consumption, efficient virtual machine is supervised simultaneously Control, manage, dispose, integrate and migration mechanism and strategy most important.The administrative mechanism of current virtual machine and deployment strategy etc. There has been an a series of achievement in research in field, but these achievements in research generally exist following part or all of problem: 1) only examine Consider one or several the factors therein such as performance, energy consumption, stability, cost, environment；2) only carry out in units of back end Deploying virtual machine and migration, really do not implement resource integration in cloud data center global scope, and active node is in cloud data The discreteness at center causes the precisely refrigeration of compartmentalization to be difficult to carry out；3) the most logically consider that the deployment of virtual machine needs with migration etc. Ask, but do not account for the actual deployment scenario of physical node of cloud data center, heat distribution and the actual fortune of temperature control system Operation mode.Linkage poor between calculating equipment and temperature control device causes achievement to be difficult in actual environment obtaining applications well.

Summary of the invention

The technical problem to be solved is: provides the virtual machine migration method of a kind of Oriented Green cloud computing, examines from the overall situation Consider, by the orderly gathering of virtual machine, it is to avoid because some node calculates the office that task is overweight or heat dispersion is poor and causes Portion is overheated.

The present invention solves above-mentioned technical problem by the following technical solutions:

The virtual machine migration method of a kind of Oriented Green cloud computing, comprises the steps:

Step 1, carries out performance detection every Preset Time to the server in cloud computation data center, and according to performance detection knot Fruit calculates the logic quality of each server, classifies server according to logic quality, predetermined threshold value γ₁、γ₂、γ₃, and γ₁＜ γ₂＜ γ₃, as logic quality≤γ₁Time, server classification is A；Work as γ₁＜ logic quality≤γ₂Time, server classification is B； Work as γ₂＜ logic quality≤γ₃Time, server classification is C；As logic quality ＞ γ₃Time, server classification is D；When cloud computing number When server classification in the minds of according to meets one of them in following condition, the virtual machine in cloud computation data center is moved Move；Described condition is: condition 1 all exists for A, B class server；Condition 2 all exists for B, D class server；Condition 3 All exist for A, B, D class server；

The logic mass formula of definition server is:

$\begin{array}{cc}M\left(s\right)=\underset{l\∈J}{\Σ}{k}_l-\underset{\∀v}{\Σ}M\left(v\right)-\mathrm{\α}\×T\left(t\right)& J\∈\{CPU,RAM,DISK,I/O\}\end{array},$

Wherein,M (s), M (v) represent the logic quality of server, virtual machine, k respectively_lFor warp Test coefficient, U_lFor virtual machine relative to the occupation rate of a certain resource of server, α is correction factor, and T (t) is the temperature of server Situation function,Represent all virtual machines in server s；

Step 2, selects the target frame of virtual machine, concretely comprises the following steps:

Step21: obtain D class server, A class server list, isolates needing the virtual machine migrated in D class server Come, make the classification of D class server revert to C, when the server classification in cloud computation data center meets condition 1, it is not necessary to Separate；

Described D class server will need the detailed process separated of virtual machine migrated be: all by D class server The logic quality of virtual machine is descending to be ranked up, and from the beginning of logic quality is maximum, is separated by virtual machine successively, often Separating once, whether the classification calculating current server returns as C, terminates if it is, separate, and otherwise continues to separate；

Step22: calculate the logic quality of each frame, is ranked up the virtual machine separated by logic quality is descending, The logic gravitation of the virtual machine to separating of the logic each frame of quality parallel computation according to frame, the machine that logic gravitation is maximum Frame is the target frame of virtual machine；

Step23: be ranked up by logic quality is ascending by the virtual machine in A class server, according to the logic quality of frame The serial computing each frame logic gravitation to the virtual machine in A class server, the frame of logic gravitation maximum is virtual machine Target frame；

The logic mass formula of definition frame is:

$M\left(f\right)=\underset{\∀s}{\Σ}M\left(s\right),$

Wherein, M (f) represents the logic quality of frame,Represent the Servers-all in frame f；

The logic gravitation formula of virtual machine is by definition frame:

$\stackrel{\→}{F}(f,v)=G\frac{M\left(f\right)M\left(v\right)}{L{(f,v)}^b},$

Wherein, (f, v) represents that virtual machine v arrives the data transmission cost that frame f produces to L, and G, b are constant coefficient；

Step 3, selects the destination server of virtual machine, concretely comprises the following steps:

B, D class server logic gravitation to virtual machine in the target frame that Step31: calculation procedure 2 obtains, and to logic Gravitation is modified, and calculates the size and Orientation of gravitation after revising；

Step32: after correction on the extended line of gravity direction, the B class server nearest apart from virtual machine is destination server；

The logic gravitation formula of virtual machine is by definition server:

$\stackrel{\→}{F}(s,v)=G\frac{M\left(s\right)M\left(v\right)}{L{(s,v)}^b},$

Wherein, (s, v) represents that virtual machine v arrives the data transmission cost that server s produces to L, and G, b are constant coefficient；

Step 4, carries out dynamic migration to needing the virtual machine migrated；

Step 5, according to the logic quality of server in current cloud computation data center, classifies to server, works as cloud computing When server classification in data center meets one of them in following condition, migration terminates, and otherwise, repeats step 1-step 4；Described condition is: condition 1 is that the classification of Servers-all is C；Condition 2 is that the classification of Servers-all is B；Bar Part 3 is that the classification of server only has A and C；Condition 4 is that the classification of server only has B and C；Condition 5 is the class of server Zhi You D.

As the preferred version of the present invention, described in step 4, the method for dynamic migration is pre-copy method.

As the preferred version of the present invention, L described in step 2 (f, formula v) is:

$L(f,v)=\underset{0}{\overset{y_1-1}{\Σ}}\frac{B\left(v\right)}{B_F\left(h_y\right)}+\underset{0}{\overset{y_2-1}{\Σ}}\frac{B\left(v\right)}{B_F\left(r_y\right)}+1,$

Wherein, B (v) represents the bandwidth that virtual machine v takies, B_F(h_y) represent switch h_yIdle bandwidth, B_F(r_y) represent router r_y Idle bandwidth, y₁The number of the switch that expression virtual machine is transferred through, y₂Represent the router that is transferred through of virtual machine Number.

As the preferred version of the present invention, after Step32 described in step 3, when the B class server that distance virtual machine is nearest have to When few one, select the server near ground as destination server.

As the preferred version of the present invention, after revising described in step 3Step31, the formula of gravitation is:

${{\stackrel{\→}{F}}^{\′}}_s\left(v\right)=\underset{i=1}{\overset{i=y}{\Σ}}\stackrel{\→}{F}(s,v)+\stackrel{\→}{G}\left(v\right),$

Wherein,Represent gravitation after revising,Represent and revise gravitation, i=1 ..., y represents 1 ..., y server.

The present invention uses above technical scheme compared with prior art, has following technical effect that

1, major part deploying virtual machine mechanism only carries out deploying virtual machine and migration in units of back end at present, the most real Implemening resource integration in cloud data center global scope, active node causes the accurate of compartmentalization in the discreteness of cloud data center Refrigeration is difficult to carry out；The present invention considers from the overall situation, by the orderly gathering of virtual machine, it is to avoid because some node calculates task mistake Weight or heat dispersion is poor and the hot-spot that causes.

2, the demand such as the machine-processed deployment the most logically considering virtual machine of major part deploying virtual machine at present and migration, does not accounts for cloud The deployment scenario of physical node, heat distribution and the actual operation pattern of temperature control system that data center is actual, calculates equipment and temperature Linkage poor between control equipment causes achievement to be difficult in actual environment obtaining applications well；The present invention is from data center's physics Framework is started with, and more agrees with the physical structure at real data center, reduces the probability of hot-spot.

3, the present invention significantly reduces consumption of data center, assembles virtual machine, thus closes part and calculate node and associated refrigeration joint thereof Point, thus save energy consumption as far as possible.

4, deploying virtual machine mechanism effect stability of the present invention, have employed the methods such as control of queue, avoids virtual machine (vm) migration as far as possible Jolt problem, it is ensured that user experience quality is good.

Accompanying drawing explanation

Fig. 1 is the flow chart of the virtual machine migration method of Oriented Green cloud computing of the present invention.

Fig. 2 (a)-Fig. 2 (c) is the schematic flow sheet that in the inventive method, virtual machine selects frame.

Fig. 3 is that in the inventive method, virtual machine is corrected the revised force diagram of gravitation.

Fig. 4 (a)-Fig. 4 (d) is the schematic flow sheet that in the inventive method, virtual machine selects server in frame.

Detailed description of the invention

Embodiments of the present invention are described below in detail, and the example of described embodiment is shown in the drawings.Below with reference to attached The embodiment that figure describes is exemplary, is only used for explaining the present invention, and is not construed as limiting the claims.

In order to realize the target such as cloud data center load balancing, energy-conservation, SLA, stability, need proposition the most virtual Machine is disposed, is migrated and management strategy.The present invention server disposition from cloud data center and the practical situation of temperature control mode, with thing Reason gravitation related notion combines, and defines the logic gravitation between physical node and virtual machine, modifying factor, correction gravitation etc. general Read, explore its computational methods, it is proposed that the virtual machine migration method of a kind of Oriented Green cloud computing.Logic gravitation realizes void Plan machine gathers on some or the back end of adjacent area in order, in order to close idle node；Modifying factor realizes local Virtual machine diffusion on overheated back end, in order to stabilisation systems, prevents system crash or the damage of node, ensures Service Quality Amount and SLA；Revise the gravitation heat distribution situation according to cloud data center, with temperature lower node preference strategy by virtual machine Preferentially converge, to realize heat distribution equilibrium, it is to avoid the generation of frame overheat condition to the frame back end that Current Temperatures is relatively low.

The virtual machine migration method that the present invention proposes is applicable to the Virtual Machine Manager of data center, and its process structure is as shown in Figure 1. By the global monitoring of cloud computing center, acquisition virtual machine (vm) migration requires, application data sheet is existing, node resource service condition, void The data such as plan machine service condition, loading condition, then pass through Surveillance center's analysis to obtaining message, it is thus achieved that virtual machine (vm) migration The destination server that target virtual machine is corresponding, then carry out the dynamic migration of virtual machine.

Need to arrange that server can bear currently takies resource and temperature upper limit threshold value, it is ensured that virtual machine creating or migration Time, the server of selection is upstate.And the server that utilization rate is too low, consider to migrate the virtual machine run on it, It is then shut off this server with energy-conservation.Therefore, data center's Servers-all is divided into four classes, class definition such as table 1.

Table 1 classification server

A class	The virtual machine (vm) migration wherein run can be walked subsequently into resting state by current server
		B class	Current server can accept virtual machine more, to reach the state of high-efficiency operation
C class	Current server is scheduling operation without the virtual machine run on which
		D class	Current server temperature is too high or calculates over loading, needs badly and is walked by some virtual machine (vm) migrations on it

Because the resource occupation related to, temperature are all relevant with logic quality, then threshold value can unify to use the logic quality of server Value delimit, classification server assignment formula is as follows:

Wherein, C (s_ti) represent server s_tiClassification, M (s_ti) represent server s_tiLogic quality, γ₁、γ₂、γ₃All represent threshold Value, and γ₁＜ γ₂＜ γ₃, γ₁、γ₂、γ₃It is to build situation with different data centers and change.In practical operation, Should model according to the situation of data center, γ₁、γ₂、γ₃Taking different values, dry run obtains optimal values.Or suggestion Use γ₁=0.3, γ₂=0.5, γ₃=0.85 calculates.

Understand technical scheme for convenience, below some concepts be defined:

Definition 1: logic quality is the measure formulas that data center's middle finger earnest reason or object logic take resource size, M (x) Represent the logic quality of object x.

The present invention includes the logic quality of virtual machine, the logic quality of server, the logic quality of frame, is defined as follows:

$\left\{\begin{array}{cc}M\left(v_{ij}\right)=\underset{l\∈J}{\mathrm{\Σ}}{k}_l\×U_l\left(v_{ij}\right)& J\∈\{CPU,RAM,DISK,I/O,...\}\\ M\left(s_{ti}\right)=\underset{l\∈J}{\mathrm{\Σ}}{k}_l-\underset{\∀v_{jk}}{\mathrm{\Σ}}M\left(v_{ij}\right)-\mathrm{\α}\×T\left(t_i\right)& J\∈\{CPU,RAM,DISK,I/O,...\}\\ M\left(f_t\right)=\underset{\∀s_{ti}}{\mathrm{\Σ}}M\left(s_{ti}\right)& \end{array}\right.---\left(2\right)$

M (v in formula (2)_ij) that calculate is the i-th station server s_tiJth platform virtual machine v_ijThe summation of the resource taken, wherein, k_lFor Empirical coefficient, U_lFor virtual machine relative to the occupation rate of this resource of server, it is however generally that, which kind of service this server belongs to Intensive server, empirical coefficient corresponding to this service is larger.By M (v_ij) computing formula can by the middle of virtual machine not The utilization of resources amount of commensurate is unified numerical value and is tried to achieve unified evaluation function, is the logic quality of virtual machine.

M (s in formula (2)_ti) that calculate is t base framework f_tThe i-th station server s_tiLogic quality, wherein, α is correction factor, That this formula calculates is server s_tiThe summation of idling-resource, through the correction of modifying factor, the logic quality of the server obtained, Performance is the follow-up attraction power to virtual machine.In the middle of real data center, modifying factor T (t_i) it is for reacting current thing The function of reason node temperature situation, T (t_i) numerical value equal to the real-time monitoring value of cloud data center.When in real time, monitoring cost is excessive, Or during analog systems, can be to T (t_i) carrying out numerical prediction, its function is:

T(t_i)=t_i+β(t_i) (3)

Wherein, t_iFor by testing and assessing this server performance thus the prediction that obtains calculates function, β (x) is deviation value correction.

M (f in formula (2)_t) that calculate is t base framework f_tThe summation of idling-resource, i.e. frame f_tLogic quality.

Definition 2: when logical reach refers to two physical node data transmission in data center, because switch, router etc. produce Data transmission cost.L(x,v_ij) represent virtual machine v_ijArriving the data transmission cost that object x produces, its formula is:

$L(x,v_{ij})=\underset{0}{\overset{y_1-1}{\Σ}}\frac{B\left(v_{ij}\right)}{B_F\left(h_y\right)}+\underset{0}{\overset{y_2-1}{\Σ}}\frac{B\left(v_{ij}\right)}{B_F\left(r_y\right)}+1---\left(4\right)$

Wherein, B (v_ij) it is virtual machine v_ijThe bandwidth taken, B_F(h_y) it is switch h_yIdle bandwidth, B_F(r_y) it is router r_y Idle bandwidth, y₁For the number of the switch that virtual machine is transferred through, y₂Number for the router that virtual machine is transferred through.

Definition 3: logic gravitation refers to any one frame has logic gravitation to any virtual machineSucking action.Greatly Little and they logic quality products are directly proportional and square being inversely proportional to of their logical reach, with two articles in data center concrete Distances etc. are unrelated, and frame is pointed to by virtual machine in direction.

Definition 4: revise gravitationRefer to virtual machine due to temperature reason, tended to frame good heat dispersion performance side To captivation.

Frame f_t, server s_tiWith virtual machine v_ijLogic gravitation computing formula be (wherein G, b are constant coefficient):

$\left\{\begin{array}{c}\stackrel{\→}{F}(f_t,v_{ij})=G\frac{M\left(f_t\right)M\left(v_{ij}\right)}{L{(f_t,v_{ij})}^b}\\ \stackrel{\→}{F}(s_{ti},v_{ij})=G\frac{M\left(s_{ti}\right)M\left(v_{ij}\right)}{L{(s_{ti},v_{ij})}^b}\end{array}\right.---\left(5\right)$

For virtual machine v_ijBy x frame data center make a concerted effort be:

${\stackrel{\→}{F}}_f\left(v_{ij}\right)=\underset{t=0}{\overset{t=x-1}{\Σ}}\stackrel{\→}{F}(f_t,v_{ij})---\left(6\right)$

Y given server is to virtual machine v_ijThrough correction make a concerted effort be:

The virtual machine (vm) migration flow process of the present invention is:

Step one, the startup of moving method

Virtual machine in cloud computing system is due to random unlatching, closedown, and through after a period of time, virtual machine will be gradually in discrete Changing distribution, whole system will appear from the state that focus is chaotic, unordered, and the accurate temperature control of localization will be difficult to carry out, and at this moment need Virtual machine is redeployed, carries out focus type migration.Preferably migration results is clearly and will remain the region that a small amount of virtual machine runs Virtual machine (vm) migration on node cluster to carry relatively multi-dummy machine Area Node cluster on, thus can be by subregion set of node The virtual machine quantity of group's carrying drops to 0, in order to dormancy in time or closedown.

If the back end in region is overheated or overload, in order to ensure operation stability and the service quality of system, virtual Machine needs to be diffused formula and migrates, and will move to other node by the virtual machine node that closed on the upper limit from temperature or resource consumption. And which virtual machine to be moved out the some virtual machines carried from node select be to need first in view of problem.Virtual machine spreads Desired result should be: after going out the fewest virtual machine with minimum cost fast transferring, temperature or the load of node can be returned Arrive acceptable degree again.

Final purpose of the present invention is the state allowing calculating node all enter C class high-efficiency operation as far as possible, and as far as possible by low speed The server closing of operating, when running into situations below when, it is necessary to considers to migrate virtual machine: in (1) data center, A, B class server all exists；(2), in data center, B, D class server all exists；(3) in data center, A, B, D Class server all exists.

Realize this monitoring, for monitored server, need t at set intervals to carry out a performance detection, once temperature mistake High, utilization rate is too high or utilization rate is too low it is necessary to send, to monitoring machine, the signal migrate needing to migrate virtual machine.And supervise Control machine once by needing the signal migrate being adjusted virtual machine more than threshold value ω, then starts the inventive method.

Step 2, selection target frame

Step1: obtain D class server, A class server list, calculate the logic quality of each frame, isolate all D A ' the part of class server so that after virtual machine (vm) migration leaves in A ', D class server can revert to C class；Feelings when step one When condition is (1), it is not necessary to separate, wherein, concretely comprising the following steps of A ' class server is isolated:

A. the logic quality of all for D apoplexy due to endogenous wind virtual machines is sorted by descending order, numbering write queue L；

B. the virtual machine deducting queue L row head sequence number corresponding in current D class server total resources being taken resource, measuring and calculating deducts After D class server by utilizing situation whether reduce to C class, and the queue L row head sequence number write queue M that will deduct；The most total Resource refers to logical china and measures the factor taken into account when calculating, such as CPU, RAM etc.；

If c. D class server will reduce to C class through prediction, then stopping measuring and calculating, in queue M, all numbering correspondences is virtual Machine forms this D isolated corresponding A of class server ' class server；

If d. D class server does not reduce to C class through prediction, return to b.

Step2: carry out priority list to the virtual machine logically quality of all A ' class servers, A class server is descending Sort, and the virtual machine priority of A ' apoplexy due to endogenous wind is above A class, shown in the original state of current whole system such as Fig. 2 (a), Wherein F1 is the frame 1 logic gravitation to virtual machine, and F2 is the frame 2 logic gravitation to virtual machine；If the situation of step one For (2), then without A class server is operated；

Step3: calculate each frame gravitation to virtual machine to be migrated, what gravitation was maximum is target frame；

Step4:A ' class server list parallel computation, Servers-all traversal in queue, it currently made number one is virtual Machine determines target frame, A class serial computing, works as A ' class and is disposed, and is the most all processed by A class server, As shown in Fig. 2 (b), it is virtual machine (vm) migration process；

Step5: current goal virtual machine the most all migrates, and returns to Step4, otherwise terminates, as shown in Fig. 2 (c), virtual machine All arriving purpose frame, migration terminates.

Step 3, each virtual machine select destination server

For the virtual machine of selected purpose frame, need to further determine that destination server, now need to revise gravitation and be modified, Select the server at the relatively low good heat dispersion performance of Current Temperatures as much as possible.As it is shown on figure 3, be virtual machine when specifying frame, The power being subject to is painted tries hard to.Virtual machine not only receives making a concerted effort from server, also receives the impact revising gravitation.By public affairs Formula (7) is tried to achieveThe server pointed to, is target and places server, can migrate after virtual machine is selected.

As shown in Fig. 4 (a)-4 (d), Fig. 4 (a) represents that a frame wherein one has the server that m × n blockage represents；Figure 4 (b) be virtual machine be subject in this frame try hard to schematic diagram, virtual machine is set as the centre initial point calculating position in frame (0,0) place calculates stress value；Fig. 4 (c) is by the selection of destination server；The server that in Fig. 4 (d), black bars represents is i.e. For choosing server.Specifically comprise the following steps that

Step1: calculate B, D class server logic gravitation to virtual machine in target frame, anticipates when logic gravitation value is negative Think of be gravity direction be the rightabout in former measuring and calculating direction, and order of magnitude is constant, and logical reach is 1；

Step2: calculate its revised gravitation, calculates resultant direction suffered by virtual machine；

Step3: on resultant direction extended line, closest B class server is destination server.If there being more than one symbol The server of conjunction condition, then choose the server near ground.

Step 4, virtual machine (vm) migration

It is necessary to start dynamic migration after each virtual machine selected target server.Present mechanism uses pre-copy method to carry out virtual machine Migration.

Step 5, the termination of deployment mechanisms an iteration

Virtual machine (vm) migration under method decision-making, all virtual machines will gradually be assembled by data center, and deployment mechanisms is once run Needing end condition, end condition is as follows: (1) data center server is C class；(2) data center server is B class；(3) data center server only has A class and C class；(4) data center server only has B class and C class；(5) data Central server only has D class.

That once disposes completes, by the frame of low power run, data center's situation is as it can be seen, include that its refrigerating machine all falls into Enter resting state and allow other server all enter high efficiency operating condition as far as possible, thus reaching energy-conservation purpose.

Above example is only the technological thought that the present invention is described, it is impossible to limit protection scope of the present invention with this, every according to this The technological thought that invention proposes, any change done on the basis of technical scheme, within each falling within scope.

Claims (5)

1. the virtual machine migration method of an Oriented Green cloud computing, it is characterised in that comprise the steps:

Step 1, carries out performance detection every Preset Time to the server in cloud computation data center, and according to performance detection knot Fruit calculates the logic quality of each server, classifies server according to logic quality, predetermined threshold value γ₁、γ₂、γ₃, and γ₁＜ γ₂＜ γ₃, as logic quality≤γ₁Time, server classification is A；Work as γ₁＜ logic quality≤γ₂Time, server classification is B； Work as γ₂＜ logic quality≤γ₃Time, server classification is C；As logic quality ＞ γ₃Time, server classification is D；When cloud computing number When server classification in the minds of according to meets one of them in following condition, the virtual machine in cloud computation data center is moved Move；Described condition is: condition 1 all exists for A, B class server；Condition 2 all exists for B, D class server；Condition 3 All exist for A, B, D class server；

The logic mass formula of definition server is:

$\begin{array}{cc}M\left(s\right)=\underset{l\∈J}{\Σ}{k}_l-\underset{\∀v}{\Σ}M\left(v\right)-\mathrm{\α}\×T\left(t\right)& J\∈\{CPU,RAM,DISK,I/O\}\end{array},$

Wherein,M (s), M (v) represent the logic quality of server, virtual machine, k respectively_lFor warp Test coefficient, U_lFor virtual machine relative to the occupation rate of a certain resource of server, α is correction factor, and T (t) is the temperature of server Situation function,Represent all virtual machines in server s；

Step 2, selects the target frame of virtual machine, concretely comprises the following steps:

Step21: obtain D class server, A class server list, isolates needing the virtual machine migrated in D class server Come, make the classification of D class server revert to C, when the server classification in cloud computation data center meets condition 1, it is not necessary to Separate；Calculate the logic quality of each frame；

Described D class server will need the detailed process separated of virtual machine migrated be: all by D class server The logic quality of virtual machine is descending to be ranked up, and from the beginning of logic quality is maximum, is separated by virtual machine successively, often Separating once, whether the classification calculating current server returns as C, terminates if it is, separate, and otherwise continues to separate；

Step22: be ranked up by logic quality is descending by the virtual machine separated, the logic quality according to frame is parallel Calculating the logic gravitation of each frame virtual machine to separating, the frame of logic gravitation maximum is the target frame of virtual machine；

Step23: be ranked up by logic quality is ascending by the virtual machine in A class server, according to the logic quality of frame The serial computing each frame logic gravitation to the virtual machine in A class server, the frame of logic gravitation maximum is virtual machine Target frame；

The logic mass formula of definition frame is:

$M\left(f\right)=\underset{\∀s}{\Σ}M\left(s\right),$

Wherein, M (f) represents the logic quality of frame,Represent the Servers-all in frame f；

The logic gravitation formula of virtual machine is by definition frame:

$\stackrel{\→}{F}(f,v)=G\frac{M\left(f\right)M\left(v\right)}{L{(f,v)}^b},$

Wherein, (f, v) represents that virtual machine v arrives the data transmission cost that frame f produces to L, and G, b are constant coefficient；

Step 3, selects the destination server of virtual machine, concretely comprises the following steps:

B, D class server logic gravitation to virtual machine in the target frame that Step31: calculation procedure 2 obtains, and to logic Gravitation is modified, and calculates the size and Orientation of gravitation after revising；

Step32: after correction on the extended line of gravity direction, the B class server nearest apart from virtual machine is destination server；

The logic gravitation formula of virtual machine is by definition server:

$\stackrel{\→}{F}(s,v)=G\frac{M\left(s\right)M\left(v\right)}{L{(s,v)}^b},$

Wherein, (s, v) represents that virtual machine v arrives the data transmission cost that server s produces to L, and G, b are constant coefficient；

Step 4, carries out dynamic migration to needing the virtual machine migrated；

Step 5, according to the logic quality of server in current cloud computation data center, classifies to server, works as cloud computing When server classification in data center meets one of them in following condition, migration terminates, and otherwise, repeats step 1-step 4；Described condition is: condition 1 is that the classification of Servers-all is C；Condition 2 is that the classification of Servers-all is B；Bar Part 3 is that the classification of server only has A and C；Condition 4 is that the classification of server only has B and C；Condition 5 is the class of server Zhi You D.

The virtual machine migration method of Oriented Green cloud computing the most according to claim 1, it is characterised in that dynamic described in step 4 The method that state migrates is pre-copy method.

The virtual machine migration method of Oriented Green cloud computing the most according to claim 1, it is characterised in that described in step 2 L (f, formula v) is:

$L(f,v)=\underset{0}{\overset{y_1-1}{\Σ}}\frac{B\left(v\right)}{B_F\left(h_y\right)}+\underset{0}{\overset{y_2-1}{\Σ}}\frac{B\left(v\right)}{B_F\left(r_y\right)}+1,$

Wherein, B (v) represents the bandwidth that virtual machine v takies, B_F(h_y) represent switch h_yIdle bandwidth, B_F(r_y) represent road By device r_yIdle bandwidth, y₁The number of the switch that expression virtual machine is transferred through, y₂Represent the route that virtual machine is transferred through The number of device.

The virtual machine migration method of Oriented Green cloud computing the most according to claim 1, it is characterised in that described in step 3 After Step32, when the B class server that distance virtual machine is nearest has at least one, select the server near ground as mesh Mark server.

The virtual machine migration method of Oriented Green cloud computing the most according to claim 1, it is characterised in that step 3Step31 After described correction, the formula of gravitation is:

${{\stackrel{\→}{F}}^{\′}}_s\left(v\right)=\underset{i=1}{\overset{i=y}{\Σ}}\stackrel{\→}{F}(s,v)+\stackrel{\→}{G}\left(v\right),$

Wherein,Represent gravitation after revising,Represent and revise gravitation, i=1 ..., y represents 1 ..., y server.