CN104008011A - Method for balancing resource load of entity machine in cluster environment - Google Patents

Abstract

The invention discloses a method for balancing resource load of a physical machine in a cluster environment, which comprises the steps of obtaining more than one resource utilization rate of a virtual machine arranged on the physical machine in the cluster environment, and further calculating the discrete degree of the resource load of the physical machine in the cluster and the arithmetic mean of the resource load according to the resource utilization rates of the virtual machines; when the dispersion degree of the resource load of the entity machines in the cluster is higher than the load dispersion degree of a target, a plurality of candidate virtual machines are obtained through a virtual machine moving algorithm for moving, the algorithm mainly comprises the steps that the entity machines with resource liability larger than the arithmetical mean of the resource load in the cluster are taken as one group, all other entity machines are taken as the other group, the virtual machines are selected to be taken as moving candidates in pairs by the entity machines with the same distance ranking from the arithmetical mean of the load, and the dispersion degree of the resource load of the entity machines in the cluster is reduced until the load dispersion degree of the target is met.

Description

The method of the balance physical machine resource load in environment of gathering together

Technical field

The invention relates to a kind of dynamic resource allocation method, relate to especially the method for balance physical machine resource load in a kind of environment of gathering together.

Background technology

(cluster) environment of gathering together is to be mainly that different equal portions use by Intel Virtualization Technology by the resource cutting being positioned on entity machine (being commonly referred to as Host).Obtain easily needed computational resource by the virtual user of allowing or application program, be not limited to mode, area, the position of initial installation, or the combination of entity.Intel Virtualization Technology allows user to move multiple operating system in a physical machine, Intel Virtualization Technology running is (or to be commonly referred to as monitor of virtual machine (virtual machine monitor by Virtual Machine Manager layer (Hypervisor), VMM)) virtual machine (Virtual Machine) is separated with physical machine hardware, in a physical machine, comprise processor, internal memory, storage facilities, the resources such as network I/O, on it, be partitioned into several virtual machines, each virtual machine is according to the demand of its application, distribute required hardware resource, in the middle of virtual machine and hardware, be responsible for communication by Virtual Machine Manager layer, Virtual Machine Manager layer is responsible for that the instruction of virtual machine is sent to required hardware resource and is carried out running, and the result of running is sent back to virtual machine, each virtual machine is mutually independently, each other and do not know sharing the resource of hardware mutually.In other words, Intel Virtualization Technology provides the classification for data, computing power, storage volume and a logic of other resources, instead of an entity boundary.

Cluster system is made up of most physical machine, and for properly utilizing physical machine resource, each physical machine is distributed to resource more several virtual machines and used to improve resource utilization.It is indefinite that virtual machine rises and falls to the demand of resource, thereby resource utilization difference in each physical machine is large, Intel Virtualization Technology provides the function of excessive use (overcommit) by Virtual Machine Manager layer, the resource utilization of virtual machine cannot be learnt by observation, for promoting server fiduciary level and efficiency of service, avoid the too high service disruption that causes of hardware burden, the necessary starting load balancing of resource between physical machine, a virtual machine is moved to another from the physical machine of a Taiwan investment source anxiety the physical machine of surplus resources.But in prior art, can only do balance to the physical machine of same hardware specification, cannot make applicable balance result according to actual conditions, have disappearance.

Summary of the invention

The object of the present invention is to provide a kind of under the environment of gathering together the method for Dynamic Resource Allocation for Multimedia, use the basis of historical data as resource scheduling by the resources of virtual machine that obtains each entity airborne device, to reach the method for designing of physical machine resource load stabilization.In what the present invention was designed gather together environment, the method for balance physical machine resource load includes following steps: there is virtual resources managing and control system to monitor that by Virtual Machine Manager layer a plurality of virtual machines the resource representation formula with the present invention's design calculate and express the resource utilization index of those virtual machines, and the various setting that need to consider in response to different situations in the time gathering together interior computing such as pre-loaded virtual machine restraint-type, default dispersion degree critical value preference server setting in the time doing load balance.

Further, according to object of the present invention, the method for balance physical machine resource load in a kind of environment of gathering together proposing, its process step comprises: A, import gather together Critical value resources and virtual machine configuration mutual exclusion and group and impose a condition; The resource use amount of B, acquisition collected each virtual machine history of a period of time; C, close to calculate the resource utilization of physical machine reality according to the resource use amount of all virtual machines in physical machine, further try to achieve the standard deviation value under the environment of gathering together by the resource utilization of calculating all physical machine again, to judge load balanced state and dispersion degree between each physical machine; Whether the standard deviation value that D, judgement are tried to achieve is greater than the Critical value resources of gathering together, if, carry out selecting of physical machine, to pick out at least one pair of physical machine, and in two physical machine of selected pairing, select virtual machine and simulate and move action, to improve the standard deviation value of physical machine resource utilization, wherein, the resource utilization of one of them physical machine in the physical machine of matching is greater than another physical machine; E, calculate simulation and move after virtual machine, judge that whether the standard deviation value of physical machine resource utilization is still higher than the Critical value resources of gathering together, if so, repeated execution of steps D and E, if not, performs step F; And F, carry out the action of moving of virtual machine, to move mutually the virtual machine in two physical machine of selected pairing.

Preferably, dispersion degree is calculated expression by following mathematical expression:

${\mathrm{\ρ}}_j^{\mathrm{resource}}=\frac{{\mathrm{\Σi}\∈s}_j{X}_{i,j}{V}_i^{\mathrm{resource}}}{H_j^{\mathrm{resource}}};$

${\mathrm{\μ}}^{\mathrm{resource}}=\frac{1}{m}{\mathrm{\Σ}}_{j=1}^m{\mathrm{\ρ}}_j^{\mathrm{resource}};$

$\mathrm{\σ}=\sqrt{\frac{1}{m}{\mathrm{\Σ}}_{j=1}^m{({\mathrm{\ρ}}_j^{\mathrm{resource}}-{\mathrm{\μ}}^{\mathrm{resource}})}^2};$

Wherein, represent the resource utilization of a numbering j physical machine; X _{i, j}representing whether a numbering i virtual machine has in the physical machine of this numbering j starts, and its value of some words is 1, and its value of the words of not enabling is 0; the resource capability that virtual machine i uses for this reason; μ ^resourcethe arithmetic mean value of resource utilization of interior m platform physical machine represents to gather together; The σ standard deviation value that above numerical value is tried to achieve of serving as reasons, in order to express the dispersion degree of this interior resource load of gathering together.

Preferably, in above-mentioned steps D, comprise the following step: the resource utilization of calculating each physical machine with arithmetic mean value μ ^resourcedifference D _j; And according to the difference D calculating _jthe physical machine of the interior resource anxiety of gathering together is classified as to one group, and other physical machine is classified as another group, and by the physical machine of two groups according to difference D _jsize carry out rank, so that the physical machine of identical rank between two groups or a certain particular rank is picked as to the object of mutual resettlement virtual machine, and then screen the physical machine of at least one pair of pairing, wherein, the one matching is source physical machine, another is target entity machine.

Preferably, the difference D of above-mentioned physical machine _jbe greater than the group that ranges the interior resource anxiety of gathering together at 1 o'clock, difference D _jminus physical machine is attributed to another group.

Preferably, above-mentioned steps D also comprises the following step: after originate physical machine and target entity machine are picked out, calculate the use resource of source entity machine and target entity machine and the gap of average, wherein, this gap U is the numerical value that difference D is multiplied by the total capacity R of physical machine target resource; And pick out a virtual machine in the physical machine of source, the resource that this virtual machine consumes need be greater than the use resource of the physical machine of originating and the gap of average, and gap is to be less than according to the predefined numerical value of target resource.

Preferably, above-mentioned steps D also comprises the following step: the use resource of basis source physical machine and target entity machine and the gap U of average _s, U _t, calculate number Thd in resource load amount; And if in the time that the virtual machine meeting cannot be picked out in physical machine in source, in the physical machine of source, select a virtual machine and in target entity machine, pick out a virtual machine, and the selected virtual machine of source entity machine is moved to target entity machine in the future, selected target entity machine virtual machine is moved to source physical machine, to make to move the resource capability V of rear two virtual machines ^resourcedifference be able to position and exist | Ut| and | between Us| and approach Thd value most.

Preferably, number Thd=(U in above-mentioned resource load amount _s-U _t)/2.

Preferably, above-mentioned steps D also comprises the following step: if cannot pick out the virtual machine meeting in source physical machine and target entity machine time, in the physical machine of source, pick out a virtual machine, the resource capability Vresource of this virtual machine need approach number Thd in resource load amount recently.

From the above, according to the present invention it, it has following one or more advantage:

1, the resource load of balance physical machine under the environment of gathering together, not only flexible, also can at will not propose the instruction of migration in the time that VM need to work.

2, under the environment of gathering together, under the prerequisite of physical machine hardware compatibility, can calculate the configuration of load balance and need not be confined to prior art and can only do balance to the physical machine of same hardware specification the different physical machine of ability.

3, the present invention on-the-fly modifies parameter according to the difference of actual environment and demand, can resiliently make applicable balance result.

Brief description of the drawings

Fig. 1 is the schematic diagram of the embodiment of the framework that method is used of balance physical machine resource load in the environment of gathering together of the present invention;

Fig. 2 is the process flow diagram of the embodiment of the method for balance physical machine resource load in the environment of gathering together of the present invention;

Fig. 3 is the thin portion process flow diagram of step S104 in Fig. 2.

Symbol description

100 virtual resources managing and control systems

200 Virtual Machine Manager layers

300 gather together

400,500 physical machine

S101～S106, S201～S205 step

Embodiment

Effect of understanding technical characterictic of the present invention, content and advantage and can reach for the expensive auditor of profit, hereby the present invention is coordinated to accompanying drawing, and be described in detail as follows with the expression-form of embodiment, and wherein used graphic, its purport is only the use of signal and aid illustration book, may not be true ratio after the invention process and precisely configuration, thus should be with regard to appended graphic ratio with configuration relation deciphering, limit the invention to the interest field in actual enforcement, close first chat bright.

The following example is that the gather together load balance scheduling method of physical machine in computing environment has progressive resource representation formula to collect the resource utilization of virtual machine and progressive load balance scheduling method to find one to multiple suitable virtual machine, move to suitable physical machine, make physical machine reach the good load balance of efficiency.

Refer to shown in Fig. 1, for executing the system architecture diagram of the method for balance physical machine resource load in a kind of environment of gathering together of preferred embodiment of the present invention.In an embodiment, there is virtual resources managing and control system 100, major function is the various entity devices in overall management and monitoring room, and the Intel Virtualization Technology that utilizes Virtual Machine Manager layer 200 is virtual resource pond by 300 the physical machine resource conversion of gathering together, use user interface and API (the Application Programming Interface of unification, commonization, application programming interface), in order to the work of operational administrative resource and the migration of execution real-time virtual machine, reach in the function of assigning load configuration order after calculating.

Gathering together 300 is normally made up of one or more physical machine, and every physical machine all can be carried out one or more virtual machine.For the purpose of simplifying the description, below only by two physical machine as for example, wherein a physical machine 400 is referred to as H1, another physical machine 500 is referred to as H2, wherein H1 is the CPU with total tens of two core, and the server host that each core arithmetic capability quantized value is 400, on it, carry out five virtual machine (virtual machines 2, virtual machine 4, virtual machine 6, virtual machine 7, virtual machine 10), an other H2 is the CPU with total tens of six core, and the server host that each core arithmetic capability quantized value is 200, on it, carry out five virtual machine (virtual machines 1, virtual machine 3, virtual machine 5, virtual machine 8, virtual machine 9), and each physical machine 400 in this embodiment, 500 all memory sources all meet carries out in this embodiment 10 virtual machines in the memory requirements of same physical machine simultaneously.The CPU arithmetic capability quantized value that wherein virtual machine has used is as shown in table 1 below.Wherein, refer to the arithmetic capability that virtual machine i has used:

Table 1

Refer to shown in Fig. 2, for the process flow diagram of the embodiment of the method for balance physical machine resource load in a kind of environment of gathering together of preferred embodiment of the present invention, its process step is: step S101, imports the restrictive condition such as Critical value resources 5% and virtual machine configuration mutual exclusion and group's setting of gathering together.Step S102, then captures a period of time to collect the historical resource use amount of each virtual machine, the calculating of resource utilization by with CPU usage for this reason the load balance of embodiment consider.Step S103, close the resource utilization with computational entity machine reality according to the resource use amount of whole virtual machines, further try to achieve the standard deviation under the environment of gathering together by the resource utilization of calculating all physical machine, to judge load balanced state and dispersion degree between each physical machine.Wherein, dispersion degree is to calculate with mathematic(al) representation below:

${\mathrm{\ρ}}_j^{\mathrm{resource}}=\frac{{\mathrm{\Σi}\∈s}_j{X}_{i,j}{V}_i^{\mathrm{resource}}}{H_j^{\mathrm{resource}}};$

${\mathrm{\μ}}^{\mathrm{resource}}=\frac{1}{m}{\mathrm{\Σ}}_{j=1}^m{\mathrm{\ρ}}_j^{\mathrm{resource}};$

$\mathrm{\σ}=\sqrt{\frac{1}{m}{\mathrm{\Σ}}_{j=1}^m{({\mathrm{\ρ}}_j^{\mathrm{resource}}-{\mathrm{\μ}}^{\mathrm{resource}})}^2};$

Wherein the resource utilization that represents a numbering j physical machine, is the arithmetic capability that virtual machine i has used, X in this embodiment _{i, j}representing whether a numbering i virtual machine has in the physical machine of this numbering j starts, and its value of some words is that 1 its value of the words of not enabling is 0, the resource capability that virtual machine i uses for this reason, μ ^resourcerepresent the to gather together arithmetic mean value of resource utilization of interior m platform physical machine, σ serves as reasons standard deviation that above numerical value tries to achieve in order to express the dispersion degree of this interior resource load of gathering together, in this embodiment σ=15.8854%.

The standard deviation calculating, higher than the critical value 5% of original definition, therefore performs step S104, selects two suitable physical machine, and selection method is to carry out following steps: calculate the resource utilization of every physical machine and the difference D of mean value _j, its formula is:

$D_j={\mathrm{\ρ}}_j^{\mathrm{resource}}-{\mathrm{\μ}}^{\mathrm{resource}};$

After obtaining the difference of every physical machine, the physical machine of the interior resource anxiety of gathering together can be classified as to one group and be referred to as SHs, being characterized as of this group object machine:

D _j>0；

Other physical machine is other one group of SHt.In the SHs of physical machine group and SHt according to D _jlarge small distance rank, and the physical machine circle of identical rank in two groups is elected as to the object of mutual resettlement virtual machine, or a pair of physical machine of only choosing the name that ranks the first according to different demands is to move mutually the object of virtual machine, namely in gathering together, finds the use of more than one pair of physical machine for load balance.After paired physical machine has been matched, the physical machine of title resource anxiety is for source physical machine Hs and claim that another physical machine is target entity machine Ht, and in this example, Hs is physical machine 500, namely H2, and Ht is physical machine 400, namely H1.Obtain its gap of the resource that uses and average, its mathematic(al) representation is expressed as follows:

U _t＝D _t*R _t；

U _s＝D _s*R _s；

The total capacity that is namely multiplied by this physical machine target resource with difference, this embodiment draws D ₁=-15.8854%; D ₂=15.8854% result, U _s=508.33 and U _t=-762.5; Then obtain number Thd in a resource load amount, its mathematic(al) representation is expressed as follows:

Thd＝(U _s-U _t)/2；

Then in these two physical machine, selecting virtual machine simulates and moves action that physical machine resource utilization standard deviation value is improved is maximum, namely connect most or lower than the critical value of nearly original definition, and then execution step S105, calculating simulation moves after virtual machine, whether the standard deviation value of physical machine resource utilization is higher than critical value, if higher than repeating step S104 and S105 of critical value, otherwise execution step S106 carries out the action of moving of the virtual machine picked out by step S104.

Fig. 3 is the thin portion process flow diagram of step S104.When execution step, S104 will disassemble into step S201～S205.Step S201 is the difference of calculating the load balance of each portion physical machine target.The result that step S202 calculates based on step S201, selects two optimal physical machine Hi and Hj, and the resource utilization of Hi must be greater than the resource utilization of Hj.Step S203 is virtual machine selection method one, by in the method for selecting virtual machine by select an optimal virtual machine VM on Hs, this VM is characterized as consumed resource and is greater than source physical machine and uses gap and the gap of resource and average to be less than according to the predefined numerical value of target resource, makes to move the load balance the most that can make total system after VM.If step S203 cannot select a suitable virtual machine, adopt step S204 virtual machine selection method two to select.Step S204 is by respectively find an optimal virtual machine VMi and VMi ' on physical machine Hs and physical machine Ht, make virtual machine VMi move to physical machine Ht and VMi ' and move to physical machine Hs, can make the load balance the most of total system, it is characterized by difference after moving not only exists position | Ut| and | between Us| but also approach Thd value most, its mathematic(al) representation is expressed as follows:

$\left|\mathrm{Ut}\right|\≤|V_i^{\mathrm{resource}}-V_{i^{\′}}^{\mathrm{resource}}|\≤\left|\mathrm{Us}\right|,\left(\mathrm{if}\right|\mathrm{Ut}|\≤|\mathrm{Us}\left|\right)\mathrm{or}\left|\mathrm{Us}\right|$ $\≤|V_i^{\mathrm{resource}}-V_{i^{\′}}^{\mathrm{resource}}|\≤\left|\mathrm{Ut}\right|,\left(\mathrm{if}\right|\mathrm{Us}|\≤|\mathrm{Ut}\left|\right);$

$({\mathrm{VM}}_i,{\mathrm{VM}}_{i^{\′}})=\arg \underset{{i\∈S}_s^{\mathrm{movable}},i^{\′}{\∈S}_t^{\mathrm{movable}},}{\min }\left(\right|\mathrm{Thd}-|V_i^{\mathrm{resource}}-V_{i^{\′}}^{\mathrm{resource}}|\left|\right);$

$(V_i^{\mathrm{resource}}>V_{{\′}^{\′}}^{\mathrm{resource}})$

If step S204 cannot select two suitable virtual machines, adopt step S205 virtual machine selection method three to select.Step S205 is by select an optimal virtual machine VMh on physical machine Hj, make to move after VMh, can make the load balance the most of total system, it is characterized by the virtual machine of selecting resource load use amount and approach most in resource load amount number in the middle of Hs, its mathematic(al) representation is expressed as follows:

$\arg \underset{{i\∈S}_s^{\mathrm{movable}}}{\min }\left(\right|\mathrm{Thd}-V_i^{\mathrm{resource}}\left|\right);$

Step from the above mentioned can be selected virtual machine 8 for meeting the applicable resettlement virtual machine of above method in method one, then utilizes the virtualized technology of Virtual Machine Manager layer that virtual machine is moved to H2 from H1; The standard deviation value of moving the load of gathering together after virtual machine is reduced to 2.6042% can make all physical machine resource load dispersion degrees lower than moving front 15.8854% and be less than critical value 5%, stop moving calculating, and transfer to virtual resources managing and control system assign migration order and then carry out virtual machine (vm) migration, to reach the target of physical machine resource load stabilization.

After selecting according to above-mentioned method the virtual machine of listing resettlement in, utilize the virtualized technology of Virtual Machine Manager layer virtual machine to be moved to the physical machine of surplus resources from the physical machine of resource anxiety; After moving virtual machine, can make all physical machine resource load dispersion degrees lower than before moving, be less than before moving and stop moving calculating until meet lower than default load dispersion degree critical value or dispersion degree, and transfer to virtual resources managing and control system assign migration order and then carry out virtual machine (vm) migration, to reach the target of physical machine resource load stabilization.

In sum, step used in the present invention can be guided and gather together computational service provider how from the useful data of the acquisition of the virtual-machine data of collecting, recycling is selected and can be made the virtual machine of physical machine resource utilization standard deviation value lower than critical value, move action, to reach the load balance design of physical machine in the computing environment of gathering together.

The foregoing is only illustrative, but not be restricted person.Anyly do not depart from spirit of the present invention and category, and equivalent modifications or change that it is carried out all should be contained in claim.

Claims (8)

1. a method for the balance physical machine resource load in environment of gathering together, is characterized in that, comprises the following step:

A, import gather together the configuration mutual exclusion of Critical value resources and virtual machine and group and impose a condition;

The resource use amount of B, acquisition collected each virtual machine history of a period of time;

C, close to calculate the resource utilization of physical machine reality according to the resource use amount of all virtual machines in physical machine, further try to achieve the standard deviation value under the environment of gathering together by the resource utilization of calculating all physical machine again, to judge load balanced state and dispersion degree between each physical machine;

Whether the standard deviation value that D, judgement are tried to achieve is greater than the Critical value resources of gathering together, if, carry out selecting of physical machine, to pick out at least one pair of physical machine, and in two physical machine of selected pairing, select virtual machine and simulate and move action, to improve the standard deviation value of physical machine resource utilization, wherein, the resource utilization of one of them physical machine in the physical machine of matching is greater than another physical machine;

E, calculate simulation and move after virtual machine, judge that whether the standard deviation value of physical machine resource utilization is still higher than the Critical value resources of gathering together, if so, repeated execution of steps D and E, if not, performs step F; And

F, carry out the action of moving of virtual machine, to move mutually the virtual machine in two physical machine of selected pairing.

2. the method for balance physical machine resource load in the environment of gathering together according to claim 1, is characterized in that, dispersion degree is calculated expression by following mathematical expression:

${\mathrm{\ρ}}_j^{\mathrm{resource}}=\frac{{\mathrm{\Σi}\∈s}_j{X}_{i,j}{V}_i^{\mathrm{resource}}}{H_j^{\mathrm{resource}}};$

${\mathrm{\μ}}^{\mathrm{resource}}=\frac{1}{m}{\mathrm{\Σ}}_{j=1}^m{\mathrm{\ρ}}_j^{\mathrm{resource}};$

$\mathrm{\σ}=\sqrt{\frac{1}{m}{\mathrm{\Σ}}_{j=1}^m{({\mathrm{\ρ}}_j^{\mathrm{resource}}-{\mathrm{\μ}}^{\mathrm{resource}})}^2};$

Wherein, represent the resource utilization of a numbering j physical machine; X _{i, j}representing whether a numbering i virtual machine has in the physical machine of this numbering j starts, and its value of some words is 1, and its value of the words of not enabling is 0; the resource capability that virtual machine i uses for this reason; μ ^resourcethe arithmetic mean value of resource utilization of interior m platform physical machine represents to gather together; The σ standard deviation value that above numerical value is tried to achieve of serving as reasons, in order to express the dispersion degree of this interior resource load of gathering together.

3. the method for balance physical machine resource load in the environment of gathering together according to claim 2, is characterized in that, comprises the following step in step D:

Calculate the resource utilization of each physical machine with arithmetic mean value μ ^resourcedifference D _j; And

According to the difference D calculating _jthe physical machine of the interior resource anxiety of gathering together is classified as to one group, and other physical machine is classified as another group, and by the physical machine of two groups according to difference D _jsize carry out rank, so that the physical machine of identical rank between two groups or a certain particular rank is picked as to the object of mutual resettlement virtual machine, and then screen the physical machine of at least one pair of pairing, wherein, the one matching is source physical machine, another is target entity machine.

4. the method for balance physical machine resource load in the environment of gathering together according to claim 3, is characterized in that the difference D of physical machine _jbe greater than the group that ranges the interior resource anxiety of gathering together at 1 o'clock, difference D _jminus physical machine is attributed to another group.

5. the method for balance physical machine resource load in the environment of gathering together according to claim 3, is characterized in that, step D also comprises the following step:

After originate physical machine and target entity machine are picked out, calculate the use resource of source entity machine and target entity machine and the gap of average, wherein, this gap U is the numerical value that difference D is multiplied by the total capacity R of physical machine target resource; And

In the physical machine of source, pick out a virtual machine, the resource that this virtual machine consumes need be greater than the use resource of the physical machine of originating and the gap of average, and gap is to be less than according to the predefined numerical value of target resource.

6. the method for balance physical machine resource load in the environment of gathering together according to claim 5, is characterized in that, step D also comprises the following step:

The use resource of basis source physical machine and target entity machine and the gap U of average _s, U _t, calculate number Thd in resource load amount; And

If cannot pick out the virtual machine meeting in the physical machine of source time, in the physical machine of source, select a virtual machine and in target entity machine, pick out a virtual machine, and the selected virtual machine of source entity machine is moved to target entity machine in the future, selected target entity machine virtual machine is moved to source physical machine, to make to move the resource capability V of rear two virtual machines ^resourcedifference be able to position and exist | Ut| and | between Us| and approach Thd value most.

7. the method for balance physical machine resource load in the environment of gathering together according to claim 6, is characterized in that, number Thd=(U in resource load amount _s-U _t)/2.

8. the method for balance physical machine resource load in the environment of gathering together according to claim 6, is characterized in that, step D also comprises the following step:

If cannot pick out the virtual machine meeting in source physical machine and target entity machine time, in the physical machine of source, pick out a virtual machine, the resource capability V of this virtual machine ^resourceneed approach recently number Thd in resource load amount.