CN104881325A - Resource scheduling method and resource scheduling system - Google Patents

Abstract

An embodiment of the invention provides a resource scheduling method and a resource scheduling system and relates to the field of cloud calculation. The resource scheduling system comprises a management node and physical nodes. According to the resource scheduling method, by the aid of the management node, a resource set corresponding to each application program is acquired according to the scheduling policy of a user when an objection of a scheduling policy is achieved and then is sent to the each physical node; by the aid of each physical node, the minimum quantity of resources distributed to each virtual machine on the corresponding physical node are determined, the left quantity of resources are distributed to every virtual machine on the corresponding physical node proportionally, and the total quantity of resources obtained by each virtual machine after distribution are sent to the management node; by the aid of the management node, resource utilization rate of each physical node is acquired according to the corresponding total quantity of resources of each virtual machine and the current resource utilization amount of each virtual machine on the corresponding physical node, and resource load of each physical node is scheduled according to the resource utilization rate. The resource scheduling method and the resource scheduling system are applied to virtualized resource scheduling and can solve the problem of unreasonable resource distribution caused by an incompatible scheduling technology.

Description

A kind of resource regulating method and resource scheduling system

Technical field

The present invention relates to field of cloud calculation, particularly relate to a kind of resource regulating method and resource scheduling system.

Background technology

The scheduling of infrastructure resources in data center, can be divided into three aspects, comprise: the scheduling of resource of the scheduling of resource of application layer, the scheduling of resource of virtual machine layer and physical machine layer.

Wherein, First Input First Output (the First Input First Output realized in software architecture MapReduce for large-scale dataset concurrent operation, FIFO), fair scheduling algorithm and computer capacity dispatching algorithm etc. belong to application layer resource scheduling algorithm, and its target is the task throughput of optimization application layer or minimizes task completion time; The dispatching algorithm representative of virtual machine layer can be credit scheduler algorithm in Xen system, this kind of algorithm is built in virtualization system inside, regulate the shared behavior of the different virtual machine in separate unit physical machine, its target ensures the fair use to competitive resource between virtual machine; The dispatching algorithm of Physical layer can be divided into static server to integrate and virtual machine live migration technology.It is to the performance of the demand of resource in conjunction with physical server according to application layer that static server is integrated, find the mapping that fixing mapping mode sets up application program and physical server and virtual machine, need human intervention when larger fluctuation appears in load by the virtual machine (vm) migration on physical server excessive for load on the lighter server of load, virtual machine live migration technology is then dynamically adjust the distribution of virtual machine in physical machine according to the service condition of resources of virtual machine.

Above-mentioned three kinds of algorithms all control for resources of virtual machine and the scheduling of resource of some characteristics of scheduling creates positive effect, but, because each algorithm has certain limitation, various resources control and dispatching technique are due to the characteristic such as target and scope of itself, may exist with other dispatching technique and restrict or incompatible situation, be difficult to integrate.Such as, when the dispatching algorithm of physical machine layer needs to reach energy-conservation object according to the service condition of resources of virtual machine, dynamically adjust the distribution of virtual machine in physical machine, such as reduce the quantity of virtual machine, if but the resource scheduling algorithm of application layer needs to reach to minimize task completion time, need again sufficient virtual machine, so just cause the restriction of two kinds of dispatching techniques, make dispatching technique incompatible, Resourse Distribute is unreasonable.

Summary of the invention

The embodiment of the present invention provides a kind of resource regulating method and resource scheduling system, can solve the irrational problem of the incompatible Resourse Distribute caused of dispatching technique.

First aspect, provides a kind of resource scheduling system, comprises management node and physical node, wherein:

Management node, the resource collection that when reaching the target of described scheduling strategy for obtaining according to the scheduling strategy of user, each application program is corresponding, described resource collection comprises the stock number that when arbitrary application program runs on virtual machine, each virtual machine takies, and described resource collection is sent to each physical node;

Described physical node, for the least resource amount determining that according to described resource collection on described physical node, each virtual machine distributes, then the surplus yield of described physical node after the described least resource amount of distribution is proportionately distributed to each virtual machine on described physical node;

Described physical node, the total resources also for being obtained by each virtual machine distributed on rear described physical node sends to described management node;

Described management node, the current resource use amount of each virtual machine also on the total resources corresponding according to each virtual machine on described physical node and described physical node obtains the resource utilization of described physical node, and then dispatches the resource load of each physical node according to described resource utilization.

In conjunction with first aspect, in the first mode in the cards of first aspect, described management node specifically for:

According to the utility function of the scheduling strategy of described user obtain reach the target of described scheduling strategy time resource collection corresponding to each application program;

Wherein, described utility function comprises:

$f\_\mathrm{app}\left({\mathrm{app}}_k\right)=\underset{t=1}{\overset{s_k}{\mathrm{\Σ}}}{\cos t}_t*R_s*f\_\mathrm{time}(R_t,\mathrm{missio}{n}_t),$ F_app (app _k) represent that the target of described scheduling strategy, described target are the shortest time that described user buys the minimum or described virtual machine of the expense of described resources of virtual machine and finishes the work, cost _trepresent t the virtual machine VM distributing to a kth application program _tthe unit resource unit interval in price, R _trepresent the stock number that t the virtual machine distributing to a described kth application program takies, f_time (R _t, mission _t) represent that described t virtual machine completes distributed task amount mission _ttime shared time.

In conjunction with the first mode in the cards of first aspect or first aspect, in the second of first aspect mode in the cards, described physical node specifically for:

The stock number that when running on described virtual machine according to arbitrary application program described in the described resource collection that described each application program is corresponding, each virtual machine takies, determine to belong to the least resource amount that each virtual machine on described physical node should distribute, the least resource amount that the arbitrary virtual machine on described physical node takies is the stock number sum that described arbitrary virtual machine is corresponding in described each application program;

Obtain the surplus yield of described physical node, described surplus yield is that the total resources of described physical node deducts the described least resource amount determined for each virtual machine on described physical node;

Described surplus yield is proportionately distributed to each virtual machine on described physical node, the described surplus yield that arbitrary virtual machine on described physical node is assigned to is that the least resource amount of described arbitrary virtual machine accounts for the ratio of the stock number that described physical node has distributed and the product of described surplus yield, and described stock number of having distributed is the least resource amount sum of each virtual machine on described physical node.

The second in conjunction with first aspect mode in the cards, in the third mode in the cards of first aspect, described management node specifically for:

According to formula obtain the resource utilization of described physical node, represent the resource utilization of i-th physical node, represent the total resources that a jth virtual machine on described i-th physical node is corresponding, represent the total resources that described i-th physical node is corresponding, represent the resource use amount that a jth virtual machine on described i-th physical node is current;

If described management node needs the load of balanced each physical node described, then when described in obtaining, the variance of the resource utilization of each physical node is minimum, the resources of virtual machine distribution on each physical node described;

If described management node needs the number reducing described physical node, then first according to the total resources order from big to small of each physical node described, each physical node described is arranged, then successively virtual machine is dispensed on physical node, again next physical node is distributed until the total resources of the physical node of current distribution has all been assigned with;

Determine, after being all dispensed on physical node by all virtual machines, whether to there is the physical node of unallocated virtual machine, if exist, then close the physical node of described unallocated virtual machine.

Second aspect, provides a kind of resource regulating method, comprising:

Management node according to the scheduling strategy of user obtain reach the target of described scheduling strategy time resource collection corresponding to each application program, described resource collection comprises the stock number that when arbitrary application program runs on virtual machine, each virtual machine takies, and described resource collection is sent to each physical node;

Described physical node determines according to described resource collection the least resource amount that on described physical node, each virtual machine distributes, then the surplus yield of described physical node after the described least resource amount of distribution is proportionately distributed to each virtual machine on described physical node;

The total resources that each virtual machine distributed on rear described physical node obtains is sent to described management node by described physical node;

The described management node resource use amount current according to each virtual machine on total resources corresponding to each virtual machine on described physical node and described physical node obtains the resource utilization of described physical node, and then dispatches the resource load of each physical node according to described resource utilization.

In conjunction with second aspect, in the first mode in the cards of second aspect, described management node according to the scheduling strategy of user obtain reach the target of described scheduling strategy time resource collection corresponding to each application program comprise:

Described management node according to the utility function of the scheduling strategy of described user obtain reach the target of described scheduling strategy time resource collection corresponding to each application program;

Wherein, described utility function comprises:

$f\_\mathrm{app}\left({\mathrm{app}}_k\right)=\underset{t=1}{\overset{s_k}{\mathrm{\Σ}}}{\cos t}_t*R_t*f\_\mathrm{time}(R_t,\mathrm{missio}{n}_t),$ F_app (app _k) represent that the target of described scheduling strategy, described target are the shortest time that described user buys the minimum or described virtual machine of the expense of described resources of virtual machine and finishes the work, cost _trepresent t the virtual machine VM distributing to a kth application program _tthe unit resource unit interval in price, R _trepresent the stock number that t the virtual machine distributing to a described kth application program takies, f_time (R _t, mission _t) represent that described t virtual machine completes distributed task amount mission _ttime shared time.

In conjunction with the first mode in the cards of second aspect or second aspect, in the second of second aspect mode in the cards, described physical node determines according to described resource collection the least resource amount that on described physical node, each virtual machine distributes, then is comprised by each virtual machine that the surplus yield of described physical node after the described least resource amount of distribution is proportionately distributed on described physical node:

The stock number that when described physical node runs on described virtual machine according to arbitrary application program described in described resource collection corresponding to described each application program, each virtual machine takies, determine to belong to the least resource amount that each virtual machine on described physical node should distribute, the least resource amount that the arbitrary virtual machine on described physical node takies is the stock number sum that described arbitrary virtual machine is corresponding in described each application program;

Described physical node obtains the surplus yield of described physical node, and described surplus yield is that the total resources of described physical node deducts the described least resource amount determined for each virtual machine on described physical node;

Described surplus yield is proportionately distributed to each virtual machine on described physical node by described physical node, the described surplus yield that arbitrary virtual machine on described physical node is assigned to is that the least resource amount of described arbitrary virtual machine accounts for the ratio of the stock number that described physical node has distributed and the product of described surplus yield, and described stock number of having distributed is the least resource amount sum of each virtual machine on described physical node.

The second in conjunction with second aspect mode in the cards, in the third mode in the cards of second aspect, the current resource use amount of each virtual machine on the described total resources corresponding according to each virtual machine on described physical node and described physical node obtains the resource utilization of described physical node, and then comprises according to the resource load that described resource utilization dispatches each physical node:

According to formula obtain the resource utilization of described physical node, represent the resource utilization of i-th physical node, represent the total resources that a jth virtual machine on described i-th physical node is corresponding, represent the total resources that described i-th physical node is corresponding, represent the resource use amount that a jth virtual machine on described i-th physical node is current;

If described management node needs the load of balanced each physical node described, then when described in obtaining, the variance of the resource utilization of each physical node is minimum, the resources of virtual machine distribution on each physical node described;

If described management node needs the number reducing described physical node, then first according to the total resources order from big to small of each physical node described, each physical node described is arranged, then successively virtual machine is dispensed on physical node, again next physical node is distributed until the total resources of the physical node of current distribution has all been assigned with;

Determine, after being all dispensed on physical node by all virtual machines, whether to there is the physical node of unallocated virtual machine, if exist, then close the physical node of described unallocated virtual machine.

The third aspect, provides a kind of management node, comprises virtual scheduler module, data memory module, communication module and data monitoring module, wherein:

Described virtual scheduler module, the resource collection that when reaching the target of described scheduling strategy for obtaining according to the scheduling strategy of user, each application program is corresponding, described resource collection comprises the stock number that when arbitrary application program runs on virtual machine, each virtual machine takies;

Described data memory module, for storing resource collection corresponding to described each application program;

Described communication module, sends to each physical node for the described resource collection stored by described data memory module;

Described communication module, also for total resources that each virtual machine receiving described physical node obtains;

Described data monitoring module, the current resource use amount of each virtual machine on the total resources that each virtual machine on the described physical node received according to described communication module is corresponding and described physical node obtains the resource utilization of described physical node;

Described virtual scheduler module, the described resource utilization also for obtaining according to described data monitoring module dispatches the resource load of each physical node.

Fourth aspect, provides a kind of physical node, comprises communication module and virtual scheduler module, wherein:

Described communication module, for the resource collection of each application program that receiving management node sends, described resource collection comprises the stock number that when arbitrary application program runs on virtual machine, each virtual machine takies;

Described virtual scheduler module, for the least resource amount determining that according to described resource collection on described physical node, each virtual machine distributes, then the surplus yield of described physical node after the described least resource amount of distribution is proportionately distributed to each virtual machine on described physical node;

Described communication module, the total resources also for being obtained by each virtual machine distributed on rear described physical node sends to described management node, dispatches the resource load of each physical node to make described management node according to described total resources.

The embodiment of the present invention provides a kind of resource regulating method and resource scheduling system, this system comprises management node and physical node, management node according to the scheduling strategy of user obtain reach the target of scheduling strategy time resource collection corresponding to each application program, resource collection comprises the stock number that when arbitrary application program runs on virtual machine, each virtual machine takies, and resource collection is sent to each physical node, the least resource amount that physical node distributes according to virtual machine each on resource collection determination physical node, again the surplus yield of physical node after distribution least resource amount is proportionately distributed to each virtual machine on physical node, the total resources that each virtual machine distributed on rear physical node obtains is sent to management node by physical node, the management node resource use amount current according to each virtual machine on total resources corresponding to each virtual machine on physical node and physical node obtains the resource utilization of physical node, and then the resource load of each physical node is dispatched according to resource utilization, like this, by this resource scheduling system, the scheduling of application layer can be paid the utmost attention to, on the basis of resource requirement ensureing user, from virtualization layer, the scheduling of Physical layer is dispatched resource, this from top under modeling scheme, based on application layer, virtual machine and physics joint and resource use amount, can to the physical node in system, virtual machine node carries out unified performance monitoring and scheduling, thus the overall situation more reasonably Resources allocation, the irrational problem of the incompatible Resourse Distribute caused of dispatching technique can be solved.

Accompanying drawing explanation

In order to be illustrated more clearly in the technical scheme of the embodiment of the present invention, be briefly described to the accompanying drawing used required in embodiment or description of the prior art below, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skill in the art, under the prerequisite not paying creative work, other accompanying drawing can also be obtained according to these accompanying drawings.

The configuration diagram of a kind of resource scheduling system that Fig. 1 provides for the embodiment of the present invention;

The schematic flow sheet of a kind of resource regulating method that Fig. 2 provides for the embodiment of the present invention;

The structural representation of a kind of management node that Fig. 3 provides for the embodiment of the present invention;

The structural representation of a kind of physical node that Fig. 4 provides for the embodiment of the present invention.

Embodiment

Below in conjunction with the accompanying drawing in the embodiment of the present invention, be clearly and completely described the technical scheme in the embodiment of the present invention, obviously, described embodiment is only the present invention's part embodiment, instead of whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art, not making the every other embodiment obtained under creative work prerequisite, belong to the scope of protection of the invention.

In order to be described the resource scheduling in virtualization system on the whole, the present invention proposes a kind of centralized resource scheduling system framework, and based on this framework, propose a kind of top-down resource scheduling system, modeling is carried out to resource scheduling, and has proposed corresponding method.

In embodiments of the present invention, as shown in Figure 1, above-mentioned centralized resource scheduling system framework can be divided into two parts: Sensor end (also can be described as slave unit) and Master end (also can be described as master control set), correspondence is present in physical node in the embodiment of the present invention and management node, and both constitute the centralized architecture centered by Master end.Master end holds with Sensor the transmission being carried out data stream and control flow check by the physical subnets of virtual cluster, jointly completes the monitoring to virtual cluster and scheduling of resource.

Wherein, it is inner that Sensor end is positioned at each physical node, and run the privileged domain Dom0 of territory physical node virtualization system, responsible content can comprise: monitor and collect the resource using information of other virtual machine on Same Physical node; According to the information of collecting, adjust the scheduling strategy of this physical node inside in real time; The information of scheduling result and collection is reported Master end in time.Master end just hold the information of scheduling result and the collection reported to form the monitor data of cluster according to Sensor, and then dispatches the resource allocation conditions of whole cluster, and scheduling strategy is transferred to Sensor and holds and Master holds self other virtual machine.

Based on above-mentioned framework, the physical node first related to the resource scheduling system of the embodiment of the present invention, virtual machine, application layer are described.

For the physical node of Physical layer, be designated as in embodiments of the present invention:

S_Node={node ₁, node ₂..., node _i..., node _n, be characterized in infrastructure layer, can for the set of the n platform physical node of virtual machine.For each node _i, there are two values: total resources R _nodeiwith resource use amount L _nodei, characterize node respectively _ithe ratio that upper assignable total resources and Current resource have taken.

For the virtual machine of virtualization layer, in embodiments of the present invention, virtual machines all in system can be represented by S set _ VM, is designated as: S_VM={VM ₁, VM ₂..., VM _j..., VM _m.For each VM _j, there are two values: total resources R _vMiwith resource use amount L _vMi, characterize VM respectively _ithe ratio that upper assignable total resources and Current resource have taken.

For application layer, in embodiments of the present invention, the application all virtual machines run can be designated as: S_APP={app ₁, app ₂..., app _i..., app _s, to characterize cloud computing user operation application program on a virtual machine.The embodiment of the present invention, in order to describe facility, a virtual machine runs an application program, but an application program can run on multiple virtual machine.Virtual machine is assigned to this relation of application program, and the embodiment of the present invention can be designated as .So, have:

$\∀{\mathrm{app}}_k\∈S\_\mathrm{APP},\∃S\_{\mathrm{VM}}_{{\mathrm{app}}_k}=\{{\mathrm{VM}}_{k_1},{\mathrm{vm}}_{k_2},...,{\mathrm{vm}}_{k_{\mathrm{sk}}}\}\⋐S\_\mathrm{WM}$

Make, set up.

Based on above-mentioned explanation, below the embodiment of the present invention is specifically described.

The embodiment of the present invention provides a kind of resource scheduling system, and with reference to system architecture as shown in Figure 1, this system comprises management node and physical node, wherein:

Management node, the resource collection that when reaching the target of scheduling strategy for obtaining according to the scheduling strategy of user, each application program is corresponding, resource collection comprises the stock number that when arbitrary application program runs on virtual machine, each virtual machine takies, and resource collection is sent to each physical node;

Physical node, for the least resource amount of distributing according to virtual machine each on resource collection determination physical node, then is proportionately distributed to each virtual machine on physical node by the surplus yield of physical node after distribution least resource amount;

Physical node, the total resources also for being obtained by each virtual machine distributed on rear physical node sends to management node;

Management node, the current resource use amount of each virtual machine also on the total resources corresponding according to each virtual machine on physical node and physical node obtains the resource utilization of physical node, and then dispatches the resource load of each physical node according to resource utilization.

Specifically, in application-level scheduling technology, management node obtain according to the scheduling strategy of user reach the target of scheduling strategy time resource collection corresponding to each application program time can be specifically for: according to the utility function of the scheduling strategy of user obtain reach the target of scheduling strategy time resource collection corresponding to each application program.

Wherein, this utility function can comprise:

$f\_\mathrm{app}\left({\mathrm{app}}_k\right)=\underset{t=1}{\overset{s_k}{\mathrm{\Σ}}}{\cos t}_t*R_T*f\_\mathrm{time}(R_t,\mathrm{missio}{n}_t)$

F_app (app _k) representing the target of scheduling strategy, target is the shortest time that user buys the minimum or virtual machine of the expense of resources of virtual machine and finishes the work, cost _trepresent t the virtual machine VM distributing to a kth application program _tthe unit resource unit interval in price, R _trepresent the stock number that t the virtual machine distributing to a kth application program takies, f_time (R _t, mission _t) represent that t virtual machine completes distributed task amount mission _ttime shared time.

This is because, in the application of cloud computing, user needs to buy certain resources of virtual machine by paying, and therefore, in order to complete the operation of submission, user needs to weigh between resource and expense.The virtual machine of user's application is more, then resource is more, and the time fulfiled assignment is shorter, and certainly total expense also can improve, and also namely how resource is chosen, and depends on the scheduling strategy of user.

Such as, when user wishes least cost, the target of its scheduling strategy can be expressed as: for each VM _t, determine R _t, make f_app (app _k, R _{t '})=min (f_app (app _k, R _t)).

If user wishes the shortest time of finishing the work, the target of its scheduling strategy can be expressed as: for each VM _t, determine R _t, make:

min(f_time(R _t',minssion _t'))＝min(min(f_time(R _t,minssion _t)))

Certainly, no matter which kind of target the scheduling strategy that user finally selects is based on, can remember that user is app _kthe resource collection selected for: wherein, s _kbe expressed as app _kthe s determined _kindividual virtual machine, represent the s distributing to a kth application program _kthe stock number that individual virtual machine takies.Also namely resource collection comprises the stock number that when arbitrary application program runs on virtual machine, each virtual machine takies.

Here resource can be understood as cpu resource, internal memory, the network bandwidth and the disk size etc. of distributing to virtual machine.

Complete the Resourse Distribute of application programs in application layer after, management node needs the resource collection determined for each application program to send to each physical node, so that physical node determines the total resources of each virtual machine according to the resource collection received, also namely carry out the scheduling of resource of virtualization layer.

First, physical node is after receiving resource collection, first according to the least resource amount that each virtual machine on resource collection determination physical node distributes, then the surplus yield of this physical node after distribution least resource amount can be proportionately distributed to each virtual machine on this physical node.

Particularly, the stock number that physical node can take according to each virtual machine when arbitrary application program runs on virtual machine in resource collection corresponding to each application program, determine to belong to the least resource amount that each virtual machine on this physical node should distribute, the least resource amount that the arbitrary virtual machine on this physical node takies is the stock number sum that arbitrary virtual machine is corresponding in each application program.Also namely, the stock number of required virtual machine when first will ensure that each application program is run.

Exemplary, because the stock number on physical node is certain, the surplus yield R_res that physical node is initial before Resources allocation can be thought _nodei=R _nodei, 1<=j<=n _i, the stock number that the jth virtual machine on i-th physical node is assigned to then from j=1, for each virtual machine distributes minimum stock number.Also namely, will be set to wherein:

and represent app _kthe stock number that a jth virtual machine of i-th physical node is assigned to, upgrades surplus yield R_res simultaneously _nodeifor:

until j=n _i.

After being assigned with least resource amount for each virtual machine on physical node, physical node can obtain its surplus yield, and this surplus yield just deducts the least resource amount determined for each virtual machine on physical node for the total resources of physical node.

Then, surplus yield can be proportionately distributed to each virtual machine on physical node, the surplus yield that arbitrary virtual machine on physical node is assigned to is that the least resource amount of arbitrary virtual machine accounts for the ratio of the stock number that physical node has distributed and the product of surplus yield, and the stock number of having distributed is the least resource amount sum of each virtual machine on physical node.

Specifically, the least resource amount be assigned to due to the virtual machine on physical node may not be identical, surplus yield can account for the ratio of the stock number of having distributed to distribute to each virtual machine according to the least resource amount of virtual machine, what namely the least resource amount that obtains of virtual machine was more can continue to distribute more surplus yield, the least resource amount that virtual machine obtains fewer can continue distribute fewer surplus yield, to ensure that resources of virtual machine is sufficient.Specifically can distribute surplus yield R_res as follows _nodei:

$R_{{\mathrm{VM}}_{i_j}}={R^{\&prime;}}_{{\mathrm{VM}}_{i_j}}\frac{{R^{\&prime;}}_{{\mathrm{VM}}_{i_j}}}{R_{{\mathrm{node}}_i}-R\_\mathrm{re}{s}_{\mathrm{nod}{e}_i}}*R\_{\mathrm{res}}_{\mathrm{no}{\mathrm{de}}_i},$ represent the total resources finally obtained surplus resources being distributed the jth virtual machine on rear i-th physical node.

For example, physical node there are 10 cpu resources, and operation has 2 virtual machines on this physical node, the least resource amount of first virtual machine needs is 5 cpu resources, the least resource amount of second virtual machine needs is 3 cpu resources, after distributing least resource amount for each virtual machine, remain 2 cpu resources, can be that first virtual machine continues to distribute surplus yield according to the algorithm of 5/8*2, be that second virtual machine continues to distribute surplus yield according to the algorithm of 3/8*2, the total resources that so first virtual machine obtains is 5/8*2+5, the total resources that second virtual machine obtains is 3/8*2+3.

Physical node is after for each virtual machine Resources allocation amount, and the total resources that each virtual machine obtains can be sent to management node, so that management node is dispatched Physical layer according to resources of virtual machine by physical node.Specifically the current resource use amount of each virtual machine on the total resources corresponding according to each virtual machine on physical node and physical node can obtain the resource utilization of physical node, and then dispatch the resource load of each physical node according to resource utilization.

Particularly, can according to formula obtain the resource utilization of physical node, represent the resource utilization of i-th physical node, represent the total resources that a jth virtual machine on i-th physical node is corresponding, represent the total resources that i-th physical node is corresponding, represent the resource use amount that a jth virtual machine on i-th physical node is current.

Then can carry out scheduling of resource according to the target of physical layer scheduling, if management node needs the load of each physical node balanced, then obtain the variance of the resource utilization of each physical node minimum time, resources of virtual machine distribution on each physical node, namely, when the target of physical layer scheduling scheme is load balancing, its utility function can be:

map M (S_VM, S_Node), make [L _nodei-E (L _nodei)] ²minimum;

When the resource utilization of a certain physical node is lower, show that the idling-resource of virtual machine on this physical node is more, now can by the virtual machine (vm) migration on other nodes to this node; When the resource utilization of a certain physical node is higher, show that the idling-resource of virtual machine on this physical node is less, now can by the virtual machine (vm) migration on this node on other node, to reach the load balancing of each node.

If management node needs the number reducing physical node, then can first according to the total resources order from big to small of each physical node, each physical node be arranged, then successively virtual machine is dispensed on physical node, again next physical node is distributed until the total resources of the physical node of current distribution has all been assigned with, determine again after all virtual machines are all dispensed on physical node, whether there is the physical node of unallocated virtual machine, if exist, then close the physical node of unallocated virtual machine, so, can need virtual machine to be dispensed on the least possible physical node, to reduce the physical node quantity being in open state, reduce the energy consumption of physical node.

By the explanation of above-mentioned 3 levels, if the resource dispatching strategy that also namely will realize in virtualization system, the scheduling scheme of application layer can be paid the utmost attention to, on the basis of resource requirement ensureing user, infrastructure providers just adjusts from the configuration status of angle to resource of virtualization layer, Physical layer.Also namely in the process of whole scheduling, priority is: application layer > virtualization layer > Physical layer, dispatching system can be made to consider cloud computing user and infrastructure resources provides the interests with both sides, more reasonably Resources allocation the overall situation.

The embodiment of the present invention provides a kind of resource scheduling system, this system comprises management node and physical node, management node according to the scheduling strategy of user obtain reach the target of scheduling strategy time resource collection corresponding to each application program, resource collection comprises the stock number that when arbitrary application program runs on virtual machine, each virtual machine takies, and resource collection is sent to each physical node, the least resource amount that physical node distributes according to virtual machine each on resource collection determination physical node, again the surplus yield of physical node after distribution least resource amount is proportionately distributed to each virtual machine on physical node, the total resources that each virtual machine distributed on rear physical node obtains is sent to management node by physical node, the management node resource use amount current according to each virtual machine on total resources corresponding to each virtual machine on physical node and physical node obtains the resource utilization of physical node, and then the resource load of each physical node is dispatched according to resource utilization, like this, by this resource scheduling system, the scheduling of application layer can be paid the utmost attention to, on the basis of resource requirement ensureing user, from virtualization layer, the scheduling of Physical layer is dispatched resource, this from top under modeling scheme, based on application layer, virtual machine and physics joint and resource use amount, can to the physical node in system, virtual machine node carries out unified performance monitoring and scheduling, thus the overall situation more reasonably Resources allocation, the irrational problem of the incompatible Resourse Distribute caused of dispatching technique can be solved.

Further embodiment of this invention provides a kind of resource regulating method, as shown in Figure 2, comprising:

201, management node according to the scheduling strategy of user obtain reach the target of scheduling strategy time resource collection corresponding to each application program, resource collection comprises the stock number that when arbitrary application program runs on virtual machine, each virtual machine takies, and resource collection is sent to each physical node.

Particularly, management node according to the scheduling strategy of user obtain reach the target of scheduling strategy time resource collection corresponding to each application program can comprise:

Management node according to the utility function of the scheduling strategy of user obtain reach the target of scheduling strategy time resource collection corresponding to each application program;

Wherein, utility function comprises:

$f\_\mathrm{app}\left({\mathrm{app}}_k\right)=\underset{t=1}{\overset{s_k}{\mathrm{\&Sigma;}}}{\cos t}_t*R_T*f\_\mathrm{time}(R_t,\mathrm{missio}{n}_t),$ F_app (app _k) representing the target of scheduling strategy, target is the shortest time that user buys the minimum or virtual machine of the expense of resources of virtual machine and finishes the work, cost _trepresent t the virtual machine VM distributing to a kth application program _tthe unit resource unit interval in price, R _trepresent the stock number that t the virtual machine distributing to a kth application program takies, f_time (R _t, mission _t) represent that t virtual machine completes distributed task amount mission _ttime shared time.

202, physical node is according to the least resource amount of virtual machine distribution each on resource collection determination physical node, then the surplus yield of physical node after distribution least resource amount is proportionately distributed to each virtual machine on physical node.

Particularly, the stock number that physical node takies according to each virtual machine when arbitrary application program runs on virtual machine in resource collection corresponding to each application program, determine to belong to the least resource amount that each virtual machine on physical node should distribute, the least resource amount that arbitrary virtual machine on physical node takies is the stock number sum that arbitrary virtual machine is corresponding in each application program, then obtain the surplus yield of physical node, surplus yield is that the total resources of physical node deducts the least resource amount determined for each virtual machine on physical node, again surplus yield is proportionately distributed to each virtual machine on physical node, the surplus yield that arbitrary virtual machine on physical node is assigned to is that the least resource amount of arbitrary virtual machine accounts for the ratio of the stock number that physical node has distributed and the product of surplus yield, the stock number of having distributed is the least resource amount sum of each virtual machine on physical node.

203, the total resources that each virtual machine distributed on rear physical node obtains is sent to management node by physical node.

204, the resource use amount that management node is current according to each virtual machine on total resources corresponding to each virtual machine on physical node and physical node obtains the resource utilization of physical node, and then dispatches the resource load of each physical node according to resource utilization.

Particularly, can according to formula obtain the resource utilization of physical node, represent the resource utilization of i-th physical node, represent the total resources that a jth virtual machine on i-th physical node is corresponding, represent the total resources that i-th physical node is corresponding, represent the resource use amount that a jth virtual machine on i-th physical node is current;

If management node needs the load of each physical node balanced, then obtain the variance of the resource utilization of each physical node minimum time, the resources of virtual machine distribution on each physical node;

If management node needs the number reducing physical node, then first according to the total resources order from big to small of each physical node, each physical node is arranged, then successively virtual machine is dispensed on physical node, again next physical node is distributed until the total resources of the physical node of current distribution has all been assigned with, and then determine after all virtual machines are all dispensed on physical node, whether there is the physical node of unallocated virtual machine, if exist, then close the physical node of unallocated virtual machine.

In the present embodiment, the specific implementation of management node and physical node with reference to the elaboration in above-mentioned resource scheduling system embodiment, can not repeat here.

The embodiment of the present invention provides a kind of resource regulating method, be applied to resource scheduling system, this system comprises management node and physical node, management node according to the scheduling strategy of user obtain reach the target of scheduling strategy time resource collection corresponding to each application program, resource collection comprises the stock number that when arbitrary application program runs on virtual machine, each virtual machine takies, and resource collection is sent to each physical node, the least resource amount that physical node distributes according to virtual machine each on resource collection determination physical node, again the surplus yield of physical node after distribution least resource amount is proportionately distributed to each virtual machine on physical node, the total resources that each virtual machine distributed on rear physical node obtains is sent to management node by physical node, the management node resource use amount current according to each virtual machine on total resources corresponding to each virtual machine on physical node and physical node obtains the resource utilization of physical node, and then the resource load of each physical node is dispatched according to resource utilization, like this, by this resource scheduling system, the scheduling of application layer can be paid the utmost attention to, on the basis of resource requirement ensureing user, from virtualization layer, the scheduling of Physical layer is dispatched resource, this from top under modeling scheme, based on application layer, virtual machine and physics joint and resource use amount, can to the physical node in system, virtual machine node carries out unified performance monitoring and scheduling, thus the overall situation more reasonably Resources allocation, the irrational problem of the incompatible Resourse Distribute caused of dispatching technique can be solved.

Further embodiment of this invention provides a kind of management node 3, as shown in Figure 3, this management node comprises master control (Master) device 30, master control set 30 comprises virtual machine monitor (Virtual Machine Manager, VMM) interface 301, interface module 302, data collection module 303, data memory module 304, data monitoring module 305, virtual scheduler module 306, communication module 307 and alarm module 308.

Wherein, the data that interface module 302 and communication module 307 are collected for other physical node in collection management node and cluster, to form the monitor data of cluster, and are centrally stored in data memory module 304.Virtual scheduler module 306, for according to history and real-time monitor data, adjusts the resource allocation conditions of whole cluster, and adjustable strategies is transferred to other virtual machines of physical node and management node self by communication module 307.

In embodiments of the present invention, virtual scheduler module 306, the resource collection that when reaching the target of scheduling strategy for obtaining according to the scheduling strategy of user, each application program is corresponding, resource collection comprises the stock number that when arbitrary application program runs on virtual machine, each virtual machine takies;

Data memory module 304, for storing resource collection corresponding to each application program;

Communication module 307, the resource collection for being stored by data memory module sends to each physical node;

Communication module 307, also for total resources that each virtual machine receiving physical node obtains;

Data monitoring module 305, the current resource use amount of each virtual machine on the total resources that each virtual machine on the physical node received according to communication module is corresponding and physical node obtains the resource utilization of physical node;

Virtual scheduler module 306, the resource utilization also for obtaining according to data monitoring module dispatches the resource load of each physical node.

Optionally, virtual scheduler module 306 can be specifically for:

According to the utility function of the scheduling strategy of user obtain reach the target of scheduling strategy time resource collection corresponding to each application program;

Wherein, utility function comprises:

$f\_\mathrm{app}\left({\mathrm{app}}_k\right)=\underset{t=1}{\overset{s_k}{\mathrm{\&Sigma;}}}{\cos t}_t*R_T*f\_\mathrm{time}(R_t,\mathrm{missio}{n}_t)$ F_app (app _k) representing the target of scheduling strategy, target is the shortest time that user buys the minimum or virtual machine of the expense of resources of virtual machine and finishes the work, cost _trepresent t the virtual machine VM distributing to a kth application program _tthe unit resource unit interval in price, R _trepresent the stock number that t the virtual machine distributing to a kth application program takies, f_time (R _t, mission _t) represent that t virtual machine completes distributed task amount mission _ttime shared time.

Optionally, virtual scheduler module 306 can be specifically for:

According to formula obtain the resource utilization of physical node, represent the resource utilization of i-th physical node, represent the total resources that a jth virtual machine on i-th physical node is corresponding, represent the total resources that i-th physical node is corresponding, represent the resource use amount that a jth virtual machine on i-th physical node is current;

If management node needs the load of each physical node balanced, then obtain the variance of the resource utilization of each physical node minimum time, the resources of virtual machine distribution on each physical node;

If management node needs the number reducing physical node, then first according to the total resources order from big to small of each physical node, each physical node is arranged, then successively virtual machine is dispensed on physical node, again next physical node is distributed until the total resources of the physical node of current distribution has all been assigned with;

Determine, after being all dispensed on physical node by all virtual machines, whether to there is the physical node of unallocated virtual machine, if exist, then close the physical node of unallocated virtual machine.

In the embodiment of the present invention, the specific implementation of each module with reference to the specific descriptions in the embodiment of above-mentioned resource scheduling system and embodiment of the method, can repeat no more here.

The embodiment of the present invention provides a kind of management node, virtual scheduler module in this management node, the resource collection that when reaching the target of scheduling strategy for obtaining according to the scheduling strategy of user, each application program is corresponding, resource collection comprises the stock number that when arbitrary application program runs on virtual machine, each virtual machine takies, data memory module, for storing resource collection corresponding to each application program, communication module, resource collection for being stored by data memory module sends to each physical node, communication module, also for total resources that each virtual machine receiving physical node obtains, data monitoring module, the current resource use amount of each virtual machine on the total resources that each virtual machine on the physical node received according to communication module is corresponding and physical node obtains the resource utilization of physical node, virtual scheduler module, resource utilization also for obtaining according to data monitoring module dispatches the resource load of each physical node, like this, the scheduling of application layer can be paid the utmost attention to, on the basis of resource requirement ensureing user, from virtualization layer, the scheduling of Physical layer is dispatched resource, this from top under modeling scheme, based on application layer, virtual machine and physics joint and resource use amount, can to the physical node in resource scheduling system, virtual machine node carries out unified performance monitoring and scheduling, thus the overall situation more reasonably Resources allocation, the irrational problem of the incompatible Resourse Distribute caused of dispatching technique can be solved.

Further embodiment of this invention provides a kind of physical node 4, as shown in Figure 4, comprise subordinate (Sensor) device 40, this slave unit 40 comprises VMM interface 401, interface module 402, Data Collection and memory module 403, virtual scheduler module 404 and communication module 405.

Wherein, Data Collection and memory module 403 can be monitored by the resource of VMM interface 401 to all virtual machines and physical node, and the data collected are passed to virtual scheduler module 404, virtual scheduler module 404 makes the decision of whether changing current Resource Allocation Formula according to scheduling strategy.Communication module 405 is responsible for transmitting control command or data message to VMM or management node, so that management node makes comprehensive scheduling strategy, simultaneously also the resource dispatching strategy of physical node inside is placed in physical node inside, mutual between the virtual machine that can simplify different physical node.

In embodiments of the present invention, communication module 405, may be used for the resource collection of each application program that receiving management node sends, resource collection comprises the stock number that when arbitrary application program runs on virtual machine, each virtual machine takies;

Virtual scheduler module 404, for the least resource amount of distributing according to virtual machine each on resource collection determination physical node, then is proportionately distributed to each virtual machine on physical node by the surplus yield of physical node after distribution least resource amount;

Communication module 405, the total resources also for being obtained by each virtual machine distributed on rear physical node sends to management node, dispatches the resource load of each physical node to make management node according to total resources.

Optionally, virtual scheduler module 404 can be specifically for:

The stock number that when running on virtual machine according to application program arbitrary in the resource collection that each application program is corresponding, each virtual machine takies, determine to belong to the least resource amount that each virtual machine on physical node should distribute, the least resource amount that the arbitrary virtual machine on physical node takies is the stock number sum that arbitrary virtual machine is corresponding in each application program;

Obtain the surplus yield of physical node, surplus yield is that the total resources of physical node deducts the least resource amount determined for each virtual machine on physical node

Surplus yield is proportionately distributed to each virtual machine on physical node, the surplus yield that arbitrary virtual machine on physical node is assigned to is that the least resource amount of arbitrary virtual machine accounts for the ratio of the stock number that physical node has distributed and the product of surplus yield, and the stock number of having distributed is the least resource amount sum of each virtual machine on physical node.

The specific implementation of each module in the embodiment of the present invention see the concrete elaboration in above-mentioned resource scheduling system and embodiment of the method, can repeat no more here.

The embodiment of the present invention provides a kind of physical node, communication module wherein, and for the resource collection of each application program that receiving management node sends, resource collection comprises the stock number that when arbitrary application program runs on virtual machine, each virtual machine takies, virtual scheduler module, for the least resource amount of distributing according to virtual machine each on resource collection determination physical node, then is proportionately distributed to each virtual machine on physical node by the surplus yield of physical node after distribution least resource amount, communication module, total resources also for being obtained by each virtual machine distributed on rear physical node sends to management node, the resource load of each physical node is dispatched according to total resources to make management node, like this, the scheduling of application layer can be paid the utmost attention to, on the basis of resource requirement ensureing user, from virtualization layer, the scheduling of Physical layer is dispatched resource, this from top under modeling scheme, based on application layer, virtual machine and physics joint and resource use amount, can to the physical node in resource scheduling system, virtual machine node carries out unified performance monitoring and scheduling, thus the overall situation more reasonably Resources allocation, the irrational problem of the incompatible Resourse Distribute caused of dispatching technique can be solved.

In several embodiments that the application provides, should be understood that disclosed node and method can realize by another way.Such as, device embodiment described above is only schematic, such as, the division of described unit, be only a kind of logic function to divide, actual can have other dividing mode when realizing, such as multiple unit or assembly can in conjunction with or another system can be integrated into, or some features can be ignored, or do not perform.Another point, shown or discussed coupling each other or direct-coupling or communication connection can be by some interfaces, and the indirect coupling of device or unit or communication connection can be electrical, machinery or other form.

The described unit illustrated as separating component or can may not be and physically separates, and the parts as unit display can be or may not be physical location, namely can be positioned at a place, or also can be distributed in multiple network element.Some or all of unit wherein can be selected according to the actual needs to realize the object of the present embodiment scheme.

In addition, each functional unit in each embodiment of the present invention can be integrated in a processing unit, also can be that the independent physics of unit comprises, also can two or more unit in a unit integrated.Above-mentioned integrated unit both can adopt the form of hardware to realize, and the form that hardware also can be adopted to add SFU software functional unit realizes.

The above-mentioned integrated unit realized with the form of SFU software functional unit, can be stored in a computer read/write memory medium.Above-mentioned SFU software functional unit is stored in a storage medium, comprising some instructions in order to make a computer equipment (can be personal computer, server, or the network equipment etc.) perform the part steps of method described in each embodiment of the present invention.And aforesaid storage medium comprises: USB flash disk, portable hard drive, ROM (read-only memory) (Read-Only Memory, be called for short ROM), random access memory (Random Access Memory, be called for short RAM), magnetic disc or CD etc. various can be program code stored medium.

Last it is noted that above embodiment is only in order to illustrate technical scheme of the present invention, be not intended to limit; Although with reference to previous embodiment to invention has been detailed description, those of ordinary skill in the art is to be understood that: it still can be modified to the technical scheme described in foregoing embodiments, or carries out equivalent replacement to wherein portion of techniques feature; And these amendments or replacement, do not make the essence of appropriate technical solution depart from the spirit and scope of various embodiments of the present invention technical scheme.

Claims (10)

1. a resource scheduling system, is characterized in that, comprises management node and physical node, wherein:

Management node, the resource collection that when reaching the target of described scheduling strategy for obtaining according to the scheduling strategy of user, each application program is corresponding, described resource collection comprises the stock number that when arbitrary application program runs on virtual machine, each virtual machine takies, and described resource collection is sent to each physical node;

Described physical node, for the least resource amount determining that according to described resource collection on described physical node, each virtual machine distributes, then the surplus yield of described physical node after the described least resource amount of distribution is proportionately distributed to each virtual machine on described physical node;

Described physical node, the total resources also for being obtained by each virtual machine distributed on rear described physical node sends to described management node;

Described management node, the current resource use amount of each virtual machine also on the total resources corresponding according to each virtual machine on described physical node and described physical node obtains the resource utilization of described physical node, and then dispatches the resource load of each physical node according to described resource utilization.

2. system according to claim 1, is characterized in that, described management node specifically for:

According to the utility function of the scheduling strategy of described user obtain reach the target of described scheduling strategy time resource collection corresponding to each application program;

Wherein, described utility function comprises:

$f\_\mathrm{app}\left({\mathrm{app}}_k\right)=\underset{t=1}{\overset{s_k}{\mathrm{\&Sigma;}}}{\cos t}_t*R_t*f\_\mathrm{time}(R_t,\mathrm{missio}{n}_t),f\_\mathrm{app}\left({\mathrm{app}}_k\right)$ Represent the target of described scheduling strategy, described target is the shortest time that described user buys the minimum or described virtual machine of the expense of described resources of virtual machine and finishes the work, cost _trepresent t the virtual machine VM distributing to a kth application program _tthe unit resource unit interval in price, R _trepresent the stock number that t the virtual machine distributing to a described kth application program takies, f_time (R _t, mission _t) represent that described t virtual machine completes distributed task amount mission _ttime shared time.

3. system according to claim 1 and 2, is characterized in that, described physical node specifically for:

The stock number that when running on described virtual machine according to arbitrary application program described in the described resource collection that described each application program is corresponding, each virtual machine takies, determine to belong to the least resource amount that each virtual machine on described physical node should distribute, the least resource amount that the arbitrary virtual machine on described physical node takies is the stock number sum that described arbitrary virtual machine is corresponding in described each application program;

Obtain the surplus yield of described physical node, described surplus yield is that the total resources of described physical node deducts the described least resource amount determined for each virtual machine on described physical node;

Described surplus yield is proportionately distributed to each virtual machine on described physical node, the described surplus yield that arbitrary virtual machine on described physical node is assigned to is that the least resource amount of described arbitrary virtual machine accounts for the ratio of the stock number that described physical node has distributed and the product of described surplus yield, and described stock number of having distributed is the least resource amount sum of each virtual machine on described physical node.

4. system according to claim 3, is characterized in that, described management node specifically for:

According to formula obtain the resource utilization of described physical node, represent the resource utilization of i-th physical node, represent the total resources that a jth virtual machine on described i-th physical node is corresponding, represent the total resources that described i-th physical node is corresponding, represent the resource use amount that a jth virtual machine on described i-th physical node is current;

If described management node needs the load of balanced each physical node described, then when described in obtaining, the variance of the resource utilization of each physical node is minimum, the resources of virtual machine distribution on each physical node described;

If described management node needs the number reducing described physical node, then first according to the total resources order from big to small of each physical node described, each physical node described is arranged, then successively virtual machine is dispensed on physical node, again next physical node is distributed until the total resources of the physical node of current distribution has all been assigned with;

Determine, after being all dispensed on physical node by all virtual machines, whether to there is the physical node of unallocated virtual machine, if exist, then close the physical node of described unallocated virtual machine.

5. a resource regulating method, is characterized in that, comprising:

Management node according to the scheduling strategy of user obtain reach the target of described scheduling strategy time resource collection corresponding to each application program, described resource collection comprises the stock number that when arbitrary application program runs on virtual machine, each virtual machine takies, and described resource collection is sent to each physical node;

Described physical node determines according to described resource collection the least resource amount that on described physical node, each virtual machine distributes, then the surplus yield of described physical node after the described least resource amount of distribution is proportionately distributed to each virtual machine on described physical node;

The total resources that each virtual machine distributed on rear described physical node obtains is sent to described management node by described physical node;

The described management node resource use amount current according to each virtual machine on total resources corresponding to each virtual machine on described physical node and described physical node obtains the resource utilization of described physical node, and then dispatches the resource load of each physical node according to described resource utilization.

6. method according to claim 5, is characterized in that, described management node according to the scheduling strategy of user obtain reach the target of described scheduling strategy time resource collection corresponding to each application program comprise:

Described management node according to the utility function of the scheduling strategy of described user obtain reach the target of described scheduling strategy time resource collection corresponding to each application program;

Wherein, described utility function comprises:

$f\_\mathrm{app}\left({\mathrm{app}}_k\right)=\underset{t=1}{\overset{s_k}{\mathrm{\&Sigma;}}}{\cos t}_t*R_t*f\_\mathrm{time}(R_t,\mathrm{missio}{n}_t),f\_\mathrm{app}\left({\mathrm{app}}_k\right)$ Represent the target of described scheduling strategy, described target is the shortest time that described user buys the minimum or described virtual machine of the expense of described resources of virtual machine and finishes the work, cost _trepresent t the virtual machine VM distributing to a kth application program _tthe unit resource unit interval in price, R _trepresent the stock number that t the virtual machine distributing to a described kth application program takies, f_time (R _t, mission _t) represent that described t virtual machine completes distributed task amount mission _ttime shared time.

7. the method according to claim 5 or 6, it is characterized in that, described physical node determines according to described resource collection the least resource amount that on described physical node, each virtual machine distributes, then is comprised by each virtual machine that the surplus yield of described physical node after the described least resource amount of distribution is proportionately distributed on described physical node:

The stock number that when described physical node runs on described virtual machine according to arbitrary application program described in described resource collection corresponding to described each application program, each virtual machine takies, determine to belong to the least resource amount that each virtual machine on described physical node should distribute, the least resource amount that the arbitrary virtual machine on described physical node takies is the stock number sum that described arbitrary virtual machine is corresponding in described each application program;

Described physical node obtains the surplus yield of described physical node, and described surplus yield is that the total resources of described physical node deducts the described least resource amount determined for each virtual machine on described physical node;

Described surplus yield is proportionately distributed to each virtual machine on described physical node by described physical node, the described surplus yield that arbitrary virtual machine on described physical node is assigned to is that the least resource amount of described arbitrary virtual machine accounts for the ratio of the stock number that described physical node has distributed and the product of described surplus yield, and described stock number of having distributed is the least resource amount sum of each virtual machine on described physical node.

8. method according to claim 7, it is characterized in that, the current resource use amount of each virtual machine on the described total resources corresponding according to each virtual machine on described physical node and described physical node obtains the resource utilization of described physical node, and then comprises according to the resource load that described resource utilization dispatches each physical node:

According to formula obtain the resource utilization of described physical node, represent the resource utilization of i-th physical node, represent the total resources that a jth virtual machine on described i-th physical node is corresponding, represent the total resources that described i-th physical node is corresponding, represent the resource use amount that a jth virtual machine on described i-th physical node is current;

If described management node needs the load of balanced each physical node described, then when described in obtaining, the variance of the resource utilization of each physical node is minimum, the resources of virtual machine distribution on each physical node described;

If described management node needs the number reducing described physical node, then first according to the total resources order from big to small of each physical node described, each physical node described is arranged, then successively virtual machine is dispensed on physical node, again next physical node is distributed until the total resources of the physical node of current distribution has all been assigned with;

Determine, after being all dispensed on physical node by all virtual machines, whether to there is the physical node of unallocated virtual machine, if exist, then close the physical node of described unallocated virtual machine.

9. a management node, is characterized in that, comprises virtual scheduler module, data memory module, communication module and data monitoring module, wherein:

Described virtual scheduler module, the resource collection that when reaching the target of described scheduling strategy for obtaining according to the scheduling strategy of user, each application program is corresponding, described resource collection comprises the stock number that when arbitrary application program runs on virtual machine, each virtual machine takies;

Described data memory module, for storing resource collection corresponding to described each application program;

Communication module, sends to each physical node for the described resource collection stored by described data memory module;

Described communication module, also for total resources that each virtual machine receiving described physical node obtains;

Described data monitoring module, the current resource use amount of each virtual machine on the total resources that each virtual machine on the described physical node received according to described communication module is corresponding and described physical node obtains the resource utilization of described physical node;

Described virtual scheduler module, the described resource utilization also for obtaining according to described data monitoring module dispatches the resource load of each physical node.

10. a physical node, is characterized in that, comprises communication module and virtual scheduler module, wherein:

Described communication module, for the resource collection of each application program that receiving management node sends, described resource collection comprises the stock number that when arbitrary application program runs on virtual machine, each virtual machine takies;

Described virtual scheduler module, for the least resource amount determining that according to described resource collection on described physical node, each virtual machine distributes, then the surplus yield of described physical node after the described least resource amount of distribution is proportionately distributed to each virtual machine on described physical node;

Described communication module, the total resources also for being obtained by each virtual machine distributed on rear described physical node sends to described management node, dispatches the resource load of each physical node to make described management node according to described total resources.