CN104184813A - Load balancing method of virtual machines, related equipment and trunking system - Google Patents

Abstract

The invention discloses a load balancing method of virtual machines, related equipment and a trunking system. The network performance of the trunking system can be improved after load balancing. According to certain feasible execution modes, the method includes the steps that a management node determines a plurality of virtual machines in which load balancing is needed in the trunking system; according to the network flow relation among the virtual machines, the virtual machines are integrated into at least one virtual machine cluster, so that the network flow between each virtual machine and at least one other virtual machine in the same virtual machine cluster is larger than or equal to a network flow threshold value, and the network flow between each virtual machine and any other virtual machine in the different virtual machine cluster is smaller than the network flow threshold value; migration suggestions are generated, and the migration suggestions are used for instructing all the virtual machines included in each virtual machine cluster to be migrated or hosted to the same target computational node; the migration suggestions are sent to source computational nodes hosted by one or more virtual machines included in each virtual machine cluster.

Description

The load-balancing method of virtual machine and relevant device and group system

Technical field

The present invention relates to computer and communication technical field, be specifically related to a kind of load-balancing method and relevant device and group system of virtual machine.

Background technology

Server virtualization technology is the key technology of cloud computing, this technology is by carrying out virtual to physical server (or being called physical host), realize at many virtual machines of separate unit physical server deploy (Virtual Machine, VM), improve the resource utilization of physical server.Many virtualized physical servers can form a virtual cluster, and virtual cluster, by abstract the physical resource in the cluster resource pool for the various resource compositions such as storage, calculating, provides virtual machine to user by the mode of on-demand application resource.

A key property of virtual cluster is dynamic resource scheduling (Dynamic Resources Scheduling, DRS).DRS can present to user by the resource such as calculating, the storage unification of many physical hosts in cluster, can utilize virtual machine (vm) migration technology, do not affecting customer service or in not perception of user in the situation that, virtual machine is moved at different physical hosts, use focus to eliminate in time resource, improve resource utilization.

In prior art, for example prestige farsighted (VMware) DRS, can carry out by calculating the resource load of central processing unit (CPU), two dimensions of internal memory (Memory, MEM) the computational resource load of computing node in gauge cluster (above said physical server or physical host).By the measurement to computational resource load in virtual cluster and calculating, VMware DRS has realized following functions: (1) initial placement: while starting virtual machine, virtual machine is placed on the lower computing node of computational resource load; (2) load balancing: monitor in real time the computational resource loading condition of each computing node in virtual cluster, generating virtual machines migration suggestion, by the virtual machine (vm) migration of high capacity computing node to the computing node of low load; (3) energy optimization: in conjunction with DPM (Distributed Power Management, distributed power supply management) function, according to the load of each computing node, computing node is carried out to lower electricity or the operation that powers on, optimize energy consumption.

Practice discovery, the DRS function of prior art is carried out load balancing according to computational resource loading condition, has following defect: if performance bottleneck appears in cluster network link, be easy to affect the network performance of virtual machine in cluster; Inappropriate virtual machine (vm) migration suggestion can cause the decline of virtual machine network performance.

Summary of the invention

The embodiment of the present invention provides a kind of load-balancing method and relevant device and group system of virtual machine, to improve the network performance of system of virtual cluster after load balancing.

First aspect present invention provides a kind of load-balancing method of virtual machine, described method is for comprising the group system of management node and multiple computing nodes, and the each computing node in described multiple computing nodes comprises hardware layer, operates in the host on described hardware layer and operates in the virtual machine on described host; Described method comprises: described management node is determined the multiple virtual machines that need to carry out load balancing in described group system, and described multiple virtual machines distribute and operate on the part or all of computing node of the described multiple computing nodes in described group system; According to the network traffics relation between described multiple virtual machines, described multiple virtual machines are divided into at least one virtual machine bunch, make, network traffics between at least one other virtual machine in each virtual machine and same virtual machine bunch are more than or equal to network traffics threshold value, and the network traffics between any other virtual machine in each virtual machine and different virtual machine bunch are less than network traffics threshold value; Cluster into migration suggestion for each virtual machine, described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host on same target computing node; The one or more virtual machine hosts' that comprise to described virtual machine bunch source computing node sends described migration suggestion so that whole virtual machines that described virtual machine bunch comprises are migrated to or host on described target computing node.

In conjunction with first aspect present invention, in the possible implementation of the first, described according to the network traffics relation between described multiple virtual machines, described multiple virtual machines are divided into at least one virtual machine bunch to be comprised: according to the network traffics relation between described multiple virtual machines, the network service that builds described multiple virtual machines is communicated with topology; Described network service is communicated with to topology to be cut apart, each the group virtual machine that network traffics is each other more than or equal to network traffics threshold value is defined as a virtual machine bunch, and, each isolated virtual machine is defined as separately to a virtual machine bunch, and the network traffics of described isolated virtual machine and other any virtual machine are all less than network traffics threshold value.

In conjunction with the possible implementation of the first of first aspect present invention or first aspect, in the possible implementation of the second, described management node is determined needs the multiple virtual machines that carry out load balancing to comprise in described group system: in the whole virtual machines that move from described group system, filter out the virtual machine each other with network traffics relation, the virtual machine each other described in determining with network traffics relation is the multiple virtual machines that need to carry out load balancing in described group system.

In conjunction with the first or the possible implementation of the second of first aspect present invention or first aspect, in the third possible implementation, before described management node is determined the multiple virtual machines that need to carry out load balancing in described group system, also comprise: described management node obtains the resource usage data of described group system, described resource usage data comprises cpu busy percentage and the memory usage of each computing node in described group system, and the network resource utilization of each link between each computing node; According to the balanced desired value of described resource usage data computational load, described load balancing desired value is for representing the load balancing state of described group system; If described load balancing desired value is greater than load balancing threshold value, judge that described group system need to carry out load balancing.

In conjunction with the third possible implementation of first aspect present invention, in the 4th kind of possible implementation, describedly comprise according to the balanced desired value of described resource usage data computational load: adopt the balanced desired value of following formula computational load; T (S)=ω ₁× σ (Util _cPU)+ω ₂× σ (Util _mEM)+ω ₃× σ (Util _nET)+ω ₄× δ _vM(S);

Wherein, S represents the Topological Mapping relation of virtual machine and computing node, and T (S) represents load balancing desired value, σ (Util _cPU) represent the variance of the cpu busy percentage of each computing node, σ (Util _mEM) represent the variance of the memory usage of each computing node, σ (Util _nET) represent the variance of the network resource utilization of each link between each computing node, δ _vM(S) represent all virtual machines that move in described group system average weighted postpone, ω ₁, ω ₂, ω ₃, ω ₄it is respectively weight coefficient.

In conjunction with the first or the possible implementation of the second of first aspect present invention or first aspect, in the 5th kind of possible implementation, describedly cluster into migration suggestion for each virtual machine, described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host comprises on same target computing node: select multiple candidate's computing nodes for each virtual machine bunch, analog computation each virtual machine bunch be migrated to or the candidate computing node of host in described multiple candidate's computing nodes after the load balancing desired value of described group system, for each virtual machine bunch, from described multiple candidate's computing nodes, determine a target computing node, make whole virtual machine (vm) migrations in each virtual machine bunch to or the load balancing desired value minimum of host's described group system after selected target computing node, cluster into migration suggestion for each virtual machine, described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host on the target computing node of described selection.

In conjunction with the 5th kind of possible implementation of first aspect present invention, in the 6th kind of possible implementation, described analog computation each virtual machine bunch be migrated to or the candidate computing node of host in described multiple candidate's computing nodes after the load balancing desired value of described group system comprise: adopt the balanced desired value of following formula computational load;

T(S)＝ω ₁×σ(Util _CPU)+ω ₂×σ(Util _MEM)+ω ₃×σ(Util _NET)+ω ₄×δ _VM(S)+ω ₅×cost(S)

Wherein, S represents the Topological Mapping relation of virtual machine and computing node, and T (S) represents load balancing desired value, σ (Util _cPU) represent the variance of the cpu busy percentage of each computing node, σ (Util _mEM) represent the variance of the memory usage of each computing node, σ (Util _nET) represent the variance of the network resource utilization of each link between each computing node, δ _vM(S) average weighted that represents all virtual machines that move in described group system postpones, and cost (S) represents group system mapping cost, ω ₁, ω ₂, ω ₃, ω ₄, ω ₅it is respectively weight coefficient.

Second aspect present invention provides a kind of management node, for comprising the group system of multiple computing nodes and described management node, the each computing node in described multiple computing nodes comprises hardware layer, operates in the host on described hardware layer and operates in the virtual machine on described host; Described management node comprises:

Determination module, for determining that described group system needs to carry out multiple virtual machines of load balancing, described multiple virtual machines distribute and operate on the part or all of computing node of the described multiple computing nodes in described group system;

Sub-clustering module, be used for according to the network traffics relation between described multiple virtual machines, described multiple virtual machines are divided into at least one virtual machine bunch, make, network traffics between at least one other virtual machine in each virtual machine and same virtual machine bunch are more than or equal to network traffics threshold value, and the network traffics between any other virtual machine in each virtual machine and different virtual machine bunch are less than network traffics threshold value;

Suggestion module, for cluster into migration suggestion for each virtual machine, described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host on same target computing node;

Sending module, sends described migration suggestion for one or more virtual machine hosts' of comprising to described virtual machine bunch source computing node so that whole virtual machines that described virtual machine bunch comprises are migrated to or host on described target computing node.

In conjunction with second aspect present invention, in the possible implementation of the first, described sub-clustering module comprises:

Construction unit, for according to the network traffics relation between described multiple virtual machines, builds the network service connection topology of described multiple virtual machines;

Division unit, for described network service connected graph is cut apart, each the group virtual machine that network traffics is each other more than or equal to network traffics threshold value is divided into a virtual machine bunch, and, each isolated virtual machine is defined as separately to a virtual machine bunch, and the network traffics of described isolated virtual machine and other any virtual machine are all less than network traffics threshold value.

In conjunction with the possible implementation of the first of second aspect present invention or second aspect, in the possible implementation of the second, described determination module specifically for: in the whole virtual machines that move from described group system, filter out the virtual machine each other with network traffics relation, the virtual machine each other described in determining with network traffics relation is the multiple plan machines of void that need to carry out load balancing in described group system.

In conjunction with the first or the possible implementation of the second of second aspect present invention or second aspect, in the third possible implementation, management node also comprises: acquisition module, for obtaining the resource usage data of described group system, described resource usage data comprises cpu busy percentage and the memory usage of each computing node in described group system, and the network resource utilization of each link between each computing node; Computing module, for according to the balanced desired value of described resource usage data computational load, described load balancing desired value is for representing the load balancing state of described group system; Judge module, if be greater than load balancing threshold value for described load balancing desired value, judges that described group system need to carry out load balancing.

In conjunction with the third possible implementation of second aspect present invention, in the 4th kind of possible implementation, described computing module is specifically for adopting the balanced desired value of following formula computational load;

T(S)＝ω ₁×σ(Util _CPU)+ω ₂×σ(Util _MEM)+ω ₃×σ(Util _NET)+ω ₄×δ _VM(S)

Wherein, S represents the Topological Mapping relation of virtual machine and computing node, and T (S) represents load balancing desired value, σ (Util _cPU) represent the variance of the cpu busy percentage of each computing node, σ (Util _mEM) represent the variance of the memory usage of each computing node, σ (Util _nET) represent the variance of the network resource utilization of each link between each computing node, δ _vM(S) represent all virtual machines that move in described group system average weighted postpone, ω ₁, ω ₂, ω ₃, ω ₄it is respectively weight coefficient.

In conjunction with the first or the possible implementation of the second of second aspect present invention or second aspect, in the 5th kind of possible implementation, described suggestion module comprises:

Analogue computer, for selecting multiple candidate's computing nodes for each virtual machine bunch, analog computation each virtual machine bunch be migrated to or the candidate computing node of host in described multiple candidate's computing nodes after the load balancing desired value of described group system;

Determining unit, be used for for each virtual machine bunch, from described multiple candidate's computing nodes, determine a target computing node, make whole virtual machine (vm) migrations in each virtual machine bunch to or the load balancing desired value minimum of host's described group system after selected target computing node;

Suggestion generation unit, for cluster into migration suggestion for each virtual machine, described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host at the target computing node of described selection.

In conjunction with the 5th kind of possible implementation of second aspect present invention, in the 6th kind of possible implementation, described analogue computer specifically for: adopt the balanced desired value of following formula computational load;

T(S)＝ω ₁×σ(Util _CPU)+ω ₂×σ(Util _MEM)+ω ₃×σ(Util _NET)+ω ₄×δ _VM(S)+ω ₅×cost(S)

Wherein, S represents the Topological Mapping relation of virtual machine and computing node, and T (S) represents load balancing desired value, σ (Util _cPU) represent the variance of the cpu busy percentage of each computing node, σ (Util _mEM) represent the variance of the memory usage of each computing node, σ (Util _nET) represent the variance of the network resource utilization of each link between each computing node, δ _vM(S) average weighted that represents all virtual machines that move in described group system postpones, cost (S) representative system mapping cost, ω ₁, ω ₂, ω ₃, ω ₄, ω ₅it is respectively weight coefficient.

Third aspect present invention provides a kind of virtual cluster, comprising: multiple computing nodes and the management node as described in second aspect present invention; The migration suggestion that described computing node sends for receiving described management node, carries out virtual machine (vm) migration operation according to the instruction of described migration suggestion.

Therefore, in some execution modes of the present invention, according to the network traffics relation between virtual machine, multiple virtual machines that need to carry out load balancing are divided at least one virtual machine bunch, make, network traffics between at least one other virtual machine in each virtual machine and same virtual machine bunch are more than or equal to network traffics threshold value, and the network traffics between any other virtual machine in each virtual machine and different virtual machine bunch are less than network traffics threshold value; Cluster into migration suggestion for the each virtual machine in described multiple virtual machines bunch, described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host on same target computing node; Obtain following technique effect:

In the time carrying out load balancing, consider the impact of network dimension, by the multiple virtual machines that have in the same virtual machine bunch of network traffics relation are focused on same computing node, thereby, network traffics between multiple virtual machines in same virtual machine bunch can not take the load of the link between computing node, the offered load that can reduce each link in group system with this, there is performance bottleneck in prevention link; And, by the multiple virtual machines in same virtual machine bunch are focused on same computing node, the network bandwidth and communication reliability between the virtual machine in same virtual machine bunch can be improved, thereby the network performance of the virtual machine bunch of traffic-intensive can be promoted.

Brief description of the drawings

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

Fig. 1 is the schematic diagram of a kind of group system of the embodiment of the present invention;

Fig. 2 a is the schematic flow sheet of the load-balancing method of a kind of virtual machine of providing of the embodiment of the present invention;

Fig. 2 b is the schematic flow sheet of the load-balancing method of the another kind of virtual machine that provides of the embodiment of the present invention;

Fig. 3 a is the structural representation of a kind of management node of providing of the embodiment of the present invention;

Fig. 3 b is the structural representation of the another kind of management node that provides of the embodiment of the present invention;

Fig. 4 a is the schematic diagram that the structure of a group system of prior art and virtual machine wherein distribute;

Fig. 4 b is the schematic diagram that the structure of a group system of the embodiment of the present invention and virtual machine wherein distribute;

Fig. 5 is the schematic diagram of the applied data center of the embodiment of the present invention;

Fig. 6 a is the structural representation of the load balancing apparatus of the management node deploy of the embodiment of the present invention;

Fig. 6 b is the schematic diagram of the cluster resource management system moved on the management node of the embodiment of the present invention;

Fig. 7 is the flow chart that the management node of the embodiment of the present invention is carried out load balance scheduling;

Fig. 8 is the flow chart that the management node of the embodiment of the present invention judges whether to carry out load balancing and generating virtual machines migration suggestion;

Fig. 9 is the schematic flow sheet of partition virtual machines bunch in the embodiment of the present invention;

Figure 10 is the schematic diagram of partition virtual machines bunch in scene example of the present invention;

Figure 11 is the structural representation of the another kind of management node that provides of the embodiment of the present invention.

Embodiment

The embodiment of the present invention provides a kind of load-balancing method and relevant device and group system of virtual machine, to improve the network performance of group system after load balancing.

In order to make those skilled in the art person understand better the present invention program, below in conjunction with the accompanying drawing in the embodiment of the present invention, technical scheme in the embodiment of the present invention is clearly and completely described, obviously, described embodiment is only the embodiment of a part of the present invention, instead of whole embodiment.Based on the embodiment in the present invention, those of ordinary skill in the art, not making the every other embodiment obtaining under creative work prerequisite, should belong to the scope of protection of the invention.

To facilitate understanding of the present embodiment of the invention the several key elements that, first can introduce in this introduces the applied system of the embodiment of the present invention and describes.

Embodiment of the present invention technical scheme, is applied to system of virtual cluster (being called for short virtual cluster or group system or cluster), and as shown in Figure 1, this group system can comprise management node 310 and computing node 320; Management node can have one or more, for example, can be two, is divided into main management node and standby management node; Computing node can have multiple.Said management node 310 and computing node 320 are all computer equipments, and management node 310 also can be described as management server, and computing node 320 also can be described as physical host.

Wherein, computing node 320 can comprise hardware layer 3201, operates in the host 3202 on hardware layer and operate at least one the virtual machine VM on described host 3202.Describe in detail below:

Virtual machine (Virtual Machine, VM):

Virtual machine refer to by software simulation have complete hardware system function, operate in a computer system in complete isolation environment.Can on a physical host, simulate one or many virtual machines by software virtual machine, and these virtual machines carry out work just as real computer, can installing operating system and application program on virtual machine, virtual machine is addressable network resource also.For the application program of moving in virtual machine, similarly be in real computer, to carry out work.

Hardware layer:

The hardware platform of virtualized environment operation.Wherein, hardware layer can comprise multiple hardwares, the for example hardware layer of certain computing node or management node can comprise processor (for example CPU) and memory (for example internal memory), can also comprise network interface card, memory etc. high speed/low speed I/O (I/O, Input/Output) equipment, and there is the miscellaneous equipment of particular procedure function, as input and output memory management unit (IOMMU, Input/Output Memory Management Unit), wherein IOMMU can be used for the conversion of virtual machine physical address and Host physical address.

Host (Host):

As management level, in order to complete management, the distribution of hardware resource; For virtual machine presents virtual hardware platform; Realize scheduling and the isolation of virtual machine.Wherein, Host may be monitor of virtual machine (VMM, Virtual Machine Monitor); In addition, VMM and 1 franchise virtual machine coordinate sometimes, and both are in conjunction with composition Host.Wherein, virtual hardware platform provides various hardware resources to each virtual machine of operation on it, as virtual cpu, internal memory, virtual disk, Microsoft Loopback Adapter etc. are provided.Wherein, this virtual disk can corresponding Host a file or a logical block equipment.Virtual machine operates on the virtual hardware platform that Host is its preparation, the one or more virtual machines of the upper operation of Host.

Below by specific embodiment, be described in detail respectively.

Please refer to Fig. 2 a, the embodiment of the present invention provides a kind of load-balancing method of virtual machine.The method can be used for comprising the group system of management node and multiple computing nodes, and wherein, computing node can comprise hardware layer, operates in the host on hardware layer and operate in the virtual machine on host.

In prior art, system of virtual cluster is only according to computational resource load, virtual machine in cluster is carried out to load balancing, can not improve the problem of the network performance of virtual machine in cluster, in embodiment of the present invention technical scheme, introduce the utilization to Internet resources, in the time carrying out equally loaded, considered the impact of network factors, improved the network performance of the virtual cluster after load balancing with this.

Please refer to Fig. 2 a, embodiment of the present invention method can comprise:

110, management node is determined the multiple virtual machines that need to carry out load balancing in described group system; Described multiple virtual machine distributes and operates on the part or all of computing node of the described multiple computing nodes in described group system;

120, management node, according to the network traffics relation between described multiple virtual machines, is divided at least one virtual machine bunch by described multiple virtual machines; Make, network traffics between at least one other virtual machine in each virtual machine and same virtual machine bunch are more than or equal to network traffics threshold value, and the network traffics between any other virtual machine in each virtual machine and different virtual machine bunch are less than network traffics threshold value;

130, management node clusters into migration suggestion for each virtual machine, and described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host on same target computing node;

140, the one or more virtual machine hosts' that comprise to described virtual machine bunch source computing node sends described migration suggestion so that whole virtual machines that described virtual machine bunch comprises are migrated to or host on described target computing node.

Please refer to Fig. 2 b, in some embodiments of the invention, before 110, can also comprise:

100, management node judges whether group system needs to carry out load balancing.

Further describe below:

100, management node judges whether group system needs to carry out load balancing.

As the manager of group system, management node provides the functions such as the life cycle management, resource scheduling management, O&M of computing node and virtual machine, is the brain of whole group system.Management node can Real-time Collection group system various performance datas, comprise resource usage data and configuration parameter etc., wherein resource usage data refers to that each resource (comprises computational resource, Internet resources, storage resources etc.) usage data (as utilance), the computational resource that for example can comprise each computing node is as CPU, the usage data of Memory, such as CPU of the computational resource of each virtual machine, the usage data of Memory, the Internet usage data of each virtual machine, network traffics relation between virtual machine, the offered load of the link between every two computing nodes etc., and can save as historical data.

Management node can be according to the system performance information of Real-time Collection, realizes dynamic resource scheduling (DRS) function, and the virtual machine in group system is carried out to load balancing.First, can carry out load-balancing algorithm, judge whether group system needs to carry out load balancing, if needed, just start to carry out subsequent step.The method that judges whether to carry out load balancing has multiple, general, and this determining step can comprise:

Obtain the resource usage data of described group system; Described resource usage data comprises cpu busy percentage and the memory usage of each computing node in described group system, and the network resource utilization of each link between each computing node; According to the balanced desired value of described resource usage data computational load, described load balancing desired value is for representing the load balancing state of described group system; If described load balancing desired value is greater than load balancing threshold value, judge that group system need to carry out load balancing.

In a kind of execution mode, the cpu busy percentage of each computing node of management node in specifically can Gains resources usage data, and then, according to the cpu busy percentage of each computing node, calculate the variance of the cpu busy percentage of each computing node in group system, as load balancing desired value; If load balancing desired value is greater than load balancing threshold value, is judged as and need to carries out load balancing.Popular point says, variance is exactly the degree departing from central value, and variance is larger, illustrates that the mutual difference of cpu busy percentage of each computing node is larger, just more need to carry out load balancing.Illustrate, the CPU capacity of supposing a computing node is 10GHz, the cpu resource that every the virtual machine moving on this computing node on average takies in a period of time is in the past 2GHz, if operation has 3 such virtual machines on this computing node, the cpu busy percentage of this computing node is 60%.Under limiting case, suppose that the cpu busy percentage of certain computing node has reached 100%, the cpu busy percentage of another computing node is 0, and variance at this moment just may be very large, exceedes threshold value, need to carry out load balancing.

In another kind of execution mode, cpu busy percentage and internal memory (MEM) utilance of each computing node of management node in specifically can Gains resources usage data, and then, according to the cpu busy percentage of each computing node, calculate the variance of the cpu busy percentage of each computing node in group system, and, the variance of the MEM utilance of each computing node, with the weighted sum of two variances calculating, as load balancing desired value.According to the difference of practical application scene, to CPU and two dimensions of MEM, can set different weights.Illustrate, the memory size of supposing a computing node is 10GB, and the memory source that every the virtual machine moving on this computing node on average takies in a period of time is in the past 2GB; If operation has 3 such virtual machines on this computing node, the memory source utilance of this computing node is 60%.The weight of supposing CPU and two dimensions of MEM is respectively a and b, with σ (Util _cPU) represent the variance of the cpu busy percentage of each computing node, with σ (Util _mEM) represent the variance of the MEM utilance of each computing node, load balancing desired value can equal a × σ (Util _cPU)+b × σ (Util _mEM).

One of the present invention preferred embodiment in, management node specifically can obtain the following resource usage data of group system: the cpu busy percentage of each computing node and memory usage, and the network resource utilization of each link between each computing node; Can adopt the balanced desired value of following formula computational load;

T(S)＝ω ₁×σ(Util _CPU)+ω ₂×σ(Util _MEM)+ω ₃×σ(Util _NET)+ω ₄×δ _VM(S)+ω ₅×cost(S)

Wherein, T (S) represents load balancing desired value; S represents the Topological Mapping relation of virtual machine and computing node, that is, which virtual machine operates on that computing node;

σ represents variance, concrete, σ (Util _cPU) represent the variance of the cpu busy percentage of each computing node, σ (Util _mEM) represent the variance of the memory usage of each computing node, σ (Util _nET) represent the variance of the network resource utilization of each link between each computing node; Here said link, refers to the link that connects any two computing nodes in group system;

δ _vM(S) average weighted of all virtual machines that move in expression group system postpones,

Cost (S) represents group system mapping cost, after specifically can referring to each virtual machine (vm) migration, and before this virtual machine (vm) migration, the mapping cost of group system; In this step, in order to judge whether to carry out load balancing, when the balanced desired value of computational load, owing to not yet carrying out virtual machine (vm) migration, cost (S)=0, therefore, can make do not comprise+ω of above-mentioned formula ₅× cost (S) this or make cost (S)=0.

ω ₁, ω ₂, ω ₃, ω ₄, ω ₅be respectively weight coefficient, can manually set according to actual needs.

In above-mentioned formula, adopt σ (Util _cPU), σ (Util _mEM), σ (Util _nET), δ _vM(S), these five calculation of parameter load balancing desired values of cost (S), can reflect more accurately the load balance degree of group system.The T (S) that this formula calculates is less, represents that load balance degree is more excellent, and T (S) is larger, represents that load balance degree is poorer.In the time that T (S) is greater than load balancing threshold value, need to carry out load balancing with regard to explanation.

In some embodiments of the invention, also can not consider δ _vM(S), one or two in these two parameters of cost (S), load balancing desired value can be: T (S)=ω ₁× σ (Util _cPU)+ω ₂× σ (Util _mEM)+ω ₃× σ (Util _nET).In other embodiment, in above-mentioned computing formula, can also increase according to actual needs other parameter.In other embodiment, also can be without variance, but utilize standard deviation to calculate, the symbol σ in above-mentioned formula also can refer to standard deviation.

Below, the σ in formula above represents that standard deviation is as example, further illustrates how according to the balanced desired value of above-mentioned formula computational load.

(1) calculate the standard deviation sigma (Util of the cpu busy percentage of each computing node _cPU)

Suppose to be numbered the computing node PM of i _ithe actual use amount of CPU be a _i, the capacity of this CPU is c _i, this computing node PM _icpu busy percentage be ρ _i=a _i/ c _i, suppose total n computing node, the average utilization μ of the CPU of this n computing node in group system _cpufor:

${\mathrm{\μ}}_{\mathrm{cpu}}=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\mathrm{\ρ}}_i$

Accordingly, the standard deviation of the cpu busy percentage of n computing node is:

$\mathrm{\σ}\left(\mathrm{Uti}{l}_{\mathrm{CPU}}\right)=\sqrt{\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{({\mathrm{\ρ}}_i-{\mathrm{\μ}}_{\mathrm{cpu}})}^2}$

(2) calculate the standard deviation sigma (Util of the memory usage of each computing node _mEM)

Suppose to be numbered the computing node PM of i _ithe actual use amount of internal memory be a _i, the capacity of this internal memory is c _i, this computing node PM _imemory usage be ρ _i=a _i/ c _i, suppose total n computing node, the average utilization μ of the internal memory of this n computing node in group system _cpufor:

${\mathrm{\μ}}_{\mathrm{MEM}}=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\mathrm{\ρ}}_i$

Accordingly, the standard deviation of the memory usage of n computing node is:

$\mathrm{\σ}\left(\mathrm{Uti}{l}_{\mathrm{MEM}}\right)=\sqrt{\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\left({\mathrm{\ρ}}_i\text{-}{\mathrm{\μ}}_{\mathrm{MEM}}\right)}^2}$

(3) calculate the standard deviation sigma (Util of the resource utilization of each link _mEM)

Link set in group system can be divided into core link and edge link, uses LK _corerepresent core link, use LK _edgerepresent edge link, and suppose LK _corein have n bar link, LK _edgein have m bar link, note LK _corein the resource utilization of i article of link be v _i, LK _corelink circuit resource utilance average be:

${\mathrm{\μ}}_{\mathrm{core}}=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{v}_i$

LK _corethe standard deviation of link circuit resource utilance be:

Note LK _edgein the resource utilization of i article of link be v _i, LK _edgelink circuit resource utilance average be:

${\mathrm{\μ}}_{\mathrm{edge}}=\frac{1}{m}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{v}_i$

LK _edgethe standard deviation of link circuit resource utilance be:

The standard deviation of total link utilization is:

σ(Util _NET)＝σ(Util _core)+σ(Util _edge)

(4) in calculating group system, the average weighted of all virtual machines postpones δ _vM(S)

Note is numbered the virtual machine vm of i _icomputing node pm (the vm at place _i), with the virtual machine vm that is numbered j _jcomputing node pm (the vm at place _j), the time delay of network service is between the two both weight coefficients are made as λ by experience _i,j,, in group system, all the average weighted of virtual machine postpones to be expressed as:

${\mathrm{\δ}}_{\mathrm{VM}}\left(S\right)=\frac{{\mathrm{\Σ}}_{{\mathrm{vm}}_i,{\mathrm{vm}}_j}({\mathrm{\λ}}_{i,j}\×{\mathrm{\δ}}_{\mathrm{pm}\left(v{m}_i\right),\mathrm{pm}\left(v{m}_j\right)})}{{\mathrm{\Σ}}_{{\mathrm{vm}}_i,{\mathrm{vm}}_j}\left({\mathrm{\λ}}_{i,j}\right)}$

Wherein, , that is, and vm _i, vm _jany two virtual machines in group system.

(5) calculate group system mapping cost cost (S)

$\cos t\left(S\right)=\frac{\mathrm{\ΣP}{C}_{\mathrm{vm}}+\mathrm{\ΣPM}{C}_{\mathrm{vm}}-\mathrm{\ΣP}{B}_{\mathrm{vm}}}{\mathrm{\α}\×\mathrm{\ΣP}{C}_{\mathrm{vm}}+\mathrm{\β}\×(\mathrm{\ΣP}{C}_{\mathrm{vm}}+\mathrm{\ΣPM}{C}_{\mathrm{vm}})}$

Wherein PC _vmrepresent the performance loss after virtual machine (vm) migration, PMC _vmrepresent the performance loss of virtual machine in transition process, PB _vmrepresent the performance boost of the rear virtual machine of migration, α, β are weight coefficients.

Group system mapping cost cost (S) operates the impact on whole group system performance for virtual machine (vm) migration.

Above, calculate respectively σ (Util _cPU), σ (Util _mEM), σ (Util _mEM), δ _vM(S), cost (S) afterwards, substitution formula:

T(S)＝ω ₁×σ(Util _CPU)+ω ₂×σ(Util _MEM)+ω ₃×σ(Util _NET)+ω ₄×δ _VM(S)+ω ₅×cost(S)

Obtain load balancing desired value T (S).

In other execution mode of the present invention, also can adopt the balanced desired value of alternate manner computational load, herein this is not construed as limiting.As long as can be used in the parameter of the load balance degree that represents group system, all can be used as load balancing desired value.As long as the load balancing desired value calculating is greater than load balancing threshold value, need to carry out load balancing with regard to explanation.This step that judges whether to carry out load balancing is optional step.

110, management node is determined the multiple virtual machines that need to carry out load balancing in group system, and described multiple virtual machines distribute and operate on the part or all of computing node of the described multiple computing nodes in described group system.

When being judged as need to carry out load balancing time, management node can be determined the multiple virtual machines that need to carry out load balancing in group system according to strategy.The definite method that need to carry out multiple virtual machines of load balancing has multiple, for example:

In a kind of execution mode, can need to carry out from the screening of computational resource angle the virtual machine of load balancing.Management node specifically can obtain the computational resource usage data of each node, for example comprise cpu busy percentage and memory usage, can first filter out the computing node (being the higher computing node of computational resource load) that computational resource utilance exceedes setting threshold, the computing node filtering out from these selects a part of virtual machine to be defined as carrying out the virtual machine of load balancing, adds in virtual machine list to be migrated.For example, can, from the computing node filtering out, select the highest one or more virtual machines of computational resource utilance, be defined as carrying out the virtual machine of load balancing.

In another kind of execution mode, can need to carry out from the screening of Internet resources angle the virtual machine of load balancing.Management node can be according to the Internet usage data of obtaining, in the whole virtual machines that move from group system, filter out the virtual machine each other with network traffics relation, be defined as carrying out the virtual machine of load balancing, that is: for all virtual machines in group system, by wherein all do not have the virtual machine of network traffics relation to reject with other any virtual machine, remaining virtual machine is defined as carrying out to the virtual machine of load balancing, add in virtual machine list to be migrated.

In the embodiment of the present invention, preferably adopt the combination of above-mentioned two kinds of modes, determine the multiple virtual machines that need to carry out load balancing in group system, that is: by the virtual machine that need to carry out load balancing from the screening of computational resource angle, with the virtual machine that need to carry out load balancing from the screening of Internet resources angle, all add in virtual machine list to be migrated, as the virtual machine that finally need to carry out load balancing.

Certainly, in other execution mode of the present invention, also can there is other screening mode, for example, can also need to carry out from the screening of storage resources angle the virtual machine of load balancing, also add in virtual machine list to be migrated.Herein for determining that the mode that need to carry out multiple virtual machines of load balancing does not limit.

120, management node is according to the network traffics relation between described multiple virtual machines, described multiple virtual machines are divided into at least one virtual machine bunch, make, network traffics between at least one other virtual machine in each virtual machine and same virtual machine bunch are more than or equal to network traffics threshold value, and the network traffics between any other virtual machine in each virtual machine and different virtual machine bunch are less than network traffics threshold value.

In the embodiment of the present invention, management node is before the virtual machine in virtual machine list to be migrated generates migration suggestion, according to the network traffics relation between virtual machine, can be one or more virtual machines bunch by the multiple virtual machine clusters that need load balancing, each virtual machine bunch comprises at least one virtual machine.

Said virtual machine bunch refers to one group of virtual machine with particular characteristic requirement, herein, network traffics between at least one other virtual machine in each virtual machine and same virtual machine bunch are more than or equal to network traffics threshold value, and the network traffics between any other virtual machine in each virtual machine and different virtual machine bunch are less than network traffics threshold value.

Or can understand like this: the multiple virtual machines that mark off bunch can be divided into two classes according to the number of included virtual machine, the virtual machine bunch that comprises at least two virtual machines is called to the first virtual machine bunch, the virtual machine bunch that only includes a virtual machine is called to the second virtual machine bunch, wherein, network traffics between at least one other virtual machine in any virtual machine in the first virtual machine bunch and same the first virtual machine bunch are greater than network traffics threshold value, the network traffics of the unique virtual machine in the second virtual machine bunch and any other virtual machine in described multiple virtual machine are less than network traffics threshold value.

In the embodiment of the present invention, Adoption Network flow compatibility algorithm, the multiple virtual machines that need load balancing are divided into at least one virtual machine bunch, specifically can comprise: carry out as required the network traffics relation between multiple virtual machines of load balancing, the network service that builds said multiple virtual machines is communicated with topology; Then network service being communicated with to topology cuts apart, each the group virtual machine that network traffics is each other more than or equal to network traffics threshold value is divided into a virtual machine bunch, and, each isolated virtual machine is defined as separately to a virtual machine bunch, and the network traffics of described isolated virtual machine and other any virtual machine are all less than network traffics threshold value.In other words, be communicated with topology for network service, establish a network traffics threshold value, if the network traffics between two virtual machines are less than this network traffics threshold value, be considered as not having network traffics relation between these two virtual machines, the connected relation between these two virtual machines is deleted; All virtual machines that network service is communicated with in topology carry out after above-mentioned processing, and all virtual machines are naturally divided into several groups, and wherein each group is exactly a virtual machine bunch.Multiple virtual machines in same virtual machine bunch may operate on same computing node, also may operate on different computing nodes.After marking off each virtual machine bunch, the further corresponding relation list of generating virtual machines bunch and main frame, list comprises the corresponding relation between each virtual machine in virtual machine bunch and each virtual machine host's computing node, thereby prepares for follow-up generation migration suggestion.

130, management node clusters into migration suggestion for each virtual machine, and described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host on same target computing node.

In the embodiment of the present invention, in management node generating virtual machines when suggestion migration,, taking virtual machine bunch as unit, the migration of generation is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host on same target computing node.General, the step of clustering into migration suggestion for each virtual machine can comprise:

Select multiple candidate's computing nodes for each virtual machine bunch, analog computation each virtual machine bunch be migrated to or the candidate computing node of host in described multiple candidate's computing nodes after the load balancing desired value of described group system; For each virtual machine bunch, from described multiple candidate's computing nodes, determine a target computing node, make whole virtual machine (vm) migrations in this virtual machine bunch to or the load balancing desired value minimum of host's group system after selected target computing node; Cluster into migration suggestion for each virtual machine, described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host on the target computing node of described selection.

In a kind of execution mode, select multiple candidate's computing nodes to comprise for each virtual machine bunch: filter out computational resource utilization rate lower than at least one computing node of computational resource load threshold as candidate's computing node.In another kind of execution mode, select multiple candidate's computing nodes to comprise for each virtual machine bunch: to suppose the virtual machine bunch to move to or in multiple random computing nodes one of host, carry out analog computation, filter out after virtual machine bunch migration, load balancing desired value is less than at least one computing node of load threshold, as candidate's computing node.In other execution mode, can also adopt other method screening candidate computing node, contrast herein does not limit.

Illustrate, suppose that group system comprises three computing nodes altogether, represents with A, B, C respectively, a virtual machine bunch that need to carry out load balancing comprises virtual machine VM1, VM2, VM3, VM4, VM5, and wherein VM1, VM2 operate in computing node A above, and it is upper that VM4, VM5 operate in computing node B, and VM3 operates on computing node C, can first filter out candidate's computing node, for example computing node B and C, then, according to the whole virtual machines in same virtual machine bunch are focused on to the principle on same computing node, can generate following candidate and move suggestion:, the candidate that virtual machine VM1, VM2, VM3 are moved on computing node B moves suggestion, and the candidate that VM1, VM2, VM4, VM5 are moved on computing node C moves suggestion, then, management node can move suggestion for each candidate and carry out analog computation, suppose that each candidate moves after suggestion execution, the load balancing desired value of analog computation group system, the candidate of load balancing desired value minimum is moved to suggestion and be defined as pending migration suggestion, for example, by virtual machine VM1, VM2, VM3 moves to migration cost that candidate on computing node B moves suggestion, and may to move the migration cost of suggestion than another candidate little, load balancing desired value after migration can be less, therefore, computing node B can be defined as to target computing node, can be by virtual machine VM1, VM2, the candidate that VM3 moves on computing node B moves suggestion, be defined as pending migration suggestion.Finally, by carrying out determined migration suggestion, the virtual machine (vm) migration of needs migration is arrived to target computing node, can realize the part virtual machine (vm) migration on high capacity computing node to low load calculation node, realize simultaneously the multiple virtual machines in same virtual machine bunch have been focused in same virtual machine, thereby reduced the load of communication link between computing node.

140, the one or more virtual machine hosts' that comprise to described virtual machine bunch source computing node sends described migration suggestion so that whole virtual machines that described virtual machine bunch comprises are migrated to or host on described target computing node.

After management node clusters into migration suggestion for each virtual machine, migration suggestion for each virtual machine bunch can be sent to each source computing node of virtual machine bunch included one or more virtual machine hosts, carry out migration suggestion by each source computing node, whole virtual machine (vm) migrations that a virtual machine bunch is comprised to or host on same target computing node.

Be appreciated that embodiment of the present invention such scheme for example can specifically implement at the computer equipment of the management node as group system.

Above, the embodiment of the invention discloses a kind of load-balancing method of virtual machine, the method adopts according to the network traffics relation between virtual machine, multiple virtual machines that need to carry out load balancing are divided at least one virtual machine bunch, make, network traffics between at least one other virtual machine in each virtual machine and same virtual machine bunch are more than or equal to network traffics threshold value, and the network traffics between any other virtual machine in each virtual machine and different virtual machine bunch are less than network traffics threshold value; Cluster into migration suggestion for the each virtual machine in described multiple virtual machines bunch, described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, moves to or the technical scheme of host on same target computing node; Obtain following technique effect:

In the time carrying out load balancing, consider the impact of network dimension, by the network traffics relation between Real-Time Monitoring virtual machine, mark off the virtual machine bunch with network traffics relation, as the thread of follow-up generation virtual machine (vm) migration suggestion;

Multiple virtual machines in same virtual machine bunch can be focused on same computing node, thereby, flow between multiple virtual machines in virtual machine bunch can not take the load of the link between computing node, the offered load that can reduce each link in group system with this, can prevent link to occur performance bottleneck;

And, by the multiple virtual machines in same virtual machine bunch are focused on same computing node, the network bandwidth and communication reliability between the virtual machine in same virtual machine bunch can be improved, thereby the network performance of the virtual machine bunch of traffic-intensive can be promoted.

In order better to implement the such scheme of the embodiment of the present invention, be also provided for coordinating the relevant apparatus of implementing such scheme below.

Please refer to Fig. 3 a, the embodiment of the present invention provides a kind of management node 200, be applied to the group system that comprises multiple computing nodes and described management node 200, the each computing node in described multiple computing nodes comprises hardware layer, operates in the host on hardware layer and operates at least one virtual machine on host; This management node 200 can comprise:

Determination module 210, for determining that described group system needs to carry out multiple virtual machines of load balancing, described multiple virtual machines distribute and operate on the part or all of computing node of the described multiple computing nodes in described group system;

Sub-clustering module 220, be used for according to the network traffics relation between described multiple virtual machines, described multiple virtual machines are divided into at least one virtual machine bunch, make, network traffics between at least one other virtual machine in each virtual machine and same virtual machine bunch are more than or equal to network traffics threshold value, and the network traffics between any other virtual machine in each virtual machine and different virtual machine bunch are less than network traffics threshold value;

Suggestion module 230, for cluster into migration suggestion for each virtual machine, described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host on same target computing node;

Sending module 240, sends described migration suggestion for one or more virtual machine hosts' of comprising to described virtual machine bunch source computing node so that whole virtual machines that described virtual machine bunch comprises are migrated to or host on described target computing node.

In some embodiments of the invention, described sub-clustering module 220 can comprise:

Construction unit, for according to the network traffics relation between described multiple virtual machines, builds the network service connection topology of described multiple virtual machines;

Division unit, for described network service connected graph is cut apart, each the group virtual machine that network traffics is each other more than or equal to network traffics threshold value is divided into a virtual machine bunch, and, each isolated virtual machine is defined as separately to a virtual machine bunch, and the network traffics of described isolated virtual machine and other any virtual machine are all less than network traffics threshold value.

In some embodiments of the invention, described determination module 210 can be specifically for: in the whole virtual machines that move from described group system, filter out the virtual machine each other with network traffics relation, the virtual machine each other described in determining with network traffics relation is the multiple plan machines of void that need to carry out load balancing in described group system.

Please refer to Fig. 3 b, in some embodiments of the invention, management node 200 can also comprise:

Acquisition module 250, for obtaining the resource usage data of described group system, described resource usage data comprises cpu busy percentage and the memory usage of each computing node in described group system, and the network resource utilization of each link between each computing node;

Computing module 260, for according to the balanced desired value of described resource usage data computational load, described load balancing desired value is for representing the load balancing state of described group system;

Judge module 270, if be greater than load balancing threshold value for described load balancing desired value, judges that described group system need to carry out load balancing.

In some embodiments of the invention, computing module 260 can be specifically for adopting the balanced desired value of following formula computational load;

T(S)＝ω ₁×σ(Util _CPU)+ω ₂×σ(Util _MEM)+ω ₃×σ(Util _NET)+ω ₄×δ _VM(S)；

Wherein, S represents the Topological Mapping relation of virtual machine and computing node, and T (S) represents load balancing desired value, σ (Util _cPU) represent the variance of the cpu busy percentage of each computing node, σ (Util _mEM) represent the variance of the memory usage of each computing node, σ (Util _nET) represent the variance of the network resource utilization of each link between each computing node, δ _vM(S) represent all virtual machines that move in described group system average weighted postpone, ω ₁, ω ₂, ω ₃, ω ₄it is respectively weight coefficient.

In some embodiments of the invention, suggestion module 230 can comprise:

Analogue computer, for selecting multiple candidate's computing nodes for each virtual machine bunch, analog computation each virtual machine bunch be migrated to or the candidate computing node of host in described multiple candidate's computing nodes after the load balancing desired value of described group system;

Determining unit, be used for for each virtual machine bunch, from described multiple candidate's computing nodes, determine a target computing node, make whole virtual machine (vm) migrations in each virtual machine bunch to or the load balancing desired value minimum of host's described group system after selected target computing node;

Suggestion generation unit, for cluster into migration suggestion for each virtual machine, described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host at the target computing node of described selection.

In some embodiments of the invention, analogue computer, can be specifically for adopting the balanced desired value of following formula computational load;

T(S)＝ω ₁×σ(Util _CPU)+ω ₂×σ(Util _MEM)+ω ₃×σ(Util _NET)+ω ₄×δ _VM(S)+ω ₅×cost(S)

Wherein, S represents the Topological Mapping relation of virtual machine and computing node, and T (S) represents load balancing desired value, σ (Util _cPU) represent the variance of the cpu busy percentage of each computing node, σ (Util _mEM) represent the variance of the memory usage of each computing node, σ (Util _nET) represent the variance of the network resource utilization of each link between each computing node, δ _vM(S) average weighted that represents all virtual machines that move in described group system postpones, and cost (S) represents group system mapping cost, ω ₁, ω ₂, ω ₃, ω ₄, ω ₅it is respectively weight coefficient.

The function that is appreciated that each functional module of the management node of the embodiment of the present invention can be according to the method specific implementation in the embodiment of the method shown in Fig. 2 a or 2b, and its specific implementation process can, with reference to the associated description in said method embodiment, repeat no more herein.

The management node of the embodiment of the present invention can be specifically computer equipment.

Therefore, in feasible execution modes more of the present invention, management node can be according to the network traffics relation between virtual machine, multiple virtual machines that need to carry out load balancing are divided at least one multiple virtual machine bunch, make, network traffics between at least one other virtual machine in each virtual machine and same virtual machine bunch are more than or equal to network traffics threshold value, and the network traffics between any other virtual machine in each virtual machine and different virtual machine bunch are less than network traffics threshold value; Cluster into migration suggestion for the each virtual machine in described multiple virtual machines bunch, described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host on same target computing node; Obtain following technique effect:

In the time carrying out load balancing, consider the impact of network dimension, by the network traffics relation between Real-Time Monitoring virtual machine, mark off the virtual machine bunch with network traffics relation, as the thread of follow-up generation virtual machine (vm) migration suggestion;

Multiple virtual machines in same virtual machine bunch can be focused on same computing node, thereby, flow between multiple virtual machines in virtual machine bunch can not take the load of the link between computing node, the offered load that can reduce each link in group system with this, can prevent link to occur performance bottleneck;

And, by the multiple virtual machines in same virtual machine bunch are focused on same computing node, the network bandwidth and communication reliability between the virtual machine in same virtual machine bunch can be improved, thereby the network performance of the virtual machine bunch of traffic-intensive can be promoted.

Please refer to Fig. 1, the embodiment of the present invention also provides a kind of group system, and this group system comprises multiple computing nodes 320 and management node 310, and management node 310 is the management nodes as shown in Fig. 3 a or Fig. 3 b embodiment.Wherein, computing node 320 comprises hardware layer 3201, operates in the host 3202 on hardware layer and operates at least one the virtual machine VM on host.The migration suggestion that described computing node 320 sends for receiving described management node 310, carries out virtual machine (vm) migration operation according to the instruction of described migration suggestion.This group system 300 can be specifically data center or virtual cluster.

This group system, in the time carrying out load balancing, has been considered the impact of network dimension, by the network traffics relation between Real-Time Monitoring virtual machine, marks off the virtual machine bunch with network traffics relation, as the thread of follow-up generation virtual machine (vm) migration suggestion;

Multiple virtual machines in same virtual machine bunch can be focused on same computing node, thereby, flow between multiple virtual machines in virtual machine bunch can not take the load of the link between computing node, the offered load that can reduce each link in group system with this, can prevent link to occur performance bottleneck;

And, by the multiple virtual machines in same virtual machine bunch are focused on same computing node, the network bandwidth and communication reliability between the virtual machine in same virtual machine bunch can be improved, thereby the network performance of the virtual machine bunch of traffic-intensive can be promoted.

The technical scheme providing for ease of better understanding the embodiment of the present invention, is example below by the execution mode under a concrete scene, and embodiment of the present invention technical scheme is introduced.

Please refer to Fig. 4 a, suppose that virtual cluster comprises four computing nodes, on four computing nodes, operation has host host respectively, and represents with host1, host2, host3, host4 respectively; Wherein, the computing node at host1, host2 place connects by switch switch2, and the computing node at host3, host4 place connects by switch switch3, and switch2 is connected with switch3 by switch1.Suppose to have higher flow between virtual machine App Server, DB Server, Mail Server, the Web Server of four different application, when these four virtual machines are distributed in again on different computing nodes simultaneously,, obviously, core link switch2 easily produces bottleneck to switch1 and switch1 to the offered load of switch3, and the network performance between virtual machine also can be affected simultaneously.

According to core concept of the present invention, if the virtual machine bunch with discharge relation is migrated on same computing node in the situation that computational resource allows, example as shown in Figure 4 b, virtual machine App Server, DB Server, these four of Mail Server, Web Server each other to high flow relation is considered as a virtual machine bunch, all move on Host1,, not only can reduce the load of core link, can improve the network performance between virtual machine simultaneously.

Please refer to Fig. 5, is the schematic diagram of the applied data center 500 of the embodiment of the present invention, and this data center 500 comprises management node 502 and multiple computing node 501.Wherein, several computing nodes 501 can form a cluster (Cluster) system 510 (as shown in 510a in figure or 510b), the part that can be considered as this group system 510 for managing the functional entity of this group system 510 of management node 502, in other words, can think that each group system also has the management node of oneself.In addition, said management node 502 can have two, and one as primary management node, and another can be used as management node for subsequent use.Embodiment of the present invention method can be used for whole data center (now whole data center being considered as to a group system), also can be for one of them group system.

Group system is a logical concept, is made up of, and provides some Premium Features as granularity, such as scheduling of resource, high available etc. multiple computing nodes.Computing node is physical host (being physical server), and unit virtualization is provided, and conventionally can move multiple virtual machines on it.Management node is managed whole data center, and the functions such as computing node and virtual machine life cycle management, resource scheduling management, O&M are provided, and is the brain of whole data center.Computing node and management node are all computer equipments.

As shown in Figure 6 a, be the schematic diagram of the load balancing apparatus 600 of management node 502 deploy, this load balancing apparatus 600 can comprise:

Performance data collection module 601, for the performance data of the whole data center of Real-time Collection, for example resource usage data, can comprise the such as CPU of computational resource of each computing node, the usage data of Memory, such as CPU of the computational resource of each virtual machine, the usage data of Memory, the Internet usage data of each virtual machine, the network traffics relation between virtual machine, the offered load of the link between every two computing nodes etc., and can save as historical data.This performance data collection module 601 is corresponding to the acquisition module 250 described in Fig. 3 b embodiment.

DRS control module 602, for other module is controlled and dispatched, carries out dynamic management to reach to the resource of data center or virtual cluster wherein.For example, obtain the performance data that performance data collection module 601 gathers, issue control command and corresponding performance data to other module, other module of instruction is processed accordingly.This DRS control module 602 is corresponding to determination module 210, computing module 260 and the judge module 270 described in Fig. 3 a or 3b embodiment.

Virtual machine flow compatibility identification module 603, for the performance data gathering according to performance data collection module 601, identify all virtual machines each other in group system with network traffics relation, also for carrying out as required the network traffics relation between multiple virtual machines of load balancing, multiple virtual machines that need to carry out load balancing are divided at least one virtual machine bunch, make, network traffics between at least one other virtual machine in each virtual machine and same virtual machine bunch are more than or equal to network traffics threshold value, network traffics between any other virtual machine in each virtual machine and different virtual machine bunch are less than network traffics threshold value.This virtual machine flow compatibility identification module 603 is corresponding to the sub-clustering module 220 described in Fig. 3 a or 3b embodiment.

Load balancing suggestion generation module 604, can specifically comprise network dimension migration suggestion generation module 604a and computational resource dimension suggestion generation module 604b, wherein: computational resource dimension suggestion generation module 604b, for need to carry out the virtual machine of load balancing from the screening of computational resource angle, generate candidate and move suggestion; Network dimension migration suggestion generation module 604a, for need to carry out the virtual machine of load balancing from the screening of Internet resources angle, generates candidate and moves suggestion; Detailed explanation please refer to the description in Fig. 1 embodiment.This load balancing suggestion generation module 604 can be corresponding to the suggestion module 230 described in Fig. 3 a or 3b embodiment.

Load balancing suggestion Executive Module 605, the candidate who generates is moved to suggestion and carry out Profit Assessment, for example calculate the balanced desired value of computational load whether minimum or whether be less than the threshold value of setting, if pass through Profit Assessment, carry out migration suggestion, that is, carry out virtual machine (vm) migration operation, on the target computing node that virtual machine (vm) migration to be migrated is specified to migration suggestion.This load balancing suggestion Executive Module 605, corresponding to the sending module 240 described in Fig. 3 a or 3b embodiment.

As shown in Figure 6 b, be the schematic diagram of the cluster resource management system of operation on the management node 502 of group system in the embodiment of the present invention.The performance data collection module 601 of above-mentioned load balancing apparatus 600 is corresponding to the input data process subsystem in Fig. 6 b herein, DRS control module 602 is corresponding to the control subsystem in Fig. 6 b, load balancing suggestion generation module 604 is corresponding to the algorithm subsystem in Fig. 6 b (load-balancing algorithm especially wherein), and load balancing suggestion Executive Module 605 is corresponding to the output subsystem in Fig. 6 b.Virtual machine flow compatibility identification module 603, is the module that the embodiment of the present invention newly joins cluster resource management system, also corresponding to the algorithm subsystem in Fig. 6 b.

Therefore, by disposing load balancing apparatus 600, management node 502 can be implemented resource scheduling management to group system, can be in the time carrying out load balancing, consider the impact of network dimension, by the network traffics relation between Real-Time Monitoring virtual machine, mark off the virtual machine bunch with network traffics relation, as the thread of follow-up generation virtual machine (vm) migration suggestion.

Please refer to Fig. 6 a and Fig. 7, is the flow process that the management node 502 that is deployed with load balancing apparatus 600 is carried out load balance scheduling, is example for system of virtual cluster, and load balance scheduling flow process can comprise:

701: the management node 502 of virtual cluster is enabled the DRS function of load balancing apparatus 600;

602 timings of 702:DRS control module or periodicity are obtained performance data from performance data collection module 601, for example, comprise performance data and parameter and the rule configuration etc. of cluster network topology information, CPU, Memory and network;

703:DRS control device 602 is carried out load-balancing algorithm, according to performance datas such as CPU, Memory and network topology and bandwidth, calculates the load balancing desired value of virtual cluster;

704:DRS control device 602 judges whether to carry out load balancing according to the load balancing desired value calculating, wherein, if the load balancing desired value of cluster is less than or equal to load balancing threshold value, illustrate that cluster is at equilibrium, finishing control flow process, if load balancing desired value is greater than load balancing threshold value, group system is in non-balanced state, need to carry out load balancing, enter next step;

705:DRS control device 602 is indicated load balancing suggestion generation module 604 and is utilized virtual machine flow compatibility identification module 603, filter out the virtual machine that need to carry out load balancing from computational resource dimension and Internet resources dimension respectively, form virtual machine list to be migrated;

706:DRS control device 602 utilizes virtual machine flow compatibility identification module 603, carries out virtual machine flow compatibility recognizer, goes out multiple virtual machines bunch to be migrated from virtual machine list cluster to be migrated;

707:DRS control device 602 is indicated load balancing suggestion generation module 604, filter out computational resource utilization rate lower than at least one computing node of computational resource load threshold as candidate's computing node, from candidate's computing node, select a target computing node for each virtual machine bunch, make whole virtual machine (vm) migrations in this virtual machine bunch to or the load balancing desired value minimum of host's group system after selected target computing node;

708:DRS control device 602 is indicated load balancing suggestion generation module 604, clusters into migration suggestion for each virtual machine, migration suggestion instruction by whole virtual machine (vm) migrations of each virtual machine bunch to or host at same target computing node;

709:DRS control device 602 is indicated load balancing suggestion Executive Module 605, virtual machine bunch migration suggestion is sent to the source computing node of each virtual machine host in virtual machine bunch, realize whole virtual machine (vm) migrations of each virtual machine bunch to same target computing node;

Then, enter next iteration control.In the embodiment of the present invention, management node 502 periodically or regularly carries out load balancing, and each iteration control cycle carries out taking turns load balancing.

By above-mentioned flow process, management node can focus on the multiple virtual machines that have in the same virtual machine bunch of network traffics relation on same computing node, thereby, network traffics between multiple virtual machines in same virtual machine bunch can not take the load of the link between computing node, the offered load that can reduce each link in group system with this, there is performance bottleneck in prevention link; And, by the multiple virtual machines in same virtual machine bunch are focused on same computing node, the network bandwidth and communication reliability between the virtual machine in same virtual machine bunch can be improved, thereby the network performance of the virtual machine bunch of traffic-intensive can be promoted.

Please refer to Fig. 8, in some embodiments of the invention, in step 703, DRS control device 602 is carried out load-balancing algorithm, judges whether to carry out the flow process of load balancing and generating virtual machines migration suggestion, specifically can comprise:

801: problem of load balancing is converted into optimization problem, sets up unified optimization aim function for the balanced desired value of computational load, optimization aim function is as follows:

T(S)＝ω ₁×σ(Util _CPU)+ω ₂×σ(Util _MEM)+ω ₃×σ(Util _NET)+ω ₄×δ _VM(S)+ω ₅×cost(S)

Wherein, S represents the Topological Mapping relation of virtual machine and computing node, and T (S) represents load balancing desired value, σ (Util _cPU) represent the variance of the cpu busy percentage of each computing node, σ (Util _mEM) represent the variance of the memory usage of each computing node, σ (Util _nET) represent the variance of the network resource utilization of each link between each computing node, δ _vM(S) weighting of expression virtual machine postpones, and cost (S) represents group system mapping cost, ω ₁, ω ₂, ω ₃, ω ₄, ω ₅it is respectively weight coefficient.

802: gather properties data by timer or manual triggers, utilize the balanced desired value of above-mentioned optimization aim function computational load;

803: judge when load balancing desired value is less than load balancing threshold value, exit load balance scheduling, otherwise, be judged as and need to carry out load balancing, continue to carry out following steps;

804: for example, select to carry out the virtual machine of load balancing from computational resource (CPU, Memory) dimension, add in virtual machine list to be migrated;

805: select to carry out the virtual machine of load balancing from Internet resources dimension, add in virtual machine list to be migrated;

806: travel through virtual machine list to be migrated, utilize virtual machine flow compatibility recognizer, divide one or more virtual machines bunch, according to optimization aim function, to virtual machine bunch select target computing node;

807: select virtual machine bunch and the target computing node of the balanced desired value minimum of migration back loading, generate pending virtual machine (vm) migration suggestion.

By above-mentioned flow process, management node can focus on the multiple virtual machines that have in the same virtual machine bunch of network traffics relation on same computing node, thereby, network traffics between multiple virtual machines in same virtual machine bunch can not take the load of the link between computing node, the offered load that can reduce each link in group system with this, there is performance bottleneck in prevention link; And, by the multiple virtual machines in same virtual machine bunch are focused on same computing node, the network bandwidth and communication reliability between the virtual machine in same virtual machine bunch can be improved, thereby the network performance of the virtual machine bunch of traffic-intensive can be promoted.

Please refer to Fig. 9, in some embodiments of the invention, in step 806, travel through virtual machine list to be migrated, utilize virtual machine flow compatibility recognizer, divide the flow process of one or more virtual machines bunch, specifically can comprise:

S1: according to the network traffics relation between virtual machine, construct the communication matrix of all virtual machines in virtual machine list to be migrated;

S2: according to the communication matrix of virtual machine, virtual machine constructor communication connected graph, for example, shown in Figure 10 (a), in figure, each circle represents a virtual machine VM, have between two virtual machines of network traffics relation with segment link, the numeral on line segment is for characterizing the size of flow;

S3: virtual machine flow compatibility identification problem is converted into weighted graph segmentation problem;

S4: set segmentation threshold, weights are carried out to cutting at the line segment of interval [0, h], successively virtual machine communication connected graph is cut apart; For example, between setting district [0,5], flow value is positioned to interval [0,5] line clipping between is removed, only retain the line segment that flow value is greater than 5, as shown in Figure 10 (b), obtain four virtual machines bunch (Cluster), wherein Cluster1 comprises virtual machine VM1, VM2, VM3, VM5, VM6, Cluster2 comprises virtual machine VM4, and Cluster3 comprises virtual machine VM9, VM10, and Cluster4 comprises virtual machine VM7, VM8, VM11, VM12.

S5: according to the attaching relation of virtual machine and computing node, the communication connected graph of virtual machine is carried out to secondary splitting, as shown in Figure 10 (c), the corresponding relation list of generating virtual machines bunch and computing node.

Above-mentioned flow process has been introduced the method for a kind of partition virtual machines bunch, and the method, by building virtual machine communication connected graph, is cut apart virtual machine communication connected graph, can realize the multiple virtual machines that need equally loaded are divided into multiple virtual machines bunch.

Follow-up, management node just can the corresponding relation list with computing node according to the virtual machine bunch of dividing and virtual machine bunch, and generating virtual machines moves and advised.For example, for virtual machine bunch Cluster1, virtual machine VM1, VM2, the VM3 that it comprises operates on computing node Host1, virtual machine VM5, VM6 operate on computing node Host3,, can generate: the candidate that virtual machine VM1, VM2, VM3 are moved on computing node Host3 moves suggestion, or, the candidate that virtual machine VM5, VM6 are moved on computing node Host1 moves suggestion, or, VM1, VM2, VM3, VM4, VM5 are moved to for example candidate on Host2 of another computing node and move suggestion.

To sum up, be example by the execution mode under a concrete scene, embodiment of the present invention technical scheme is introduced, do not know that part can be with reference to the description of other parts above.

Therefore, in feasible execution modes more of the present invention, in the time carrying out load balancing, consider the impact of network dimension, by the multiple virtual machines that have in the same virtual machine bunch of network traffics relation are focused on same computing node, thereby the network traffics between the multiple virtual machines in same virtual machine bunch can not take the load of the link between computing node, the offered load that can reduce each link in group system with this, there is performance bottleneck in prevention link; And, by the multiple virtual machines in same virtual machine bunch are focused on same computing node, the network bandwidth and communication reliability between the virtual machine in same virtual machine bunch can be improved, thereby the network performance of the virtual machine bunch of traffic-intensive can be promoted.

The embodiment of the present invention also provides a kind of computer-readable storage medium, and this computer-readable storage medium can have program stored therein, and this program comprises the part or all of step of the load-balancing method of the virtual machine of recording in above-mentioned embodiment of the method as shown in Fig. 2 a or 2b while execution.

Please refer to Figure 11, the embodiment of the present invention also provides a kind of management node 900.

The management node 900 of the embodiment of the present invention can be applicable to comprise the group system of this management node 900 and multiple computing nodes, and as the management node in described group system, should be understood that, the computing node in this group system can comprise processor and memory; In some embodiments, the memory stores that the computing node in this group system comprises following element, executable module or data structure, or their subset, or their superset:

Host: as management level, in order to complete management, the distribution of hardware resource; For virtual machine presents virtual hardware platform; Realize scheduling and the isolation of virtual machine.Wherein, Host may be monitor of virtual machine (VMM); In addition, VMM and 1 franchise virtual machine coordinate sometimes, and both are in conjunction with composition Host.Wherein, virtual hardware platform provides various hardware resources to each virtual machine of operation on it, as virtual processor, internal memory, virtual disk, Microsoft Loopback Adapter etc. are provided.Wherein, this virtual disk can corresponding Host a file or a logical block equipment.Virtual machine operates on the virtual hardware platform that Host is its preparation, the one or more virtual machines of the upper operation of Host.

One or more virtual machines: can simulate one or many virtual computers by software virtual machine on a physical computer, and these virtual machines carry out work just as real computer, can installing operating system and application program on virtual machine, virtual machine is addressable network resource also.For the application program of moving in virtual machine, virtual machine similarly is in real computer, to carry out work.

The management node 900 of the embodiment of the present invention can comprise input equipment 901 (optionally), output equipment 904 (optionally), processor 902 and memory 903.Wherein, memory 903 can comprise read-only memory and random access memory, and provides instruction and data to processor 902.A part for memory 903 can also comprise nonvolatile RAM (NVRAM).

Memory 903 has been stored following element, executable module or data structure, or their subset, or their superset:

Operational order: comprise various operational orders, for realizing various operations.

Operating system: comprise various system programs, for realizing various basic businesses and processing hardware based task.

In embodiments of the present invention, the operational order (this operational order can be stored in operating system) that processor 902 is stored by calling memory 903, carry out following operation: determine the multiple virtual machines that need to carry out load balancing in described group system, described multiple virtual machines distribute and operate on the part or all of computing node of the described multiple computing nodes in described group system; According to the network traffics relation between described multiple virtual machines, described multiple virtual machines are divided into at least one virtual machine bunch, make, network traffics between at least one other virtual machine in each virtual machine and same virtual machine bunch are more than or equal to network traffics threshold value, and the network traffics between any other virtual machine in each virtual machine and different virtual machine bunch are less than network traffics threshold value; Cluster into migration suggestion for each virtual machine, described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host on same target computing node; And the one or more virtual machine hosts' that comprise to described virtual machine bunch source computing node sends described migration suggestion so that whole virtual machines that described virtual machine bunch comprises are migrated to or host on described target computing node.

In the embodiment of the present invention, in the time carrying out load balancing, consider the impact of network dimension, by the multiple virtual machines that have in the same virtual machine bunch of network traffics relation are focused on same computing node, thereby, network traffics between multiple virtual machines in same virtual machine bunch can not take the load of the link between computing node, the offered load that can reduce each link in group system with this, and there is performance bottleneck in prevention link; And, by the multiple virtual machines in same virtual machine bunch are focused on same computing node, the network bandwidth and communication reliability between the virtual machine in same virtual machine bunch can be improved, thereby the network performance of the virtual machine bunch of traffic-intensive can be promoted.

Processor 902 is controlled the operation of computer equipment 900, and processor 902 can also be called CPU (Central Processing Unit, CPU).Memory 903 can comprise read-only memory and random access memory, and provides instruction and data to processor 902.A part for memory 903 can also comprise nonvolatile RAM (NVRAM).In concrete application, each assembly of gateway 600 is coupled by bus system 905, and wherein bus system 905, except comprising data/address bus, can also comprise power bus, control bus and status signal bus in addition etc.But for the purpose of clearly demonstrating, in the drawings various buses are all designated as to bus system 905.

The method that the invention described above embodiment discloses can be applied in processor 902, or is realized by processor 902.Processor 902 may be a kind of integrated circuit (IC) chip, has the disposal ability of signal.In implementation procedure, each step of said method can complete by the instruction of the integrated logic circuit of the hardware in processor 902 or software form.Above-mentioned processor 902 can be general processor, digital signal processor (DSP), application-specific integrated circuit (ASIC) (ASIC), ready-made programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic device, discrete hardware components.Can realize or carry out disclosed each method, step and logic diagram in the embodiment of the present invention.General processor can be that microprocessor or this processor can be also the processors of any routine etc.Can directly be presented as that in conjunction with the step of the disclosed method of the embodiment of the present invention hardware decoding processor is complete, or complete with the hardware in decoding processor and software module combination.Software module can be positioned at random asccess memory, and flash memory, read-only memory, in the storage medium of this area maturations such as programmable read only memory or electrically erasable programmable memory, register.This storage medium is positioned at memory 903, and the information in processor 902 read memories 903 completes the step of said method in conjunction with its hardware.

Alternatively, according to the network traffics relation between described multiple virtual machines, described multiple virtual machines are divided into the aspect of at least one virtual machine bunch, described processor 902 specifically for: according to the network traffics relation between described multiple virtual machines, build described multiple virtual machines network service be communicated with topology; Described network service is communicated with to topology to be cut apart, each the group virtual machine that network traffics is each other more than or equal to network traffics threshold value is defined as a virtual machine bunch, and, each isolated virtual machine is defined as separately to a virtual machine bunch, and the network traffics of described isolated virtual machine and other any virtual machine are all less than network traffics threshold value.

Alternatively, determine need in described group system to carry out multiple virtual machines of load balancing aspect, described processor 902 is specifically in the whole virtual machines that move from described group system, filter out the virtual machine each other with network traffics relation, the virtual machine each other described in determining with network traffics relation is the multiple virtual machines that need to carry out load balancing in described group system.

Alternatively, before needing to carry out multiple virtual machines of load balancing in the definite described group system of described processor 902, also for the resource usage data that obtains described group system, described resource usage data comprises cpu busy percentage and the memory usage of each computing node in described group system, and the network resource utilization of each link between each computing node; According to the balanced desired value of described resource usage data computational load, described load balancing desired value is for representing the load balancing state of described group system; If described load balancing desired value is greater than load balancing threshold value, judge that described group system need to carry out load balancing.

Alternatively, aspect the balanced desired value of described resource usage data computational load, described processor 902 specifically for:

Adopt the balanced desired value of following formula computational load;

T(S)＝ω ₁×σ(Util _CPU)+ω ₂×σ(Util _MEM)+ω ₃×σ(Util _NET)+ω ₄×δ _VM(S)

Wherein, S represents the Topological Mapping relation of virtual machine and computing node, and T (S) represents load balancing desired value, σ (Util _cPU) represent the variance of the cpu busy percentage of each computing node, σ (Util _mEM) represent the variance of the memory usage of each computing node, σ (Util _nET) represent the variance of the network resource utilization of each link between each computing node, δ _vM(S) represent all virtual machines that move in described group system average weighted postpone, ω ₁, ω ₂, ω ₃, ω ₄it is respectively weight coefficient.

Alternatively, cluster into migration suggestion for each virtual machine, described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host on same target computing node, described processor 902 specifically for: select multiple candidate's computing nodes for each virtual machine bunch, analog computation each virtual machine bunch is migrated to or the load balancing desired value of host's described group system after candidate's computing node; For each virtual machine bunch, from described multiple candidate's computing nodes, determine a target computing node, make whole virtual machine (vm) migrations in each virtual machine bunch to or the load balancing desired value minimum of host's described group system after selected target computing node; Cluster into migration suggestion for each virtual machine, described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host on the target computing node of described selection.

Alternatively, analog computation each virtual machine bunch be migrated to or the load balancing desired value of host's described group system after candidate's computing node aspect, described processor 902 specifically for:

Adopt the balanced desired value of following formula computational load;

T(S)＝ω ₁×σ(Util _CPU)+ω ₂×σ(Util _MEM)+ω ₃×σ(Util _NET)+ω ₄×δ _VM(S)+ω ₅×cost(S)

Wherein, S represents the Topological Mapping relation of virtual machine and computing node, and T (S) represents load balancing desired value, σ (Util _cPU) represent the variance of the cpu busy percentage of each computing node, σ (Util _mEM) represent the variance of the memory usage of each computing node, σ (Util _nET) represent the variance of the network resource utilization of each link between each computing node, δ _vM(S) average weighted that represents all virtual machines that move in described group system postpones, and cost (S) represents group system mapping cost, ω ₁, ω ₂, ω ₃, ω ₄, ω ₅it is respectively weight coefficient.

Therefore, in some execution modes of the present invention, management node, be applied to the group system that comprises this management node and multiple computing nodes, can be according to the network traffics relation between virtual machine in group system, multiple virtual machines that need to carry out load balancing are divided at least one virtual machine bunch, make, network traffics between at least one other virtual machine in each virtual machine and same virtual machine bunch are more than or equal to network traffics threshold value, network traffics between any other virtual machine in each virtual machine and different virtual machine bunch are less than network traffics threshold value, cluster into migration suggestion for the each virtual machine in described multiple virtual machines bunch, described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host on same target computing node, obtain following technique effect:

In the time carrying out load balancing, consider the impact of network dimension, by the multiple virtual machines that have in the same virtual machine bunch of network traffics relation are focused on same computing node, thereby, network traffics between multiple virtual machines in same virtual machine bunch can not take the load of the link between computing node, the offered load that can reduce each link in group system with this, there is performance bottleneck in prevention link; And, by the multiple virtual machines in same virtual machine bunch are focused on same computing node, the network bandwidth and communication reliability between the virtual machine in same virtual machine bunch can be improved, thereby the network performance of the virtual machine bunch of traffic-intensive can be promoted.

In the above-described embodiments, the description of each embodiment is all emphasized particularly on different fields, in certain embodiment, there is no the part of detailed description, can be referring to the associated description of other embodiment.

It should be noted that, for aforesaid each embodiment of the method, for simple description, therefore it is all expressed as to a series of combination of actions, but those skilled in the art should know, the present invention is not subject to the restriction of described sequence of movement, because according to the present invention, some step can adopt other order or carry out simultaneously.Secondly, those skilled in the art also should know, the embodiment described in specification all belongs to preferred embodiment, and related action and module might not be that the present invention is necessary.

One of ordinary skill in the art will appreciate that all or part of step in the whole bag of tricks of above-described embodiment is can carry out the hardware that instruction is relevant (for example processor) by program to complete, this program can be stored in a computer-readable recording medium, and storage medium can comprise: ROM, RAM, disk or CD etc.

The load-balancing method of a kind of the virtual machine above embodiment of the present invention being provided and relevant device and group system are described in detail, applied specific case herein principle of the present invention and execution mode are set forth, the explanation of above embodiment is just for helping to understand method of the present invention and core concept thereof; , for one of ordinary skill in the art, according to thought of the present invention, all will change in specific embodiments and applications, in sum, this description should not be construed as limitation of the present invention meanwhile.

Claims (15)

1. the load-balancing method of a virtual machine, it is characterized in that, described method is for comprising the group system of management node and multiple computing nodes, and the each computing node in described multiple computing nodes comprises hardware layer, operates in the host on described hardware layer and operates in the virtual machine on described host;

Described method comprises:

Described management node is determined the multiple virtual machines that need to carry out load balancing in described group system, and described multiple virtual machines distribute and operate on the part or all of computing node of the described multiple computing nodes in described group system;

According to the network traffics relation between described multiple virtual machines, described multiple virtual machines are divided into at least one virtual machine bunch, make, network traffics between at least one other virtual machine in each virtual machine and same virtual machine bunch are more than or equal to network traffics threshold value, and the network traffics between any other virtual machine in each virtual machine and different virtual machine bunch are less than network traffics threshold value;

Cluster into migration suggestion for each virtual machine, described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host on same target computing node;

The one or more virtual machine hosts' that comprise to described virtual machine bunch source computing node sends described migration suggestion so that whole virtual machines that described virtual machine bunch comprises are migrated to or host on described target computing node.

2. method according to claim 1, is characterized in that, described according to the network traffics relation between described multiple virtual machines, described multiple virtual machines is divided into at least one virtual machine bunch and comprises:

According to the network traffics relation between described multiple virtual machines, the network service that builds described multiple virtual machines is communicated with topology;

Described network service is communicated with to topology to be cut apart, each the group virtual machine that network traffics is each other more than or equal to network traffics threshold value is defined as a virtual machine bunch, and, each isolated virtual machine is defined as separately to a virtual machine bunch, and the network traffics of described isolated virtual machine and other any virtual machine are all less than network traffics threshold value.

3. method according to claim 1, is characterized in that, described management node is determined needs the multiple virtual machines that carry out load balancing to comprise in described group system:

In the whole virtual machines that move from described group system, filter out the virtual machine each other with network traffics relation, the virtual machine each other described in determining with network traffics relation is the multiple virtual machines that need to carry out load balancing in described group system.

4. according to arbitrary described method in claims 1 to 3, it is characterized in that, described management node also comprises before determining the multiple virtual machines that need to carry out load balancing in described group system:

Described management node obtains the resource usage data of described group system, described resource usage data comprises cpu busy percentage and the memory usage of each computing node in described group system, and the network resource utilization of each link between each computing node;

According to the balanced desired value of described resource usage data computational load, described load balancing desired value is for representing the load balancing state of described group system;

If described load balancing desired value is greater than load balancing threshold value, judge that described group system need to carry out load balancing.

5. method according to claim 4, is characterized in that, describedly comprises according to the balanced desired value of described resource usage data computational load:

Adopt the balanced desired value of following formula computational load;

T(S)＝ω ₁×σ(Util _CPU)+ω ₂×σ(Util _MEM)+ω ₃×σ(Util _NET)+ω ₄×δ _VM(S)

Wherein, S represents the Topological Mapping relation of virtual machine and computing node, and T (S) represents load balancing desired value, σ (Util _cPU) represent the variance of the cpu busy percentage of each computing node, σ (Util _mEM) represent the variance of the memory usage of each computing node, σ (Util _nET) represent the variance of the network resource utilization of each link between each computing node, δ _vM(S) represent all virtual machines that move in described group system average weighted postpone, ω ₁, ω ₂, ω ₃, ω ₄it is respectively weight coefficient.

6. according to arbitrary described method in claims 1 to 3, it is characterized in that, describedly cluster into migration suggestion for each virtual machine, described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, moves to or host comprises on same target computing node:

Select multiple candidate's computing nodes for each virtual machine bunch, analog computation each virtual machine bunch be migrated to or the candidate computing node of host in described multiple candidate's computing nodes after the load balancing desired value of described group system;

For each virtual machine bunch, from described multiple candidate's computing nodes, determine a target computing node, make whole virtual machine (vm) migrations in each virtual machine bunch to or the load balancing desired value minimum of host's described group system after selected target computing node;

Cluster into migration suggestion for each virtual machine, described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host on the target computing node of described selection.

7. method according to claim 6, is characterized in that, described analog computation each virtual machine bunch be migrated to or the candidate computing node of host in described multiple candidate's computing nodes after the load balancing desired value of described group system comprise:

Adopt the balanced desired value of following formula computational load;

T(S)＝ω ₁×σ(Util _CPU)+ω ₂×σ(Util _MEM)+ω ₃×σ(Util _NET)+ω ₄×δ _VM(S)+ω ₅×cost(S)

Wherein, S represents the Topological Mapping relation of virtual machine and computing node, and T (S) represents load balancing desired value, σ (Util _cPU) represent the variance of the cpu busy percentage of each computing node, σ (Util _mEM) represent the variance of the memory usage of each computing node, σ (Util _nET) represent the variance of the network resource utilization of each link between each computing node, δ _vM(S) average weighted that represents all virtual machines that move in described group system postpones, and cost (S) represents group system mapping cost, ω ₁, ω ₂, ω ₃, ω ₄, ω ₅it is respectively weight coefficient.

8. a management node, it is characterized in that, for comprising the group system of multiple computing nodes and described management node, the each computing node in described multiple computing nodes comprises hardware layer, operates in the host on described hardware layer and operates in the virtual machine on described host; Described management node comprises:

Determination module, for determining that described group system needs to carry out multiple virtual machines of load balancing, described multiple virtual machines distribute and operate on the part or all of computing node of the described multiple computing nodes in described group system;

Sub-clustering module, be used for according to the network traffics relation between described multiple virtual machines, described multiple virtual machines are divided into at least one virtual machine bunch, make, network traffics between at least one other virtual machine in each virtual machine and same virtual machine bunch are more than or equal to network traffics threshold value, and the network traffics between any other virtual machine in each virtual machine and different virtual machine bunch are less than network traffics threshold value;

Suggestion module, for cluster into migration suggestion for each virtual machine, described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host on same target computing node;

Sending module, sends described migration suggestion for one or more virtual machine hosts' of comprising to described virtual machine bunch source computing node so that whole virtual machines that described virtual machine bunch comprises are migrated to or host on described target computing node.

9. management node according to claim 8, is characterized in that, described sub-clustering module comprises:

Construction unit, for according to the network traffics relation between described multiple virtual machines, builds the network service connection topology of described multiple virtual machines;

Division unit, for described network service connected graph is cut apart, each the group virtual machine that network traffics is each other more than or equal to network traffics threshold value is divided into a virtual machine bunch, and, each isolated virtual machine is defined as separately to a virtual machine bunch, and the network traffics of described isolated virtual machine and other any virtual machine are all less than network traffics threshold value.

10. management node according to claim 8, it is characterized in that, described determination module specifically for: in the whole virtual machines that move from described group system, filter out the virtual machine each other with network traffics relation, the virtual machine each other described in determining with network traffics relation is the multiple plan machines of void that need to carry out load balancing in described group system.

In 11. according to Claim 8 to 10, arbitrary described management node, is characterized in that, also comprises:

Acquisition module, for obtaining the resource usage data of described group system, described resource usage data comprises cpu busy percentage and the memory usage of each computing node in described group system, and the network resource utilization of each link between each computing node;

Computing module, for according to the balanced desired value of described resource usage data computational load, described load balancing desired value is for representing the load balancing state of described group system;

Judge module, if be greater than load balancing threshold value for described load balancing desired value, judges that described group system need to carry out load balancing.

12. management nodes according to claim 11, is characterized in that, described computing module is specifically for adopting the balanced desired value of following formula computational load;

T(S)＝ω ₁×σ(Util _CPU)+ω ₂×σ(Util _MEM)+ω ₃×σ(Util _NET)+ω ₄×δ _VM(S)

Wherein, S represents the Topological Mapping relation of virtual machine and computing node, and T (S) represents load balancing desired value, σ (Util _cPU) represent the variance of the cpu busy percentage of each computing node, σ (Util _mEM) represent the variance of the memory usage of each computing node, σ (Util _nET) represent the variance of the network resource utilization of each link between each computing node, δ _vM(S) represent all virtual machines that move in described group system average weighted postpone, ω ₁, ω ₂, ω ₃, ω ₄it is respectively weight coefficient.

In 13. according to Claim 8 to 10, arbitrary described management node, is characterized in that, described suggestion module comprises:

Analogue computer, for selecting multiple candidate's computing nodes for each virtual machine bunch, analog computation each virtual machine bunch be migrated to or the candidate computing node of host in described multiple candidate's computing nodes after the load balancing desired value of described group system;

Determining unit, be used for for each virtual machine bunch, from described multiple candidate's computing nodes, determine a target computing node, make whole virtual machine (vm) migrations in each virtual machine bunch to or the load balancing desired value minimum of host's described group system after selected target computing node;

Suggestion generation unit, for cluster into migration suggestion for each virtual machine, described migration is proposed to be used in whole virtual machines that instruction comprises each virtual machine bunch, move to or host at the target computing node of described selection.

14. methods according to claim 13, is characterized in that, described analogue computer specifically for: adopt the balanced desired value of following formula computational load;

T(S)＝ω ₁×σ(Util _CPU)+ω ₂×σ(Util _MEM)+ω ₃×σ(Util _NET)+ω ₄×δ _VM(S)+ω ₅×cost(S)

Wherein, S represents the Topological Mapping relation of virtual machine and computing node, and T (S) represents load balancing desired value, σ (Util _cPU) represent the variance of the cpu busy percentage of each computing node, σ (Util _mEM) represent the variance of the memory usage of each computing node, σ (Util _nET) represent the variance of the network resource utilization of each link between each computing node, δ _vM(S) average weighted that represents all virtual machines that move in described group system postpones, cost (S) representative system mapping cost, ω ₁, ω ₂, ω ₃, ω ₄, ω ₅it is respectively weight coefficient.

15. 1 kinds of group systems, is characterized in that, comprise multiple computing nodes and as the management node as described in arbitrary in claim 8-14;

The migration suggestion that described computing node sends for receiving described management node, carries out virtual machine (vm) migration operation according to the instruction of described migration suggestion.