CN102567080B - Virtual machine position selection system facing load balance in cloud computation environment - Google Patents

Virtual machine position selection system facing load balance in cloud computation environment Download PDF

Info

Publication number
CN102567080B
CN102567080B CN201210001315.9A CN201210001315A CN102567080B CN 102567080 B CN102567080 B CN 102567080B CN 201210001315 A CN201210001315 A CN 201210001315A CN 102567080 B CN102567080 B CN 102567080B
Authority
CN
China
Prior art keywords
module
node
load
virtual machine
position selection
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201210001315.9A
Other languages
Chinese (zh)
Other versions
CN102567080A (en
Inventor
阮利
肖利民
祝明发
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
SHANGHAI ZHIRUI ELECTRONIC TECHNOLOGY Co.,Ltd.
Original Assignee
Beihang University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beihang University filed Critical Beihang University
Priority to CN201210001315.9A priority Critical patent/CN102567080B/en
Publication of CN102567080A publication Critical patent/CN102567080A/en
Application granted granted Critical
Publication of CN102567080B publication Critical patent/CN102567080B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Landscapes

  • Computer And Data Communications (AREA)

Abstract

The invention discloses a virtual machine position selection system facing load balance in a cloud computation environment, which comprises a front end module and a rear end module, wherein the front end module comprises a comprehensive load computation module and a host selection module, and the rear end module comprises a plurality of distributed node agent modules in an integrated manner. Each node agent module comprises a node load collection module, a node load predication module and a node load incremental computation module. Cloud computation serving as a novel computation and service method is put forward and is increasingly widely paid attention to, popularized and applied, and the virtual machine position selection system serves as a vital module in resource allocation of cloud computation, so that the virtual machine position selection system has the advantages that the virtual machine network throughput rate in the cloud computation environment is increased, the resource preemption probability of virtual machines is decreased, the service quality of a virtual host is improved, and the load balance of a distributed system in the cloud computation environment is enhanced. Hence, the virtual machine position selection system is wide in application range and capable of generating remarkable economic benefit.

Description

The virtual machine position selection system of the facing load balance in a kind of cloud computing environment
Technical field
The invention discloses a kind of virtual machine position selection system, particularly relate to the virtual machine position selection system of the facing load balance in a kind of cloud computing environment.Belong to field of computer technology.
Background technology
In recent years, the develop rapidly of network application makes constantly to increase the demand of computing power, and along with the development of grid computing, parallel computation, Distributed Calculation, cloud computing is arisen at the historic moment.According to the definition of National Institute of Standards and Technology (NIST), current cloud computing service can be divided into 3 levels, respectively: namely (1) infrastructure serve (IaaS), as the elastic calculation cloud (elastic compute cloud, be called for short EC2) of Amazon, the blue cloud (blue cloud) of IBM and the cloud Infrastructure platform (IAAS) etc. of Sun; (2) namely platform serves (PaaS), as the Google App Engine of Google and the Azure platform etc. of Microsoft; (3) namely software serve (SaaS), as the CRM service etc. of Salesforce company.Cloud computing, as a kind of emerging business computation schema, is listed in the following technique direction given priority to of every country, and becomes the hot research problem of computer nowadays research circle and industry member
Along with the application kind in cloud computing environment get more and more, resource extent increasing, in cloud computing environment, the difficulty of resource management is also in remarkable increase.Especially along with Intel Virtualization Technology is in recent years revived again, virtual resource is as the arena of history in a kind of keystone resources form again secondary, and the virtual data center that facing cloud calculates obtains to be paid close attention to more widely.So-called virtual data center, foreign scholar is called Virtual Datacenter, refer to and utilize server virtualization technology, adopt fictitious host computer (abbreviation virtual machine) that is independent, isolation mutually to provide the function of equivalent physical host, and its cost will be lower than typical data center.So-called virtual machine, is exactly computer software, runs on physical hardware or physical computer, and it can operation system (being called client operating system) and application program, has the virtual hardware of oneself.Virtual machine is not emulator and simulator, and they are real computing machines, can realize the function that even exceed physical computer identical with physical computer.Easy-to-use flexibly in view of virtual machine, the function of physical machine will transfer the virtual machine holding and provide these to serve to, the namely host (host) of relative virtual machine from providing service (application program, database).Along with the development of cloud computing, the large-scale IT company such as current IBM, HP, Amazon, Google, Microsoft and data center are all actively setting up the virtual data center also externally providing various facing cloud calculation services.Visible, virtual data center technology has become the hot research problem of the outer research circle of Present Domestic and industry member, and the management of resource virtualizing and virtual resource has also become one of important channel solving system resource utilization bottleneck in cloud computing, there is important Research Significance and using value.
Be that the utilization of resources in cloud computing environment brings convenience at Intel Virtualization Technology, namely physical machine is combined externally by while the independent virtual main frame that provides considerably beyond physical host quantity with Intel Virtualization Technology, also a difficult problem for virtual machine position selection is brought, namely how rightly for virtual machine selects host.For this problem, the virtual machine position selection method in cloud computing environment and the progress of product and case study as follows:
Some Virtual Machine Manager products of knowing clearly are researched and developed both at home and abroad, as the OpenNebula increased income, the systems such as Eucalyptus, the Virtual Machine Manager 2008 of industrial community and VMware ESX Server, virtual machine position selection embeds wherein as subfunction, and visible virtual machine position selection is a kernel subsystems.From architecture aspect, an existing system of selecting can be divided into 1) centralized, as virtual machine quantity on physical host at most or load the maximum preferential; 2) distributing, as virtual machine minimum number on physical host or least-loaded person preferential.From implementation, an existing system of selecting can be divided into two classes, and one is manually select position, by keeper perceptual be virtual machine select host, Existence dependency in keeper's subjective factor, the shortcomings such as robotization is high not; Two is automatically select position, selects host by background process based on to the virtual machine that is thought of as of physical host environmental information, resource utilization etc.An implementation algorithm is selected from core, the robotization virtual machine position selection core algorithm of current main flow mainly contains packing (Packing) method, itemize (Striping) method, Load-aware (Load-aware) method, method that internal memory is got close to (Memory Buddies).
● the basic thought of packing (Packing) method is to use node for target to the greatest extent less, to be concentrated on by virtual machine on the node in part cloud computing and run.In implementation, adopt virtual machine to run the maximum priority principle of number, namely when host selected by needs for newly-built virtual machine, select the host having the operation of maximum quantity virtual machine.Achieve this virtual machine position selection method in the Scheduler program of OpenNebula platform, its configuration file is labeled as " RANK=RUNNING_VMS ".The advantage of packaging method is that the method can make a large amount of virtual machine concentrate in minority physical nodes to run, can reduce physical server cost.The deficiency of packaging method is that virtual machine is too concentrated, causes resources of virtual machine to seize probability excessive, and for ensureing the service quality of virtual server, it is essential that a large amount of virtual machine (vm) migrations and resource adjust behavior, can produce a large amount of expense like this.
● itemize (Striping) method basic thought is that to maximize individual server node available resources be target, is dispersed in by virtual machine on all nodes and runs.In implementation, adopt virtual machine to run minimum number priority principle, namely when host selected by needs for newly-built virtual machine, select the host having the operation of minimum number virtual machine.Also achieve this virtual machine position selection method in the Scheduler program of OpenNebula platform, its configuration file is labeled as " RANK=-RUNNING_VMS ".Its basic ideas inspire by group of planes load balancing, and virtual machine is distributed by uniform amount, reduces resources of virtual machine and seize probability.Its shortcoming carries out selecting position according to virtual machine quantity, only abstract for virtual machine quantity be load value on node, too single, and different resource cannot be distinguished (as CPU, internal memory, disk etc.) actual loading that the virtual machine of asking causes node, inadequate refinement, is difficult to realize more fine granularity and accurate resource allocation requirements.
● the target of Load-aware (Load-aware) method is identical with itemize (Striping) method, makes every effort to maximize the available resources on single node.Basic Design thinking inspires by node least-loaded, is placed on by newly-built virtual machine on the node with minimum load and runs.In implementation, adopt maximum CPU idleness priority principle, namely when host selected by needs for newly-built virtual machine, select the host that CPU idleness is maximum.Also achieve this method in the Scheduler program of OpenNebula platform, its configuration file is labeled as " RANK=FREECPU ".Its advantage comprises identical with itemize method, inspires by load balancing, owing to considering the cpu resource service condition on node, can reach the cpu resource load balancing within the scope of distributed computing system.Its shortcoming comprises the method and only considered cpu resource, to the important composition of the Cluster nodes loads such as the internal memory of node, network and disk service condition, lacks and considers.
● method that internal memory is got close to (Memory Buddies) is the one proposed by Timothy doctor Wood shares perception (memory sharing-aware) virtual machine position selection system based on internal memory, comprise an internal memory recognition system (memory fingerprinting system), effectively can judge that the internal memory between one group of virtual machine shares potential (sharing potential), and calculate more effective modes of emplacement.In addition, along with load variations, system also will utilize online migration (live migration) to optimize virtual machine and place.Its advantage finds the identical virtual machine of virtual memory page from group of planes scope, and make them move to same Cluster nodes, shared virtual memory, can improve physical memory utilization factor, save internal memory, the virtual machine promoting a group of planes holds quantity.Its shortcoming is that procedure is more complicated because needs realize selecting position by virtual machine (vm) migration, and if the virtual machine of shared virtual memory is too much, the service quality of virtual server cannot be ensured.
Sum up existing virtual machine position selection method known, bolus dressing takies the virtual machine position selection method that minimum Cluster nodes is target, causes resources of virtual machine to seize probability problems of too, and the service quality of virtual server also can reduce greatly; Itemize method and Load-aware method all with group of planes load balancing for target, respectively it is considered that the virtual machine number that Cluster nodes runs and Cluster nodes cpu resource utilization factor, have ignored other resource service conditions such as the different same of each virtual machine request resource and node internal memory, network to the impact of node load, also namely only consider cpu resource utilization power, lack the consideration of comprehensive resources utilization factor situation; The internal memory method of getting close to maximizes virutal machine memory within the scope of a group of planes to be shared as target, though improve the open ended virtual machine quantity of a group of planes, selects process need virtual machine (vm) migration, and have the loss of virtual server service quality.
Sum up existing invention present situation known, cloud computing is as service mode emerging in recent years, increasingly extensive attention is obtained in recent years in research circle and industry member, but due to cloud computing newer, still lack the invention of the virtual machine position selection system of facing cloud computing environment at present, especially lack the invention of the virtual machine position selection system of the facing load balance in cloud computing environment.
Therefore, namely the present invention is for this emerging Development Technology of cloud computing, and the existing above problem of selecting a technology and existing, and has invented the virtual machine position selection system of the facing load balance in a kind of cloud computing environment.
Summary of the invention
1, object
The object of the invention is the problem such as appropriate distribution for virtual resource in cloud computing environment, especially for the problem of virtual machine position selection (namely how rightly for virtual machine selects host), resourceoriented load balancing target, invent the virtual machine position selection system of the facing load balance in a kind of cloud computing environment, virtual machine network throughput in final lifting cloud computing environment, the resource reducing virtual machine seizes probability, improves the service quality of fictitious host computer and improves system load balancing in cloud computing environment.
2, technical scheme
Technical scheme of the present invention is as follows: the virtual machine position selection system of the facing load balance in a kind of cloud computing environment, on module composition, primarily of front-end module and rear module composition.
1) front-end module: front-end module comprises sub-function module further: integrated load computing module, module selected by host.
2) rear module: run on each distributed rear terminal node, mainly carries out the load information work for the treatment of of each distributed node.Rear module is made up of the set of multiple distributed node proxy module.Node proxy module is made up of node load collection module, node load prediction module, node incremental loading computing module again.
Shown in wherein the function of each module in charge and deployed position are described below:
● front-end module: play comprehensive management role, be responsible for accepting newly-built virtual machine request and responding.Run on front leaf.
● integrated load computing module: the integrated load carrying out cloud computing system calculates, and integrated load result is transmitted.Run on front leaf.
● module selected by host: with integrated load value vector for input, meet within the scope of the node of request in available resources, perform linear search algorithm, according to load balancing principle, selects the node that integrated load value is minimum, as the target host of newly-built virtual machine.
● rear module: run on each distributed rear terminal node, mainly carry out the load information work for the treatment of of each distributed node.Rear module is made up of the set of multiple distributed node proxy module.
● node proxy module: as the agency of each distributed physical nodes.Mainly comprise node load collection module, node incremental loading computing module three sub-module compositions.Be deployed on each node in cloud computing environment.
● node load collection module: each node obtains the load information on this node.As CPU idleness, memory usage etc.
● node load prediction module: each node is by the current original load information predicting subsequent time with the original load information of previous moment.
● node incremental loading computing module: first each node carries out incremental computations to load, then the load data of prediction compares with previous moment load information by node, then perform and select operation as follows: if increment is greater than threshold value, predict the outcome and be transmitted to the front-end module of front leaf as this node load information, the current original load information on this node is replaced the original load information of previous moment simultaneously.If increment is not more than threshold value, the front-end module not to leaf before a group of planes transmits load information, directly on this node, current original load information is replaced with the original load information of previous moment.
Based on above module composition, on core algorithm, the present invention discloses the virtual machine position selection algorithm of the facing load balance in a kind of cloud computing environment.The basic procedure of this algorithm is:
S1: what front-end module acceptance was new selects a request instruction.
S2: front-end module triggers each node proxy module.
S3: each node proxy module calls node load collection module: each node proxy module accepts to select a request instruction from front-end module, calls node load collection module, performs node load and collect.
S4: each node proxy module calls node load prediction module, performs node load estimation.
S5:S5 forms primarily of three sub-steps, i.e. S5:S51: each node proxy module calls node incremental loading computing module, S5:S52: judge whether load changes, S5:S53: transmission is loaded to front-end module by node incremental computations module.Perform node incremental loading to calculate.
S6: front-end module calls integrated load computing module, performs integrated load and calculates.
S7: front-end module calls host and selects module, completes host and selects.
3, advantage and effect
The virtual machine position selection system of the facing load balance in a kind of cloud computing environment that the present invention announces, it compared with prior art, its main advantage is: (1) just selects host according to load balancing principle in the initial virtual machine creating phase, the later stage can be reduced well due to node load imbalance each in cloud computing, the possibility that virtual machine (vm) migration and resource are redistributed need be carried out, promote load equilibrium and the fault-tolerance of cloud computing system.(2) because cloud computing is in recent years suggested as novel calculating and method of servicing, and it is increasingly extensive to obtain industry member (finance, government, enterprise etc.), promotion and application, and virtual machine position selection system is the vital module in its Resourse Distribute, therefore the present invention has very strong practicality and very wide range of application.
Accompanying drawing explanation
The virtual machine position selection system overall framework schematic diagram of the facing load balance in Fig. 1 cloud computing environment of the present invention
The virtual machine position selection system schematic flow sheet of the facing load balance in Fig. 2 cloud computing environment of the present invention
Embodiment
Express clearly clear for making the object, technical solutions and advantages of the present invention, below in conjunction with drawings and the specific embodiments, the present invention is further described in more detail.As shown in Figure 2, the virtual machine position selection system of the facing load balance in cloud computing environment, concrete implementation step is as follows:
A virtual machine position selection system for facing load balance in cloud computing environment, on module composition, primarily of front-end module MF1 and rear module ME1 composition.
1) front-end module MF1: front-end module comprises sub-function module further: integrated load computing module MF11, module MF12 selected by host.Run on the front terminal node of cloud computing system.
2) rear module ME1: rear module ME1 is made up of multiple distributed node proxy module ME2 set.Node proxy module is made up of load collection module ME211, load prediction module ME212, incremental computations module ME213 again.Run on each distributed rear terminal node, mainly carry out the load information work for the treatment of of each distributed node.
Shown in wherein the function of each module in charge and deployed position are described below:
3) front-end module MF1: play comprehensive management role, is responsible for accepting newly-built virtual machine request and responding.Run on front leaf.
4) integrated load computing module MF11: the integrated load carrying out cloud computing system calculates, and integrated load result is transmitted.Run on front leaf.
5) module MF13 selected by host: with integrated load value vector for input, meet within the scope of the node of request in available resources, perform linear search algorithm, according to load balancing principle, select the node that integrated load value is minimum, as the target host of newly-built virtual machine.
6) rear module ME1: run on each distributed rear terminal node, mainly carries out the load information work for the treatment of of each distributed node.Rear module ME1 is made up of multiple distributed node proxy module ME2 set.
7) node proxy module ME2: as the agency of each distributed physical nodes.Mainly comprise node load collection module ME211, node incremental loading computing module ME213 tri-sub-module compositions.Be deployed on each node in each cloud computing environment.
8) node load collection module ME211: each node obtains the load information on this node.As CPU idleness, memory usage etc.
9) node load prediction module ME212: each node is by the current original load information predicting subsequent time with the original load information of previous moment.
10) node incremental loading computing module ME213: first each node carries out incremental computations to load, then the load data of prediction compares with previous moment load information by node, then perform and select operation as follows: if increment is greater than threshold value, predict the outcome and be transmitted to the front-end module (MF1) of front leaf as this node load information, the current original load information on this node is replaced the original load information of previous moment simultaneously.If increment is not more than threshold value, the front-end module MF1 not to leaf before a group of planes transmits load information, directly on this node, current original load information is replaced with the original load information of previous moment.
Based on above module composition, on core algorithm, the virtual machine position selection algorithm of the facing load balance in a kind of cloud computing environment that the present invention announces.The basic implementing procedure of this algorithm and example are:
What S1: front-end module MF1 acceptance was new selects a request instruction, and in this example, instruction example is " virtual machine creating " and " virtual machine position selection ".
S2: front-end module MF1 triggers each node proxy module ME2, logins other nodes and long-distance support codes implement " triggering " in this example by arranging front leaf user without password.
S3: each node proxy module ME2 calls node load collection module ME211.In this example, each node proxy module ME2 calls and accepts to select a request instruction (" virtual machine creating " and " virtual machine position selection ") from front-end module MF1, then calls node load collection module ME211, performs node load and collects.Concrete collection method is that ME2 obtains the load information of node by multithread mode and SAR instrument, IOSTAT method.Load information comprises CPU idleness, memory usage, disk read-write speed, network reception and transmission rate.
S4: each node proxy module ME2 calls node load prediction module ME212, performs node load estimation.
In this example, the load information normalization first will obtained in S3, derives CPU usage (α cR), memory usage (α mR), network utilization (α nR) and disk transfer rate (P).Wherein CPU usage, memory usage get actual numerical value, and network utilization is network reception speed (v nRS), transmission rate (v nSS) sum and transmission network maximum rate (v mNS) ratio, because the platform in this example adopts hundred Broadcoms, so v mNS=100Mbps, and α nR=(v nRS+ v nSS)/v mNS.Disk transfer rate (P) is similar, adopts disk read-write speed (v dRS, v dWS) sum is divided by disk optimal transmission rate (v mDS), wherein obtain disk iptimum speed by hdparm order.Through the repeatedly test to 3 station servers in this example, find that hdparm order is tested the Timing cached reads speed returned and is no more than 4170MB/s, Timing buffered disk reads speed is no more than 92Mbps, therefore adopts 92Mbps as disk iptimum speed.Then, the block size BLKS (blocksize) of server disk is determined.By checking that disk unit order obtains the details of subregion, determine that experiment porch server B lock size is 4096B and 4KB.Can determine according to above information
P=(v DRS+v DWS)B×8/v MDS
cR,α mR,α nR,p) tbe expressed as current (t) load data, node preserves previous moment (t simultaneously 0moment) load data (α cR0,α mR0,α nR0,p 0) t 0, by the linear prediction model of two groups of data, we are known subsequent time (be defined as next second according to the ageing of load data herein, be designated as the n moment) load data (α cRn,α mRn,α nRn,pn) nwith t, t 0there is following relation in moment load data:
α CRn - α CR n - t = α CR - α CR 0 t - t 0 , α MRn - α MR n - t = α MR - α MR 0 t - t 0 , α NRn - α NR n - t = α NR - α NR 0 t - t 0 , P n - P n - t = P - P 0 t - t 0 . - - - ( 1 )
From n-t=1, the solution of aforesaid equation is obtained by following system of equations
α CRn = α CR + α CR - α CR 0 t - t 0 , α MRn = α MR + α MR - α MR 0 t - t 0 , α NRn = α NR + α NR - α NR 0 t - t 0 , P n = P + P - P 0 t - t 0 . - - - ( 2 )
Above-mentioned solution of equations is exactly the subsequent time node load data of prediction.
S5: each node proxy module ME2 calls node incremental loading computing module ME213, performs node incremental loading and calculates.
In this example, this step mainly needs to calculate n moment load (α cRn,α mRn,α nRn,pn) whether relative to t 0moment load (α cR0,α mR0,α nR0,p 0) there is significant change, if every variable quantity is all less than threshold value, be designated as unchanged, node sends prediction load data, to reduce the network overhead in cloud computing system, simultaneously by t load data (α without the need to forward end node cR,α mR,α nR,p) t is assigned to 0in the moment, preserved by node module; Otherwise, the relative t of definition predicted data 0moment load changes, and node module answers front end module to send prediction load data, and by t load data (α cR,α mR,α nR,p) t is assigned to 0in the moment, preserved by node module.According to reservation 2 significant digits significant figure and the principle that rounds up in this example, when every load data variable quantity is less than 0.05 (namely threshold value is set as 0.05), can think that load data is unchanged.So, if the relative t of predicted data 0moment load is unchanged, adopts following inequality group
| &alpha; CRn - &alpha; CR 0 | < 0.05 , | &alpha; MRn - &alpha; MR 0 | < 0.05 , | &alpha; NRn - &alpha; NR 0 | < 0.05 , | P n - P 0 | < 0.05 . - - - ( 3 )
S6: front-end module MF1 calls integrated load computing module MF11, performs integrated load and calculates.It is CPU that web services due to a virtual server shows maximally related part, next is internal memory, network interface card again, it is finally disk, therefore in this example, every load information weight can be defined as 40 percent, 30 percent, 20 percent and 10 successively, and integrated load value is produced by cpu busy percentage, memory usage, network utilization and the weighting of disk transfers utilization factor.In this example, to each node load data weighting respectively of preserving, each node integrated load value is calculated.According to the reasoning in S1-S5 formula, the every integrated load information weight of CPU usage, memory usage, network utilization, disk transfer rate is followed successively by 0.4,0.3,0.2 and 0.1.So node integrated load value, is designated as LV, meet
LV=0.4×α CR+0.3×α MR+0.2×α NR+0.1×P. (4)
Finally, each node integrated load value gone out according to above formulae discovery is sent to host and select module MF12.
S7: front-end module MF1 calls host selects module MF12, completes host and selects.
Computation structure in this example in Main Basis S6, according to load balancing principle, meets within the scope of the node of virtual machine request in available resources, select the node that integrated load value (LV) is minimum, as the target host of newly-built virtual machine.So far, virtual machine position selection flow process is completed.
It should be noted last that: above embodiment is the unrestricted technical scheme of the present invention in order to explanation only, although with reference to above-described embodiment to invention has been detailed description, those of ordinary skill in the art is to be understood that: still can modify to the present invention or equivalent replacement, and not departing from any modification or partial replacement of the spirit and scope of the present invention, it all should be encompassed in the middle of right of the present invention.

Claims (2)

1. a virtual machine position selection system for the facing load balance in cloud computing environment, on module composition, is made up of front-end module (MF1) and rear module (ME1);
1) front-end module (MF1): front-end module comprises sub-function module further: integrated load computing module (MF11), module (MF12) selected by host;
2) rear module (ME1): run on each distributed rear terminal node, carries out the load information work for the treatment of of each distributed node; Rear module (ME1) was made up of multiple distributed node proxy module (ME2) set; Node proxy module is made up of node load collection module (ME211), node load prediction module (ME212), node incremental loading computing module (ME213) again;
Wherein, shown in the function of each module in charge and deployed position are described below:
● front-end module (MF1): play comprehensive management role, be responsible for accepting newly-built virtual machine request and responding; Run on front leaf;
● integrated load computing module (MF11): the integrated load carrying out cloud computing system calculates, and integrated load result is transmitted; Run on front leaf;
● module (MF13) selected by host: with integrated load value vector for input, meet within the scope of the node of request in available resources, perform linear search algorithm, according to load balancing principle, select the node that integrated load value is minimum, as the target host of newly-built virtual machine;
● rear module (ME1): run on each distributed rear terminal node, carry out the load information work for the treatment of of each distributed node; Rear module (ME1) was made up of multiple distributed node proxy module (ME2) set;
● node proxy module (ME2): as the agency of each distributed physical nodes; Comprise node load collection module (ME211), node incremental loading computing module (ME213) and node load prediction module (ME212) three sub-module compositions; Be deployed on each rear terminal node in cloud computing environment;
● node load collection module (ME211): each rear terminal node obtains the load information on this node;
● node load prediction module (ME212): each rear terminal node is by the current original load information predicting subsequent time with the original load information of previous moment;
● node incremental loading computing module (ME213): first each rear terminal node carries out incremental computations to load, then the load data of prediction compares with previous moment load information by node, then perform and select operation as follows: if increment is greater than threshold value, predict the outcome and be transmitted to the front-end module (MF1) of front leaf as this node load information, the current original load information on this node is replaced the original load information of previous moment simultaneously; If increment is not more than threshold value, does not transmit load information to the front-end module (MF1) of leaf before a group of planes, directly on this node, current original load information is replaced with the original load information of previous moment.
2., based on the virtual machine position selection method of the facing load balance in a kind of cloud computing environment of virtual machine position selection system according to claim 1, basic procedure is:
S1: front-end module (MF1) accepts new to select a request instruction;
S2: front-end module (MF1) triggers each rear terminal node proxy module (ME2);
S3: each rear terminal node proxy module (ME2) calls node load collection module (ME211): each rear terminal node proxy module (ME2) accepts to select a request instruction from front-end module (MF1), call node load collection module (ME211), perform node load and collect;
S4: each rear terminal node proxy module (ME2) calls node load prediction module (ME212), performs node load estimation;
S5:S5 is made up of three sub-steps, i.e. S5:S51: each rear terminal node proxy module (ME2) calls node incremental loading computing module (ME213), S5:S52: judge whether load changes, S5:S53: transmission is loaded to front-end module (MF1) by node incremental computations module (ME213); Perform node incremental loading to calculate;
S6: front-end module (MF1) calls integrated load computing module (MF11), performs integrated load and calculates;
S7: front-end module (MF1) calls host and selects module (MF12), completes host and selects.
CN201210001315.9A 2012-01-04 2012-01-04 Virtual machine position selection system facing load balance in cloud computation environment Active CN102567080B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201210001315.9A CN102567080B (en) 2012-01-04 2012-01-04 Virtual machine position selection system facing load balance in cloud computation environment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201210001315.9A CN102567080B (en) 2012-01-04 2012-01-04 Virtual machine position selection system facing load balance in cloud computation environment

Publications (2)

Publication Number Publication Date
CN102567080A CN102567080A (en) 2012-07-11
CN102567080B true CN102567080B (en) 2015-03-04

Family

ID=46412578

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201210001315.9A Active CN102567080B (en) 2012-01-04 2012-01-04 Virtual machine position selection system facing load balance in cloud computation environment

Country Status (1)

Country Link
CN (1) CN102567080B (en)

Families Citing this family (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103885831B (en) * 2012-12-19 2017-06-16 中国电信股份有限公司 The system of selection of virtual machine host machine and device
CN103473113B (en) * 2013-09-04 2017-02-08 国云科技股份有限公司 Universal virtual-machine adopting method
CN103595763B (en) * 2013-10-15 2016-08-24 北京航空航天大学 resource scheduling processing method and device
CN104133727A (en) * 2014-08-08 2014-11-05 成都致云科技有限公司 Load distribution method based on real-time resources
KR102123422B1 (en) 2014-09-12 2020-06-16 인텔 코포레이션 Memory and resource management in a virtual computing environment
CN104750541B (en) * 2015-04-22 2018-01-16 成都睿峰科技有限公司 A kind of virtual machine migration method
WO2017100578A1 (en) 2015-12-10 2017-06-15 Affirmed Networks, Inc. Data driven automated provisioning of telecommunication applications
US10129769B2 (en) 2015-12-31 2018-11-13 Affirmed Networks, Inc. Adaptive peer overload control in mobile networks
CN108464031B (en) 2016-01-15 2019-11-08 阿弗梅德网络公司 The redundancy based on database in telecommunication network
CN108028850B (en) * 2016-08-08 2021-03-26 深圳秦云网科技有限公司 Private cloud management platform
CN108536525B (en) * 2017-03-02 2021-01-29 北京金山云网络技术有限公司 Host machine scheduling method and device
CN110058966B (en) * 2018-01-18 2023-11-14 伊姆西Ip控股有限责任公司 Method, apparatus and computer program product for data backup
CN109710269A (en) * 2018-09-07 2019-05-03 天翼电子商务有限公司 A kind of list application discrete type clustered deploy(ment) device and method
EP3912035A1 (en) 2019-01-15 2021-11-24 Microsoft Technology Licensing, LLC Dynamic auto-configuration of multi-tenant paas components
CN112764918B (en) * 2020-12-29 2022-02-18 赛韵网络科技(上海)有限公司 Working method for carrying out space search on available area by cloud platform

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101593133A (en) * 2009-06-29 2009-12-02 北京航空航天大学 Load balancing of resources of virtual machine method and device
CN102004671A (en) * 2010-11-15 2011-04-06 北京航空航天大学 Resource management method of data center based on statistic model in cloud computing environment
CN102195890A (en) * 2011-06-03 2011-09-21 北京大学 Internet application dispatching method based on cloud computing

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101593133A (en) * 2009-06-29 2009-12-02 北京航空航天大学 Load balancing of resources of virtual machine method and device
CN102004671A (en) * 2010-11-15 2011-04-06 北京航空航天大学 Resource management method of data center based on statistic model in cloud computing environment
CN102195890A (en) * 2011-06-03 2011-09-21 北京大学 Internet application dispatching method based on cloud computing

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
《虚拟计算环境下虚拟机资源负载均衡方法》;刘媛媛等;《计算机工程》;20100831;第36卷(第16期);第30-32页 *

Also Published As

Publication number Publication date
CN102567080A (en) 2012-07-11

Similar Documents

Publication Publication Date Title
CN102567080B (en) Virtual machine position selection system facing load balance in cloud computation environment
Ge et al. GA-based task scheduler for the cloud computing systems
CN103713956B (en) Method for intelligent weighing load balance in cloud computing virtualized management environment
CN104375897B (en) Cloud computing resource scheduling method based on the unbalanced degree of minimum relative load
CN102236582A (en) Method for balanced distribution of virtualization cluster load in a plurality of physical machines
CN103401939A (en) Load balancing method adopting mixing scheduling strategy
CN105975340A (en) Deployment algorithm of application distribution of virtual machine
CN106572500B (en) Scheduling method of hardware accelerator in C-RAN
Li et al. An effective scheduling strategy based on hypergraph partition in geographically distributed datacenters
CN103425536A (en) Test resource management method oriented towards distributed system performance tests
CN107291550A (en) A kind of Spark platform resources dynamic allocation method and system for iterated application
CN110990154A (en) Big data application optimization method and device and storage medium
CN105760227B (en) Resource regulating method and system under cloud environment
Subalakshmi et al. Enhanced hybrid approach for load balancing algorithms in cloud computing
Su et al. Variable-sized map and locality-aware reduce on public-resource grids
He et al. Research on task scheduling strategy optimization based onaco in cloud computing environment
Yan et al. Affinity-aware virtual cluster optimization for mapreduce applications
Kumar et al. A priority based dynamic load balancing approach in a grid based distributed computing network
CN106155782B (en) Virtual machine migration method and device
Zhang et al. Speeding up vm startup by cooperative vm image caching
CN104166593A (en) Method for computing asynchronous and concurrent scheduling of multiple application functions
Adhikari et al. An enhanced dynamic load balancing mechanism for task deployment in IaaS cloud
CN114237866A (en) Cloud platform computing resource scheduling method and system for relay protection setting
Jaikar et al. Priority-based virtual machine load balancing in a scientific federated cloud
Zhao et al. A holistic cross-layer optimization approach for mitigating stragglers in in-memory data processing

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
TR01 Transfer of patent right

Effective date of registration: 20210220

Address after: Room 203-204, building Y2, 112 liangxiu Road, Pudong New Area, Shanghai, 201203

Patentee after: SHANGHAI ZHIRUI ELECTRONIC TECHNOLOGY Co.,Ltd.

Address before: 100191 No. 37, Haidian District, Beijing, Xueyuan Road

Patentee before: BEIHANG University

TR01 Transfer of patent right