CN105430049B - A kind of virtual streaming media cluster collaboration moving method based on DCN - Google Patents
A kind of virtual streaming media cluster collaboration moving method based on DCN Download PDFInfo
- Publication number
- CN105430049B CN105430049B CN201510718803.5A CN201510718803A CN105430049B CN 105430049 B CN105430049 B CN 105430049B CN 201510718803 A CN201510718803 A CN 201510718803A CN 105430049 B CN105430049 B CN 105430049B
- Authority
- CN
- China
- Prior art keywords
- cluster
- virtual machine
- machine
- subregion
- physical machine
- 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
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/56—Provisioning of proxy services
- H04L67/563—Data redirection of data network streams
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L65/00—Network arrangements, protocols or services for supporting real-time applications in data packet communication
- H04L65/60—Network streaming of media packets
- H04L65/75—Media network packet handling
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/104—Peer-to-peer [P2P] networks
- H04L67/1044—Group management mechanisms
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/131—Protocols for games, networked simulations or virtual reality
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Multimedia (AREA)
- Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Theoretical Computer Science (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The present invention cooperates with moving method for a kind of virtual streaming media cluster based on DCN, first, builds the traffic communication topological diagram between virtual streaming media server；According to the traffic communication topological diagram between virtual streaming media server divide towards the small cluster topology division of physical machine and towards the big cluster topology of subregion；According to the Topology partition of virtual streaming media server position and 2 levels as a result, assign respectively towards subregion big cluster and subregion correspondence and towards the small cluster of physical machine and the correspondence of physical machine；Finally, moving method is cooperateed with by virtual streaming media cluster, on all virtual machine (vm) migrations to be migrated to purpose physical machine.
Description
Technical field
The invention belongs to cloud computing DCN technical fields, be related to the division of streaming media cluster under DCN environment, place and
Migration field, more particularly to a kind of virtual streaming media cluster collaboration moving method based on DCN.
Background technology
Nowadays the development of cloud computing is subject to people more and more to pay close attention to, and the related application based on cloud computing also gradually incorporates
Among the daily life of people, therefore in order to meet the various needs of users, cloud service supplier constantly disposes in cloud platform
Large-scale application service.With increase of the user to the resource quantity and resource category demand of data center, simply expand
Data center's scale has not been a kind of effective solution because this can greatly increase data center construction into
This, while must not also reduce application service quality.Therefore, in order to solve this problem, researcher proposes cloud computing ring
The solution that multi-dummy machine migrates under border.Retrieve following several related to the present invention to belong to cloud computing DCN fields
Patent, they are respectively：
1. Chinese patent 201510036992.8, a kind of data migration method of data center's multi-dummy machine；
2. Chinese patent 201210551631.3, a kind of application cluster moving method and device；
A kind of data migration method of data center's multi-dummy machine is provided in above-mentioned patent 1, belongs to field of cloud calculation.
The present invention is based on the migrations of the parallel multi-dummy machine based on precloning that Franco Callegati and Walter Cerroni are proposed
Based on strategy and the serial multi-dummy machine migration strategy based on rear duplication proposed in the present invention, introduce parameter m and repair
Positive divisor α is proposed under a kind of constraints for meeting the maximum downtime that service provider consults with user, the gross migration time
Minimum data migration method, this method is applied widely, can further reduce the transit time cost of service provider.
A kind of application cluster moving method and device, this method are disclosed in above-mentioned patent 2 to be included：In cloud computing platform
In, receive the application cluster migration request that user sends according to the agreement of migration frame；To the application cluster migration request into
Row parsing determines the migration attribute information for the application cluster that request is migrated；According to the migration attribute information determined, in sky
The node resource of application cluster migration demand is determined for compliance in not busy resource pool, according to the node resource determined and migrates frame,
The application cluster for asking migration is moved in the node resource determined.Using the above program, cloud meter can preferably be improved
Calculate the transport efficiency of application cluster in platform.
It is looked into newly according to above-mentioned, the problems of prior art is not account for the high resource consumption of streaming media server
And the characteristics of close traffic communication, this can cause the heavy congestion of entire system for cloud computing, the inside band of a large amount of consumption networks
Width, and then seriously affect the performance of Stream Media Application service.
The content of the invention
The shortcomings that in order to overcome the above-mentioned prior art, it is an object of the invention to provide a kind of virtual stream matchmakers based on DCN
Body cluster cooperates with moving method, and on the premise of the external service performance of Stream Media Application is ensured, the interior of DCN networks is greatly lowered
Portion's bandwidth consumption, while reduce the bulk migration time cost of streaming media cluster.
To achieve these goals, the technical solution adopted by the present invention is：
A kind of virtual streaming media cluster collaboration moving method based on DCN,
First, the traffic communication topological diagram between virtual streaming media server is built；
According to the traffic communication topological diagram between virtual streaming media server draw towards the small cluster topology of physical machine
Divide and divided towards the big cluster topology of subregion；
According to the Topology partition of virtual streaming media server position and 2 levels as a result, assigning towards the big of subregion
It the correspondence of cluster and subregion and assigns towards the small cluster of physical machine and the correspondence of physical machine；
Finally, moving method is cooperateed with by virtual streaming media cluster, all virtual machine (vm) migrations to be migrated to purpose thing
On reason machine.
Traffic communication topological diagram between the virtual streaming media server of structure is expressed as Gvm=(V, E), wherein V=
{Vm1,Vm2......Vmn, i.e., each virtual machine is a point in traffic communication topological diagram, and n is virtual machine number, and E is
Side collection is represented between virtual machine with the presence or absence of communication；It communicates if existing between virtual machine i and virtual machine j, Eij=1, otherwise Eij
=0；Side right value WijCommunication size between virtual machine i and virtual machine j；Virtual machine i is expressed as Vmi=(c, b), wherein c tables
Show the cpu resource needed for virtual machine i, b represents the bandwidth resources needed for virtual machine i.
The small cluster topology division towards physical machine is as follows：
Step1：Each virtual machine is initialized as a small cluster towards physical machine；
Step2：Side gain is calculated, side gain accounts for all of institute's tie point to square directly proportional and side right value of side right value
The percentage of the sum of side right value is directly proportional, side tie point resource and is inversely proportional；
Step3：It is more than 0 if there is side gain, then it is one towards object to merge 2 points connect when gain is maximum
The small cluster of reason machine；
Step4：Step2 is repeated, until all side gains are all not more than 0；
Step5：Merge remaining isolated node for a small cluster towards physical machine.
The big cluster topology division of the basis towards subregion is as follows：
Step1：The small cluster initialized towards physical machine is a big cluster towards subregion；
Step2：Side gain is calculated, side gain accounts for all of institute's tie point to square directly proportional and side right value of side right value
The percentage of the sum of side right value is directly proportional, side tie point resource and is inversely proportional；
Step3：It is more than 0 if there is side gain, then it is a face to merge 2 small clusters connect when gain is maximum
To the big cluster of subregion；
Step4：Step2 is repeated, until all side gains are all not more than 0；
Step5：Merge remaining isolated small cluster for a big cluster towards subregion.
The appointment is as follows towards the big cluster of subregion and the correspondence of subregion：
Step1：Internal storage data total amount formula Data (the i)=B*Vm copied during virtual machine (vm) migrationi(c)/(B-Vmi
(b)) represent, in formula：B represents transmission rate, Vmi(c) virutal machine memory size, Vm are representedi(b) virutal machine memory containing dirty pages are represented
Generation rate；
Step2：The distance virtual machine (vm) migration of virtual machine and subregion copy internal storage data total amount is passed through with it to subregion
Communication switchboard number product representation；
Step3：It is all virtual to be included towards the big cluster of subregion towards the big cluster of subregion and the distance of subregion
The sum of machine and subregion distance；
Step4：According to the distance of the big cluster towards subregion to subregion, optimum allocation is acquired using Hungary's dispatching algorithm
Scheme.
The appointment is as follows towards the small cluster of physical machine and the correspondence of physical machine：
Step1：It is otherwise 1 if deploying virtual machine is 0 with physical machine distance in physical machine, virtual machine；
Step2：The distance of small cluster and physical machine towards physical machine is the institute that is included towards the small cluster of physical machine
There are the sum of virtual machine and physical machine distance；
Step3：According to the distance of the small cluster towards physical machine to physical machine, acquired using Hungary's dispatching algorithm optimal
Allocative decision.
It is described that moving method is cooperateed with by virtual streaming media cluster, all virtual machine (vm) migrations to be migrated to purpose physics
It is as follows on machine：
Step1：Judge whether the destination host resource of virtual machine to be migrated meets, virtual machine to be migrated is divided into transportable
Queue and etc. queue to be migrated；
Step2：The network gain that the virtual machine (vm) migration in transportable queue is brought is calculated, the network brought by its migration increases
The arrangement of beneficial descending, virtual machine (vm) migration gain by after virtual machine (vm) migration reduction DCN inner stream flow size；
Step3：If transportable queue is sky, Step6 is gone to；Otherwise, virtual machine is taken in order from transportable queue,
The migration path of virtual machine is calculated, migration path bandwidth is expressed as B；
Step4：If migration path bandwidth B is more than 1.5 times of virutal machine memory containing dirty pages generation rate, migration item is unsatisfactory for
Part goes to Step3；Otherwise, its migration task is started；
Step5：Virtual machine completes migration, updates the migration gain for the virtual machine being attached thereto, updates source physical host
Available resources, the transportable queue of update and wait resource queue, update DCN network topology matrix available bandwidths；Go to Step3；
Step6：Update purpose physical host can with resource, update it is transportable and etc. queue to be migrated, update DCN networks open up
Flutter matrix available bandwidth；Go to Step3；
Step7：Whether the queues to be migrated such as judgement are empty, and if it is empty, then algorithm terminates；Otherwise, by virtual machine to be moved out
Virtual machine (vm) migration in the physical machine of number minimum is to free physical machine.
Compared with prior art, the present invention considers the spy of the high resource consumption of streaming media server and close traffic communication
Point carries out virtual streaming media cluster collaboration migration by the preferable migration scheme for acquiring virtual streaming media cluster.
Description of the drawings
Fig. 1 is the flow chart that a kind of virtual streaming media cluster based on DCN of the present invention cooperates with moving method.
Specific embodiment
The present invention is described in further details with reference to the accompanying drawings and examples.
As shown in Figure 1, the present invention cooperates with moving method for a kind of virtual streaming media cluster based on DCN, first, structure
Traffic communication topological diagram between virtual streaming media server；According to the traffic communication topological diagram between virtual streaming media server
Divided towards the small cluster topology division of physical machine and towards the big cluster topology of subregion；According to virtual streaming media server
The Topology partition of position and 2 levels as a result, assign the correspondence knead dough of the big cluster and subregion towards subregion respectively
To the small cluster of physical machine and the correspondence of physical machine；Finally, moving method is cooperateed with by virtual streaming media cluster, all
Virtual machine (vm) migration to be migrated is in purpose physical machine.
Technical scheme is described in detail step by step below.
1st, traffic communication topological diagram between the virtual streaming media server of structure
It builds the traffic communication topological diagram between virtual streaming media server and is expressed as Gvm=(V, E), wherein V={ Vm1,
Vm2......Vmn, i.e., each virtual machine is a point in traffic communication topological diagram, and n is virtual machine number, and E is side collection,
It represents between virtual machine with the presence or absence of communication；It communicates if existing between virtual machine i and virtual machine j, Eij=1, otherwise Eij=0；Side
Weights WijCommunication size between virtual machine i and virtual machine j；Virtual machine i is expressed as Vmi=(c, b), wherein c represent virtual
Cpu resource needed for machine i, b represent the bandwidth resources needed for virtual machine i.
2nd, divided towards the small cluster topology of physical machine
According to the traffic communication topological diagram between virtual streaming media server draw towards the small cluster topology of physical machine
Point thinking be：
Initialization small cluster network topological representation be：Gsc=(Vsc, E), midpoint for Imaginary Mechanism into it is small
Cluster, i.e. each virtual machine are initialized as a small cluster as a point in figure first, are expressed as Vsci=(c, b,
D), c represents cpu resource needed for small cluster again simultaneously at this time, and b represents bandwidth resources needed for small cluster again simultaneously, and d represents small cluster
The sum of internal all side right values.Virtual machine i is initialized as small cluster i, and small cluster is initialized as the point i, virtual machine j in figure
J, as the point j in figure, if there are a line e, weights between small cluster i and small cluster j to be expressed as wij=Evscij+
EvscjiThe sum of uninterrupted between i.e. small cluster, if there is no side, w between small cluster i and small cluster jij=0.
Small assemblage classification towards physical machine is to reachIt is maximum.
eijAnd aiCalculation formula be：
E in formulaijRepresent that the side right value of tie point i and point j put a spot the sum of i and point j internal edges weights.aiRepresent point i institutes
There are the sum of the weights on side.Due to needing the demand of consideration node size, i.e. resources of virtual machine, the calculation formula of side gain delta Q：
M=max { (Vsci(c)+Vscj(c))/C,(Vsci(b)+Vscj(b))/B}
In formula, wijRepresent the weights on the side of tie point i and point j；C represents the memory of physical machine；B represents the band of physical machine
It is wide；M represents that point i and point j occupies the percentage of physical machine resource after merging.Physical machine resource is discontented with after if two small cluster merges
Its resource requirement in total of foot, then Δ Q is minimum value 0.
It concretely comprises the following steps：
Step1：Input data：Virtual machine traffic relational graph (matrix) Gvm=(V, E), physical machine memory size C, physics
Machine amount of bandwidth B.
Step2：The each virtual machines of Gvm=(V, E) as a small cluster, Gsc=(Vsc, E) is initialized as, in Gsc
Each point is a small cluster.
Step3：Calculate eijAnd aiMeet：
Step4：Calculating each has the gain delta Q while after the small cluster being connected merges successively：
M=max { (Vsci(c)+Vscj(c))/C,(Vsci(b)+Vscj(b))/B}
Step5：If there is no Δ Q ＞ 0, terminate algorithm；Otherwise, the point i and point j corresponding to the side of Δ Q maximums are merged.
Point is Vsc after mergingnew：
Vscnew(c)=Vsci(c)+Vscj(c)
Vscnew(b)=Vsci(b)+Vscj(b)
Vscnew(d)=Vsci(d)+Vscj(d)+wij
Step6：Update the side right value between all points being connected with point i with point j, wnew,k=wik+wjk, go to Step3.
3rd, divided towards the big cluster topology of subregion
The Gsc=(Vsc, E) obtained after Gvm=(V, E) is divided into towards the small cluster of physical machine, VsciRepresent small collection
Group i, WscijRepresent the weights on the side between small cluster i and small cluster j, Vpci=(h, d) represents big cluster i, Vpci(h) represent
The number of the small cluster included in big cluster i, Vpci(d) between the small cluster that the big cluster i internal edges weights of expression include
Side right value, WpcijRepresent the weights between big cluster i and big cluster j, N represents the number of the physical machine contained in a subregion.
Community detecting algorithm has equally been used towards the big assemblage classification algorithm of subregion, has provided e ' firstijWith a 'iIt calculates public
Formula：
E ' in formulaijRepresent that the side right value of tie point i and point j put a spot the sum of i and point j internal edges weights.a′iRepresent point i
The sum of the weights on all sides.Provide the calculation formula of side gain delta Q：
M=(Vpci(h)+Vscj(h))/N
In formula, WpcijRepresent the weights on the side of tie point i and point j；N represents the number of physical machine in subregion；M represents point i
Physical machine number percentage in subregion is occupied after merging with point j.Physical machine number is more than in subregion after if two big cluster merges
Physical machine number, then Δ Q is minimum value 0.
It is as follows：
Step1：Input data：Towards the small cluster topology graph Gsc'=(Vsc, E) of physical machine, physical machine number N in subregion.
Step2：Each small cluster is expressed as, each point is one in Gpc as a big cluster in Gsc'=(Vsc, E)
A small cluster.
Step3：Calculate e 'ijWith a 'iMeet：
Step4：Calculating each has the gain delta Q while after the big cluster being connected merges successively：
M=(Vpci(h)+Vscj(h))/N
Step5：If there is no Δ Q ＞ 0, terminate algorithm.Otherwise, the point i, j corresponding to the side of Δ Q maximums are merged.It closes
And point is Vpc afterwardsnew：
Vpcnew(h)=Vpci(h)+Vpcj(h)
Vpcnew(d)=Vpci(d)+Vpcj(d)+wij
Step6：Update side right value between all points being connected with point i with point j, wnew,k=wik+wjk, go to Step3.
4th, assign towards the big cluster of subregion and the correspondence of subregion
Gpc=(Vpc, E) big cluster topology graph, wherein VpciInclude multiple small clusters, small cluster there are multiple virtual machines；
Dpartion(i, j) represents the interchanger number between subregion in DCN, and different DCN frameworks are different, are handed over for Guinier-Preston zone
Port change planes as p mouthfuls of Tree frameworks, can be formulated：
In formula：I, j represent subregion label, from left to right number consecutively；Represent that subregion is handed in same aggregation
It changes planes down.With intra-area communication is divided only to pass through time access switch, different subregions communication is by 3 friendships under same aggregation switch
It changes planes, other situations will pass through 5 interchangers.
CostVm (i, j) represents the matrix between virtual machine and each subregion；CostCluster (i, j) represents big cluster
The distance between each subregion matrix；Assigned (i, j) represents appointment matrix, this is the solution required by this algorithm,
Assigned (i, j)=0 represents that big cluster i is assigned to subregion j.PartionNow (i) represents the subregion at the present place of virtual machine
Function, this is known quantity.PartionTo (i) represents the partition functions to be migrated of virtual machine, if assigned (i, j)=1
PartionTo (i)=j；Data (i) represents virtual machine (vm) migration transmitted data amount；M represents big cluster number；N represents subregion
Number.
1) big cluster and subregion distance are defined and calculate
Virtual machine (vm) migration process needs to copy entire memory, and the transition process virtual machine (vm) migration to subregion is estimated in DCN
Distance of the migrating data amount of generation as virtual machine and subregion.Virtual machine (vm) migration process will transmit entire memory, constantly copy
Shellfish containing dirty pages and the remaining containing dirty pages of shutdown copy.If virtual machine VmiIn save as Vmi(c), it is Vm that containing dirty pages, which generate rate,i(r), link passes
Defeated rate B, then virtual machine VmiMigration needs the transmitted data amount to be altogether：
Virtual machine (vm) migration first time iteration will send entire memory, take as Vmi(c)/B, this process generate (Vmi
(c)*Vmi(b)) data of/B, second of iteration send these data.Third time iteration will be sent time-consuming Vmi(c)*(Vmi
(b)/B)2Data.If without considering background traffic, that is, give tacit consent to link bandwidth and be B and can use.Stopping criterion for iteration is to reach most
Big iterations nmaxOr the data transmitted is needed to be less than presetting threshold value.
In order to facilitate calculating, it is proposed that the simplified model of total transmission data.Simplified data transfer model is formulated：
Datasimple(i)=B*Vmi(c)/(B-Vmi(b))
In formula：Vmi(c)/(B-Vmi(b)) illustrate transmission time total in simplified model, i.e., simply think in containing dirty pages not
Stopping pregnancy life and memory are constantly transmitted under double action (containing dirty pages generation adds memory, and memory, which is transmitted, reduces memory),
Memory, which is all transmitted, completes the memory copying stage.The time that the memory copying stage uses is multiplied i.e. with the rate transmitted
For the data volume always transmitted.
The distance of virtual machine and subregion is expressed as：
CostVm (i, j)=Data (i) * Dpartion(PartionNow(i),j)
That is virtual machine i migrations need the data volume transmitted to be multiplied by its place subregion and the communication switchboard number of j.
The distance of big cluster and subregion is with being expressed as：
The sum of all virtual machines and subregion distance that i.e. big cluster is included.
2) according to the distance of the big cluster of calculating and each subregion, the allocative decision of total distance minimum is sought
Acquire CostCluster (i, j) needs to be allocated big cluster and subregion afterwards, and one big cluster can only be put into
One subregion, a subregion can only put a big cluster.Optimization aim is：
In formula：Represent that each subregion can only place a big cluster；Represent that each big cluster can only be put into a subregion.assignedijBig cluster i is represented for 1
It is assigned to subregion j.
5th, assign towards the small cluster of physical machine and the correspondence of physical machine
Gsc=(Vsc, E) represent that chapters and sections 3.3 generate towards the small cluster topology of physical machine；CostVm (i, j) represents virtual
The distance of machine i and physical machine j；CostCluster (i, j) represents the distance of small cluster i and physical machine j, and value is equal to small cluster
The sum of interior all virtual machines and physical machine j distances；Assigned (i, j) represents virtual machine and the assignment matrix of physical machine,
assignedij=1 expression virtual machine i has been assigned to physical machine j.The physical machine that HostNow (i) virtual machines are now disposed；M represents big
Small cluster number in cluster, under normal circumstances m==n；N represents physical machine number in subregion, this value is a constant；
ListVsc represents all small clusters contained of specific big cluster；Hosts represents all physical machines in particular zones.
1) small cluster and physical machine distance
Virtual machine, which must move to the migration cost of physical machine to be selected and virtual machine present position, relation, if there are certain
Physical machine need not then migrate, and virtual machine is 0 with this physical machine migration cost.The physical machine that virtual machine (vm) migration is not present to it
Upper cost is the same, is equal to any of which distance of one because these physical machines are located in same subregion.Therefore
Virtual machine is with physical machine distance：
The small cluster and the range formula of physical machine that can be arrived be：
2) according to small cluster and physical machine distance, the allocative decision apart from summation minimum is found
CostCluster (i, j) is the coefficient matrix of input, and using Hungary Algorithm, just optimum distribution solves.
6th, Hungary's dispatching algorithm
(1) if from coefficient matrix cijRow (or row) each element in be individually subtracted after the least member of the row (or row)
To a new matrix bij, then with bijThere is identical optimal solution for the assignment problem and former problem of coefficient matrix.By above-mentioned conversion
Afterwards, bijIn often row and each column all at least contain there are one 0 element, 0 element for being located at different lines of not going together is claimed to be independent 0 yuan
Element.
(2) if bijThere are n 0 independent elements, it can thus be concluded that a dematrix, method is that order corresponds to b in Xij0
The element of element position is 1, and the element of other positions is 0, then X is the optimal solution of assignment problem.
7th, virtual streaming media cluster collaboration moving method
Moving method is cooperateed with by virtual streaming media cluster, on all virtual machine (vm) migrations to be migrated to purpose physical machine
It is as follows：
Step1：Judge whether the destination host resource of virtual machine to be migrated meets, virtual machine to be migrated is divided into transportable
Queue and etc. queue to be migrated；
Step2：The network gain that the virtual machine (vm) migration in transportable queue is brought is calculated, the network brought by its migration increases
The arrangement of beneficial descending, virtual machine (vm) migration gain by after virtual machine (vm) migration reduction DCN inner stream flow size；
Step3：If transportable queue is sky, Step6 is gone to；Otherwise, virtual machine is taken in order from transportable queue,
The migration path of virtual machine is calculated, migration path bandwidth is expressed as B；
Step4：If migration path bandwidth B is more than 1.5 times of virutal machine memory containing dirty pages generation rate, migration item is unsatisfactory for
Part goes to Step3；Otherwise, its migration task is started；
Step5：Virtual machine completes migration, updates the migration gain for the virtual machine being attached thereto, updates source physical host
Available resources, the transportable queue of update and wait resource queue, update DCN network topology matrix available bandwidths；Go to Step3；
Step6：Update purpose physical host can with resource, update it is transportable and etc. queue to be migrated, update DCN networks open up
Flutter matrix available bandwidth；Go to Step3；
Step7：Whether the queues to be migrated such as judgement are empty, and if it is empty, then algorithm terminates；Otherwise, by virtual machine to be moved out
Virtual machine (vm) migration in the physical machine of number minimum is to free physical machine.
Claims (3)
1. a kind of virtual streaming media cluster collaboration moving method based on DCN, which is characterized in that include the following steps：
First, the traffic communication topological diagram between virtual streaming media server is built, is expressed as Gvm=(V, E), wherein V=
{Vm1,Vm2......Vmn, i.e., each virtual machine is a point in traffic communication topological diagram, and n is virtual machine number, and E is
Side collection is represented between virtual machine with the presence or absence of communication；It communicates if existing between virtual machine i and virtual machine j, Eij=1, otherwise Eij
=0；Side right value WijCommunication size between virtual machine i and virtual machine j；Virtual machine i is expressed as Vmi=(c, b), wherein c tables
Show the cpu resource needed for virtual machine i, b represents the bandwidth resources needed for virtual machine i；
Then, according to the traffic communication topological diagram between virtual streaming media server draw towards the small cluster topology of physical machine
Divide and divided towards the big cluster topology of subregion；
Wherein, the small cluster topology division towards physical machine is as follows：
Step1：Each virtual machine is initialized as a small cluster towards physical machine；The small cluster network topological representation of initialization
For：Gsc=(Vsc, E), midpoint for Imaginary Mechanism into small cluster, i.e. each virtual machine is initialized as one first
Small cluster is expressed as Vsc as a point in figurei=(c, b, d), at this time c simultaneously again represent cpu resource needed for small cluster, b
Represent bandwidth resources needed for small cluster again simultaneously, d represents the sum of small all side right values of cluster internal, and virtual machine i is initialized as small
Cluster i is initialized as small cluster j as the point i, virtual machine j in figure, as the point j in figure, if small cluster i and small cluster j it
Between there are a line e, then weights are expressed as wij=Evscij+EvscjiThe sum of uninterrupted between i.e. small cluster, if small cluster i
Side is not present between small cluster j, then wij=0, wijRepresent the weights on the side of tie point i and point j；
Step2：Side gain is calculated, side gain accounts for all side rights of institute's tie point to square directly proportional and side right value of side right value
The percentage of the sum of value is directly proportional, side tie point resource and is inversely proportional, and calculation formula is：
M=max { (Vsci(c)+Vscj(c))/C,(Vsci(b)+Vscj(b))/B}
C represents the memory of physical machine；B represents the bandwidth of physical machine；M represents that point i and point j occupies the hundred of physical machine resource after merging
Divide ratio, if physical machine resource is unsatisfactory for its resource requirement in total after two small cluster merges, △ Q are minimum value 0, eijAnd ai
Calculation formula be：
E in formulaijRepresent that the side right value of tie point i and point j put a spot the sum of i and point j internal edges weights, aiRepresent all sides of point i
The sum of weights；
Step3：It is more than 0 if there is side gain, then it is one towards physical machine to merge 2 points connect when gain is maximum
Small cluster；
Step4：Step2 is repeated, until all side gains are all not more than 0；
Step5：Merge remaining isolated node for a small cluster towards physical machine；
The big cluster topology division towards subregion is as follows：
Step1：The small cluster initialized towards physical machine is a big cluster towards subregion；Gvm=(V, E) is divided into face
The Gsc=(Vsc, E) obtained after the small cluster of physical machine, VsciRepresent small cluster i, WscijRepresent small cluster i and small cluster j
Between side weights, Vpci=(h, d) represents big cluster i, Vpci(h) number of the small cluster included in big cluster i is represented,
Vpci(d) the side right value between the small cluster that big cluster i internal edges weights include, Wpc are representedijRepresent big cluster i and great Ji
Weights between group j, N represent the number of the physical machine contained in a subregion；
Step2：Side gain is calculated, side gain accounts for all side rights of institute's tie point to square directly proportional and side right value of side right value
The percentage of the sum of value is directly proportional, side tie point resource and is inversely proportional, and calculation formula is：
M=(Vpci(h)+Vscj(h))/N
eijAnd aiMeet：
E in formulaijRepresent that the side right value of tie point i and point j put a spot the sum of i and point j internal edges weights； aiRepresent all sides of point i
The sum of weights, WpcijRepresent the weights on the side of tie point i and point j；N represents the number of physical machine in subregion；M represent point i with
Point j occupies physical machine number percentage in subregion after merging；Physical machine number is more than object in subregion after if two big cluster merges
Reason machine number, then △ Q are minimum value 0；
Step3：It is more than 0 if there is side gain, then merges 2 small clusters connect when gain is maximum for one towards dividing
The big cluster in area；
Step4：Step2 is repeated, until all side gains are all not more than 0；
Step5：Merge remaining isolated small cluster for a big cluster towards subregion；
Then, according to the Topology partition of virtual streaming media server position and 2 levels as a result, assigning towards the big of subregion
It the correspondence of cluster and subregion and assigns towards the small cluster of physical machine and the correspondence of physical machine；
Finally, moving method is cooperateed with by virtual streaming media cluster, all virtual machine (vm) migrations to be migrated to purpose physical machine
On, it is as follows：
Step1：Judge whether the destination host resource of virtual machine to be migrated meets, virtual machine to be migrated is divided into transportable queue
With etc. queue to be migrated；
Step2：The network gain that the virtual machine (vm) migration in transportable queue is brought is calculated, the network gain brought by its migration drops
Sequence arrange, virtual machine (vm) migration gain by after virtual machine (vm) migration reduction DCN inner stream flow size；
Step3：If transportable queue is sky, Step6 is gone to；Otherwise, virtual machine is taken in order from transportable queue, calculate
The migration path of virtual machine, migration path bandwidth are expressed as B；
Step4：If migration path bandwidth B is more than 1.5 times of virutal machine memory containing dirty pages generation rate, transition condition is unsatisfactory for, is turned
To Step3；Otherwise, its migration task is started；
Step5：Virtual machine completes migration, updates the migration gain for the virtual machine being attached thereto, updates the available of source physical host
Resource, the transportable queue of update and wait resource queue, update DCN network topology matrix available bandwidths；Go to Step3；
Step6：Update purpose physical host can with resource, update it is transportable and etc. queue to be migrated, update DCN network topology squares
Battle array available bandwidth；Go to Step3；
Step7：Whether the queues to be migrated such as judgement are empty, and if it is empty, then algorithm terminates；Otherwise, by virtual machine number to be moved out
Virtual machine (vm) migration in minimum physical machine is to free physical machine.
2. the virtual streaming media cluster collaboration moving method based on DCN according to claim 1, which is characterized in that the finger
Group is as follows towards the big cluster of subregion and the correspondence of subregion：
Step1：Internal storage data total amount formula Data (the i)=B*Vm copied during virtual machine (vm) migrationi(c)/(B-Vmi(b))
It represents, in formula：B represents transmission rate, Vmi(c) virutal machine memory size, Vm are representedi(b) represent that virutal machine memory containing dirty pages generate
Rate；
Step2：The distance virtual machine (vm) migration of virtual machine and subregion copies internal storage data total amount and leads to it to what subregion was passed through
Believe the product representation of interchanger number；
Step3：Towards subregion big cluster and subregion distance for all virtual machines for being included towards the big cluster of subregion with
The sum of subregion distance；
Step4：According to the distance of the big cluster towards subregion to subregion, optimum allocation side is acquired using Hungary's dispatching algorithm
Case.
3. the virtual streaming media cluster collaboration moving method according to claim 1 based on DCN, which is characterized in that described
Appointment is as follows towards the small cluster of physical machine and the correspondence of physical machine：
Step1：It is otherwise 1 if deploying virtual machine is 0 with physical machine distance in physical machine, virtual machine；
Step2：The distance of small cluster and physical machine towards physical machine is all void for being included towards the small cluster of physical machine
The sum of plan machine and physical machine distance；
Step3：According to the distance of the small cluster towards physical machine to physical machine, optimum allocation is acquired using Hungary's dispatching algorithm
Scheme.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510718803.5A CN105430049B (en) | 2015-10-29 | 2015-10-29 | A kind of virtual streaming media cluster collaboration moving method based on DCN |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510718803.5A CN105430049B (en) | 2015-10-29 | 2015-10-29 | A kind of virtual streaming media cluster collaboration moving method based on DCN |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105430049A CN105430049A (en) | 2016-03-23 |
| CN105430049B true CN105430049B (en) | 2018-05-18 |
Family
ID=55507986
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510718803.5A Active CN105430049B (en) | 2015-10-29 | 2015-10-29 | A kind of virtual streaming media cluster collaboration moving method based on DCN |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105430049B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107370679A (en) * | 2017-07-20 | 2017-11-21 | 郑州云海信息技术有限公司 | A kind of method and device for realizing migration process |
| CN109960564B (en) * | 2017-12-22 | 2023-04-11 | 上海盛霄云计算技术有限公司 | Virtual machine migration method |
| CN108541081B (en) * | 2018-04-17 | 2021-09-28 | 福建北峰通信科技股份有限公司 | Analog and digital DMR compatible talkback small cluster building method |
| CN110058970B (en) * | 2019-04-22 | 2022-10-18 | 重庆邮电大学 | Double-virtual-machine quasi-synchronous evacuation method under disaster risk model |
| CN111381934B (en) * | 2020-03-11 | 2023-07-21 | 重庆邮电大学 | Method for improving quick migration reliability of virtual machine under disaster risk model |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103559084A (en) * | 2013-10-17 | 2014-02-05 | 电子科技大学 | Virtual machine migration method of energy-saving data center |
| CN104008018A (en) * | 2014-05-22 | 2014-08-27 | 南京邮电大学 | Virtual machine online migration method in cloud computing environment |
| CN104901999A (en) * | 2015-03-31 | 2015-09-09 | 电子科技大学 | Method for reliability perceptive cross-domain mapping of virtual data center |
| CN104935628A (en) * | 2015-04-20 | 2015-09-23 | 电子科技大学 | Method of migrating a plurality of associated virtual machines among a plurality of data centers |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8478878B2 (en) * | 2010-03-11 | 2013-07-02 | International Business Machines Corporation | Placement of virtual machines based on server cost and network cost |
| US8667171B2 (en) * | 2010-05-28 | 2014-03-04 | Microsoft Corporation | Virtual data center allocation with bandwidth guarantees |
| US9444689B2 (en) * | 2013-06-03 | 2016-09-13 | Microsoft Technology Licensing, Llc | Dynamically migrating virtual machines and gateways |
-
2015
- 2015-10-29 CN CN201510718803.5A patent/CN105430049B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103559084A (en) * | 2013-10-17 | 2014-02-05 | 电子科技大学 | Virtual machine migration method of energy-saving data center |
| CN104008018A (en) * | 2014-05-22 | 2014-08-27 | 南京邮电大学 | Virtual machine online migration method in cloud computing environment |
| CN104901999A (en) * | 2015-03-31 | 2015-09-09 | 电子科技大学 | Method for reliability perceptive cross-domain mapping of virtual data center |
| CN104935628A (en) * | 2015-04-20 | 2015-09-23 | 电子科技大学 | Method of migrating a plurality of associated virtual machines among a plurality of data centers |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105430049A (en) | 2016-03-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105430049B (en) | A kind of virtual streaming media cluster collaboration moving method based on DCN | |
| CN105610715B (en) | A kind of cloud data center multi-dummy machine migration scheduling method of planning based on SDN | |
| CN113315700B (en) | Computing resource scheduling method, device and storage medium | |
| CN113193984B (en) | Air-space-ground integrated network resource mapping method and system | |
| WO2021237996A1 (en) | Fuzzy c-means-based adaptive energy consumption optimization vehicle clustering method | |
| US20120158858A1 (en) | Resource Optimization for Online Services | |
| WO2015105987A1 (en) | System and method for zoning in software defined networks | |
| CN103457852B (en) | A kind of survivability mapping method of multicast virtual network | |
| CN104104621B (en) | A kind of virtual network resource dynamic self-adapting adjusting method based on Nonlinear Dimension Reduction | |
| CN106936705B (en) | Software defined network routing method | |
| CN107196806B (en) | Topological proximity matching virtual network mapping method based on sub-graph radiation | |
| Cai et al. | Failure-resilient DAG task scheduling in edge computing | |
| CN108512772A (en) | Quality-of-service based data center's traffic scheduling method | |
| CN104935628A (en) | Method of migrating a plurality of associated virtual machines among a plurality of data centers | |
| CN106775949A (en) | A kind of Application of composite feature that perceives migrates optimization method online with the virtual machine of the network bandwidth | |
| CN109710406A (en) | Data distribution and its model training method, device and computing cluster | |
| CN106681815A (en) | Concurrent migration method of virtual machines | |
| CN106572170A (en) | Controller and dynamic load balancing method under SDN hierarchical multiple controllers | |
| CN105704054A (en) | Data center network flow migration method and system thereof | |
| CN104537682A (en) | Medical image segmenting and dispatching method | |
| CN108923979A (en) | Software defined network mapping method of virtual network | |
| CN106656555A (en) | Dynamic adjustment method of service resources of cloud computing system | |
| CN104539744A (en) | Two-stage media edge cloud scheduling method and two-stage media edge cloud scheduling device | |
| Hussain et al. | Clonal selection algorithm for energy minimization in software defined networks | |
| Singh et al. | Energy efficient load balancing hybrid priority assigned laxity algorithm in fog computing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |