CN105930202A - Migration policy for virtual machine with three thresholds - Google Patents
Migration policy for virtual machine with three thresholds Download PDFInfo
- Publication number
- CN105930202A CN105930202A CN201610291246.8A CN201610291246A CN105930202A CN 105930202 A CN105930202 A CN 105930202A CN 201610291246 A CN201610291246 A CN 201610291246A CN 105930202 A CN105930202 A CN 105930202A
- Authority
- CN
- China
- Prior art keywords
- virtual machine
- cycle
- main frame
- platform
- platform main
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/44—Arrangements for executing specific programs
- G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
- G06F9/45533—Hypervisors; Virtual machine monitors
- G06F9/45558—Hypervisor-specific management and integration aspects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/44—Arrangements for executing specific programs
- G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
- G06F9/45533—Hypervisors; Virtual machine monitors
- G06F9/45558—Hypervisor-specific management and integration aspects
- G06F2009/4557—Distribution of virtual machine instances; Migration and load balancing
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Hardware Redundancy (AREA)
Abstract
The present invention discloses a migration policy for a virtual machine with three thresholds. The migration policy is characterized in comprising: 1. using virtual machine activeness to quantify virtual machine load fluctuations, and using virtual machine activeness to determine whether the virtual machine is active; 2. traversing hosts in a cluster, and migrating heterogeneous virtual machines different from a host active mark in the hosts to an overload migration queue; 3. traversing hosts in the cluster, and migrating some virtual machines from the overload hosts to the overload migration queue until the hosts are no longer overloaded; 4. migrating the virtual machines in the overload migration queue to the suitable hosts in the cluster; 5. traversing hosts in the cluster, and migrating all virtual machines in the light-load hosts to a light-load migration queue; and 6. migrating the virtual machines in the light-load migration queue to the suitable hosts in the cluster. According to the migration policy disclosed by the present invention, in the case that virtual machine load fluctuations are intense in the cloud computing environment, the number of times of virtual machine migration between hosts in the cluster can be reduced, so that the quality of cloud services can be improved.
Description
Technical field
The present invention relates to field of cloud computer technology, the virtual machine (vm) migration strategy of a kind of three threshold values.
Background technology
Cloud computing is a kind of based on the Internet novel of generation after grid computing, parallel computation, Distributed Calculation
Computation schema.Intel Virtualization Technology be cloud computing dynamic resource adjust provide solution route (BARHAM P, DRAGOVIC B,
FRASER K, et al.Xen and the Art of Virtualization [J] .ACM, 2003 (08): 164-177.), it is
One of core technology that cloud computing is important.
Physical resource can be abstracted into a resource pool by Intel Virtualization Technology, and it is required that virtual machine obtains operation from resource pool
Resource.Use Intel Virtualization Technology can run multiple stage virtual machine in a physical machine simultaneously, this not only adds physics money
Utilization rate (NICOLAE B, the CAPPELLO F.A hybrid local storage transfer scheme in source
live migration of I/O intensive workloads[C].Proc of the 21st International
Symposium on High-performance Parallel and Distributed Computing.New York:ACM
Press, 2012), but also improve motility (the Ma Sugang.A review on of resource distribution in cloud computing environment
cloud computing development[J].Journal of Networks,2012,7(2):305-310.).Cloud computing
The Virtual Cluster built centrally through Intel Virtualization Technology, making the calculating resource dynamically organizing isomery obtain may (Mauro
Andreolini,Sara Casolari,Michele Colajanni.Dynamic load management of virtual
machines in a cloud architecture[J].Department of Information Engineering,
2010:201-204).Intel Virtualization Technology has reached its maturity, and is no lack of many revolutionary breakthroughs in its evolution, virtual
Machine migrating technology is exactly one of them.
Virtual machine migration technology is to be moved out from source host by virtual machine and migrate into another destination host, passes through virtual machine
Resource in virtual cluster can be carried out the most reasonable must allocating by migration.Virtual machine migration technology is moved from initial off-line
Shifting develops into online migration of today, and this makes virtual machine be held open during migrating, discontinuities will not be deployed on virtual machine
On application (CLARK C, FRASER K, HAND S, et al.Live migration ofvirtual machines [C] .Proc
of the 2nd Symposium on Networked Systems Design&Implementation.Berkeley:USENIX
Association, 2005:273-286.), whole transition process is transparent for cloud user.The online migrating technology of virtual machine is given
Cloud computing brings revolutionary breakthrough, and in Shi Yun data center, the dynamically management of resource may obtain.
In actual IaaS application, virtual machine is not unalterable in demand for resource the most in the same time.?
Super order in environment, when physical host is not enough to provide virtual machine resource required when being transferred to higher load condition by idle condition,
So this will reduce the quality of cloud service.Now it is accomplished by moving out a part of virtual machine to other main frame from overload main frame
In, to ensure the virtual machine demand for resource.
Existing dual threshold virtual machine (vm) migration strategy mainly triggers moving of virtual machine by setting upper threshold value with lower threshold value
Mobile work.When the load of main frame exceedes upper threshold value, a part of virtual machine of moving out from main frame is in other main frame in cluster
Go, reduce the risk violating SLA.When the load of main frame is less than lower threshold value, collection that all virtual machines in main frame are all moved out to
In other main frame in Qun, and by this Host Shutdown to save the energy.
Virtual machine load in main frame changes under more stable state, uses the virtual machine (vm) migration strategy of dual threshold not
Can often trigger the migration action of virtual machine.If but when the virtual machine fluctuation of load is violent, using dual threshold virtual machine (vm) migration plan
Slightly will often trigger the migration action of virtual machine, cause the Quality Down of cloud service.
Summary of the invention
The present invention is for avoiding the weak point existing for prior art, proposes the virtual machine (vm) migration plan of a kind of three threshold values
Slightly, to reducing in the case of in cloud computing environment, the virtual machine fluctuation of load is violent between virtual machine main frame in the cluster
Migration number of times so that cloud service quality is improved.
The present invention solves that technical problem adopts the following technical scheme that
The virtual machine (vm) migration strategy of a kind of three threshold values of the present invention, is for the migration between M platform main frame of the N platform virtual machine
Cheng Zhong, is characterized in:
M platform main frame is made to be designated as gathering PM={pm1,pm2,...,pmm,...,pmM, 1≤m≤M；pmmRepresent the m platform main frame in PM,
Make the m platform main frame pm in the T cyclemIt is designated asAnd
pcmRepresent m platform main frame pmmTotal CPU computing capability；Represent m platform main frame in the T cycleLoad, and Represent m platform main frame in the T cycleThe CPU computing capability used, and Represent m platform main frame in the T cycleIdle CPU computing capability, and Represent
M platform main frame in the T cycleEnliven labelling, andutm、ltmAnd mtmRepresent m platform main frame respectively
pmmUpper threshold value, lower threshold value and mobility threshold；Wherein utm=α；ltm=β；And α ＜ γ ＜ μ≤β；
Assume the m platform main frame in T cycleIn comprisePlatform virtual machine, is designated as virtual machine setAndBy described m in the T cycle
Platform main frameInPlatform virtual machineAbstract it is Represent m platform main frame pmmInPlatform virtual machineTotal CPU computing capability；Represent m in the T cycle
Platform main frameInPlatform virtual machineLoad, and Represent m platform master in the T cycle
MachineInPlatform virtual machineThe CPU computing capability used, and Represent m platform main frame in the T cycle?Platform virtual machineIdle CPU computing capability, and Represent m platform main frame in the T cycle?Platform virtual machineActive
Degree,WhereinS represents definite value, 1≤w≤s；Represent
Represent m platform main frame in the T cycle?Platform virtual machineEnliven labelling,
Wherein, at represents the liveness threshold value of virtual machine；
Described virtual machine (vm) migration strategy is to carry out as follows:
Step 1, define the T cycle overload migrate army unit be classified as L(T)；It is defined on m platform main frame in the T cycleInstitute
The overload used migrates queueInitialize T=1；Initialize m=1；
If step 2 m ＞ M sets up, then jump to step 7；Otherwise, initialize overload and migrate queueFor sky；Mark
In virtual machine setIn the m platform main frame enlivening labelling and T cycle of all virtual machinesEnliven labellingThe virtual machine differed；And by the virtual machine that marks from virtual machine setIn move out, put into overload migrate team
It is classified asIn；Thus update m platform main frame in the described T cycleThe overload used migrates queueAnd T
Cycle m platform main frameMiddle virtual machine set
If the m platform main frame in T cycle of step 3In virtual machine set after middle renewal, all virtual machines use
The summation of computing capability of CPU more than or equal to pcm×utmSet up, then jump to step 4；Otherwise jump to step 6；
Step 4, will update after virtual machine set in all virtual machines be ranked up according to load order from small to large,
Obtain the orderly virtual machine set after updating
Step 5, orderly virtual machine set after described renewalIn successively virtual machine is moved into update after mistake
Carry and migrate in queue, until the orderly virtual machine set after updatingIn the CPU computing capability of use of all virtual machines
Summation less than or equal to pcm×utmTill；Thus again update m platform main frame in the T cycleThe overload used is moved
Move queueAnd the T cycle m platform main frameMiddle virtual machine set
Step 6, m+1 is made to be assigned to m；Jump to step 2；
Step 7, make T cycle overload migrate army unit be classified asDescribed T
The overload in cycle migrates total queue L(T)In virtual machine quantity be designated asThe overload in described T cycle is migrated
Total queue L(T)In virtual machine be ranked up according to its CPU computing capability used order from big to small, thus the row of acquisition
The overload in T cycle after sequence migrates total queue Represent sequence
After T cycle overload migrate total queue L '(T)In i-th(T)Platform virtual machine；I-th(T)Platform virtual machineUse
CPU computing capability be designated asI-th(T)Platform virtual machineActive sign be designated as1≤i(T)≤I(T)；
Step 8, make i(T)=1；
If step 9 i(T)＞ I(T)Set up, then it represents that in the T cycle, the All hosts in host complexes does not all transship, by T
The host complexes all not transshipped in the individual cycle is designated asThe T cycle
In m platform do not transship main frame and be designated asIn the T cycle
M platform does not transships main frameIn virtual machine set be designated as
AndAnd jump to step 12；Otherwise, by M platform main frame
All hosts in PM is ranked up from big to small according to its CPU computing capability used, thus obtains in the T cycle i-th(T)M platform main frame after minor sort
Represent in the T cycle i-th(T)M platform main frame after minor sortIn m platform main frame；And
Step 10, make m=1；
Step 11, to sequence after T cycle overload migration total queue L '(T)In i-th(T)Platform virtual machineSentence
Break and whether meetAndAs met condition, then by i-th(T)Platform
Virtual machinePut into i-th(T)M platform main frame after minor sortIn, and make i(T)+ 1 is assigned to i(T)After, jump to
Step 9；As being unsatisfactory for condition, then, after making m+1 be assigned to m, jump to step 11；
Step 12, define the T cycle underloading migrate army unit be classified as L*(T), it is defined on m platform main frame in the T cycle
The underloading used migrates queueInitialize m=1；
If step 13 m ＞ M sets up, then jump to step 14；Otherwise, initialize underloading and migrate queueFor sky；If T
In the individual cycle, m platform does not transships main frameIn virtual machine setIn the calculating energy of CPU that uses of all virtual machines
The summation of power is less than or equal to ltm×pcm, then willIn virtual machine all migrate into underloading migrate queueIn, thus
Update the T cycle m platform main frameThe underloading used migrates queueMake m+1 be assigned to m, jump to step 13；
Step 14, make T cycle underloading migrate army unit be classified asDescribed
The underloading in T cycle migrates total queue L*(T)In virtual machine quantity be designated as I*(T)；The underloading in described T cycle is migrated total
Queue L*(T)In virtual machine be ranked up according to its CPU computing capability used order from big to small, thus the row of acquisition
The total queue of underloading in T cycle after sequenceAfter representing sequence
T cycle underloading migrate total queue L '*(T)In i-th*(T)Platform virtual machine；I-th*(T)Platform virtual machineUse
CPU computing capability be designated asI-th*(T)Platform virtual machineActive sign be designated as1≤i*(T)≤I*(T)；
Step 13, make i*(T)=1；
If step 14 i*(T)＞ I*(T)Set up, then it represents that the most non-underloading of the All hosts in host complexes in the T cycle,
And jump to step 17；Otherwise, the All hosts in M platform main frame PM is entered from big to small according to its CPU computing capability used
Row sequence, thus obtain in the T cycle i-th*(T)M platform main frame after minor sort Represent the T week
Interim i-th*(T)M platform main frame after minor sortIn m platform main frame；And
Step 15, make m=1；
Step 16, to sequence after T cycle underloading migration total queue L '*(T)In i-th*(T)Platform virtual machine
Judge whether to meetAndAs met condition, then by
i*(T)Platform virtual machinePut into i-th*(T)M platform main frame after minor sortIn, and make i*(T)+ 1 is assigned to i*(T)
After, jump to step 14；As being unsatisfactory for condition, then, after making m+1 be assigned to m, jump to step 16；
Step 17, make T+1 be assigned to T, and jump to step 2.
Compared with the prior art, the present invention has the beneficial effect that:
1, the present invention is by mobility threshold newly-increased on the basis of dual threshold virtual machine (vm) migration strategy, be used as by virtual from
Migrate queue move into main frame time judge main frame whether suitably condition such that it is able to what the reduction fluctuation of load was violent enlivens virtual machine
Make this main frame immediately reach the probability of mobility threshold after main frame of moving into, and then reduce the secondary of triggering virtual machine (vm) migration
Number, improves cloud service quality.
2, the present invention use virtual machine liveness model to the fluctuation of load situation quantifying in virtual machine life cycle, thus
Can intuitively reflect the fluctuation size that virtual machine loads.
3, the present invention is by treating the two class virtual machines distinguished by virtual machine liveness with a certain discrimination, by setting different moving
Move threshold value to reach increase deploying virtual machine density and reduce the effect that CPU calculates the waste of resource.
Accompanying drawing explanation
Fig. 1 is the inventive method flow chart.
Detailed description of the invention
In the present embodiment, the virtual machine (vm) migration strategy of a kind of three threshold values, for N platform virtual machine between M platform main frame
In transition process,
M platform main frame is made to be designated as gathering PM={pm1,pm2,...,pmm,...,pmM, 1≤m≤M；pmmRepresent the m platform master in PM
Machine, makes the m platform main frame pm in the T cyclemIt is designated asAnd
pcmRepresent m platform main frame pmmTotal CPU computing capability, such as main frame has 2 CPU, the computing capability of each CPU is
2000MIPS, then the CPU computing capability that this main frame is total is 4000；Represent m platform main frame in the T cycleLoad,
AndSuch as main frame CPU usage in current period is 50%, then this main frame load in current period
It is 0.5；Represent m platform main frame in the T cycleThe CPU computing capability used, and Represent m platform main frame in the T cycleIdle CPU computing capability, and Table
Show m platform main frame in the T cycleEnliven labelling, andIn the present embodiment, 0 expression inactive, 1
Represent active；utm、ltmAnd mtmRepresent m platform main frame pm respectivelymUpper threshold value, lower threshold value and mobility threshold；Wherein utm=α；
ltm=β；And α ＜ γ ＜ μ≤β, rule of thumb α is set to 0.2, β is set to 0.8, and γ is arranged
Being 0.6, μ is set to 0.8；
Assume the m platform main frame in T cycleIn comprisePlatform virtual machine, is designated as virtual machine setAndWill in the T cycle m platform
Main frameInPlatform virtual machineAbstract it is Represent m platform main frame pmmInPlatform virtual machineTotal CPU computing capability；Represent m platform in the T cycle
Main frameInPlatform virtual machineLoad, and Represent m platform main frame in the T cycleInPlatform virtual machineThe CPU computing capability used, and Represent m platform main frame in the T cycle?Platform virtual machineIdle CPU computing capability, and Represent m platform main frame in the T cycle?Platform virtual machineWork
Jerk,WhereinS represents definite value, 1≤w≤s,
In the present embodiment, s is set to 6, represents that the liveness of virtual machine is to be calculated by the loading condition in nearest 6 cycles
's；Represent m platform main frame in the T cycle?Platform virtual machineEnliven labelling,Wherein, at represents the liveness threshold value of virtual machine, in the present embodiment, at is set to 8, i.e.
The virtual machine liveness virtual machine more than 8 is marked as the virtual machine enlivened, virtual machine liveness less than 8 virtual machine then by
It is labeled as inactive virtual machine；
As it is shown in figure 1, virtual machine (vm) migration strategy is to carry out as follows:
Step 1, define the T cycle overload migrate army unit be classified as L(T)；It is defined on m platform main frame in the T cycleInstitute
The overload used migrates queueInitialize T=1；Initialize m=1；
If step 2 m ＞ M sets up, represent and had stepped through All hosts, and by the foreign peoples in current period main frame
Virtual machine adds each overload to and migrates in queue, then jump to step 7；Otherwise, initialize overload and migrate queueFor sky；Mark
Remember and in virtual machine setIn the m platform main frame enlivening labelling and T cycle of all virtual machinesActive mark
NoteThe virtual machine differed, referred to as foreign peoples's virtual machine；And by the virtual machine that marks from virtual machine setIn move
Going out, putting into overload migration queue isIn；Thus update m platform main frame in the T cycleThe overload used migrates queueAnd the T cycle m platform main frameMiddle virtual machine setSuch as in the T cycle, in m platform main frame
Have 4 virtual machines, be designated asIfFor foreign peoples's virtual machine, then by heterogeneous host
After moving outIt is updated toAnd transship migration queueIt is updated to
If the m platform main frame in T cycle of step 3In virtual machine set after middle renewal, all virtual machines use
The summation of computing capability of CPU more than or equal to pcm×utmSet up, represent that this main frame has transshipped, then jump to step 4；No
Then jump to step 6；
Step 4, will update after virtual machine set in all virtual machines be ranked up according to load order from small to large,
Obtain the orderly virtual machine set after updating
Step 5, from update after orderly virtual machine setIn successively virtual machine is moved into update after overload move
Move in queue, until the orderly virtual machine set after updatingIn the summation of CPU computing capability of use of all virtual machines
Less than or equal to pcm×utmTill；Thus again update m platform main frame in the T cycleThe overload used migrates queueAnd the T cycle m platform main frameMiddle virtual machine setSuch asIn have three orderly virtual
Machine, i.e.Overload migrates queueIf the CPU that three virtual machines use
The summation of computing capability more than or equal to pcm×utm, first willMove out, if the meter of the CPU of remaining two virtual machines use
The summation of calculation ability is less than pcm×utm, then update
Step 6, m+1 is made to be assigned to m；Jump to step 2；
Step 7, make T cycle overload migrate army unit be classified asThe T cycle
Overload migrate total queue L(T)In virtual machine quantity be designated asThe overload in T cycle is migrated total queue L(T)In virtual machine be ranked up according to its CPU computing capability used order from big to small, thus after obtaining sequence
The overload in T cycle migrates total queue Represent the T after sequence
The overload in individual cycle migrates total queue L '(T)In i-th(T)Platform virtual machine；I-th(T)Platform virtual machineThe CPU used meter
Calculation ability is designated asI-th(T)Platform virtual machineActive sign be designated as1≤i(T)≤I(T)；
Step 8, make i(T)=1；
If step 9 i(T)＞ I(T)Set up, then it represents that in the T cycle, the All hosts in host complexes does not all transship, by the
The host complexes all not transshipped in T cycle is designated asThe T cycle
In m platform do not transship main frame and be designated asIn the T cycle
M platform does not transships main frameIn virtual machine set be designated as
AndAnd jump to step 12；Otherwise, by M platform master
All hosts in machine PM is ranked up from big to small according to its CPU computing capability used, thus obtains in the T cycle
I-th(T)M platform main frame after minor sort Represent in the T cycle i-th(T)M platform main frame after minor sortIn m platform main frame；And
Step 10, make m=1；
Step 11, to sequence after T cycle overload migration total queue L '(T)In i-th(T)Platform virtual machineSentence
Break and whether meetAndAs met condition, then by i-th(T)Platform
Virtual machinePut into i-th(T)M platform main frame after minor sortIn, and make i(T)+ 1 is assigned to i(T)After, jump to
Step 9；As being unsatisfactory for condition, then, after making m+1 be assigned to m, jump to step 11；
Step 12, define the T cycle underloading migrate army unit be classified as L*(T), it is defined on m platform main frame in the T cycle
The underloading used migrates queueInitialize m=1；
If step 13 m ＞ M sets up, then jump to step 14；Otherwise, initialize underloading and migrate queueFor sky；If T
In the individual cycle, m platform does not transships main frameIn virtual machine setIn the calculating energy of CPU that uses of all virtual machines
The summation of power is less than or equal to ltm×pcm, then willIn virtual machine all migrate into underloading migrate queueIn, thus
Update the T cycle m platform main frameThe underloading used migrates queueMake m+1 be assigned to m, jump to step 13；
Step 14, make T cycle underloading migrate army unit be classified asT
The underloading in cycle migrates total queue L*(T)In virtual machine quantity be designated as I*(T)；The underloading in T cycle is migrated total queue L*(T)
In virtual machine be ranked up according to its CPU computing capability used order from big to small, thus obtain the T after sequence
The total queue of underloading in individual cycle Represent the T week after sequence
The underloading of phase migrates total queue L '*(T)In i-th*(T)Platform virtual machine；I-th*(T)Platform virtual machineThe CPU used calculate
Ability is designated asI-th*(T)Platform virtual machineActive sign be designated as1≤i*(T)≤I*(T)；
Step 13, make i*(T)=1；
If step 14 i*(T)＞ I*(T)Set up, then it represents that the most non-underloading of the All hosts in host complexes in the T cycle,
And jump to step 17；Otherwise, the All hosts in M platform main frame PM is entered from big to small according to its CPU computing capability used
Row sequence, thus obtain in the T cycle i-th*(T)M platform main frame after minor sort Represent the T week
Interim i-th*(T)M platform main frame after minor sortIn m platform main frame；And
Step 15, make m=1；
Step 16, to sequence after T cycle underloading migration total queue L '*(T)In i-th*(T)Platform virtual machine
Judge whether to meetAndAs met condition, then by
i*(T)Platform virtual machinePut into i-th*(T)M platform main frame after minor sortIn, and make i*(T)+ 1 is assigned to i*(T)
After, jump to step 14；As being unsatisfactory for condition, then, after making m+1 be assigned to m, jump to step 16；
Step 17, make T+1 be assigned to T, and jump to step 2.
In order to verify three threshold value virtual machine (vm) migration strategy validity, use Cloudsim to three threshold value virtual machine (vm) migration plans
Slightly carry out analogue simulation.1 data center is set up in experiment in the cloud environment that CloudSim builds, and has 500 in data center
Main frame.The CPU computing capability of every main frame is 3000MIPS.Data center there are 500 virtual machines run in main frame, every
The CPU computing capability of virtual machine is 1000MIPS.
Simulation experiment uses (Beloglazov A, Buyya R.Optimal Online Deterministic
Algorithms and Adaptive Heuristics for Energy and Performance Efficient
Dynamic Consolidation of Virtual Machines in Cloud Data Centers[J]
.Concurrency and Computation:Practice and c Experience,2012,24(13):1397-
1420.) data set of the virtual machine load announced in.Server in 1052 production environments of whole data set record,
The initial data of the cpu load that every 5 minute time intervals record in the operation of 240 hours.
Experiment carries out 5 times, and experimental result takes the meansigma methods of 5 experiments.Each experiment three test data set of use (D1,
D2, D3) three threshold value virtual machine (vm) migration strategies to be verified, each data are concentrated with 500 virtual machine load datas.Three numbers
According to collection three kinds of different virtual machine fluctuations of load of simulation (high, low, in) respectively.
Test practical dual threshold virtual machine (vm) migration strategy and three threshold value virtual machine (vm) migration policy virtual machine migrate number of times
Rate of violation with SLA.Whenever having a virtual machine to move out from main frame, virtual machine (vm) migration number of times adds one.SLA rate of violation is logical
The CPU computing capability crossing virtual machine request deducts the CPU that the difference of the CPU computing capability that main frame can provide is asked divided by virtual machine
Computing capability calculates.Experimental result is as shown in Table 1 and Table 2:
Table 1 virtual machine (vm) migration number of times compares
Table 2 SLA rate of violation compares
Virtual machine of can reading a book from Tables 1 and 2 uses the virtual machine (vm) migration strategy of three threshold values can efficiently reduce void
The migration number of times of plan machine and the rate of violation of SLA.
Claims (1)
1. a virtual machine (vm) migration strategy for three threshold values, is in N platform virtual machine transition process between M platform main frame, its
Feature is:
M platform main frame is made to be designated as gathering PM={pm1,pm2,...,pmm,...,pmM, 1≤m≤M；pmmRepresent the m platform main frame in PM, order
M platform main frame pm in the T cyclemIt is designated asAnd
pcmRepresent m platform main frame pmmTotal CPU computing capability；Represent m platform main frame in the T cycleLoad, and Represent m platform main frame in the T cycleThe CPU computing capability used, and Represent m platform main frame in the T cycleIdle CPU computing capability, and Table
Show m platform main frame in the T cycleEnliven labelling, andutm、ltmAnd mtmRepresent m platform respectively
Main frame pmmUpper threshold value, lower threshold value and mobility threshold；Wherein utm=α；ltm=β；And α ＜ γ ＜
μ≤β；
Assume the m platform main frame in T cycleIn comprisePlatform virtual machine, is designated as virtual machine set AndBy described m in the T cycle
Platform main frameInPlatform virtual machineAbstract it is Represent m platform main frame pmmInPlatform virtual machineTotal CPU computing capability；Represent m platform in the T cycle
Main frameInPlatform virtual machineLoad, and Represent m platform main frame in the T cycle
InPlatform virtual machineThe CPU computing capability used, and Represent m platform main frame in the T cycle?Platform virtual machineIdle CPU computing capability, and Represent m platform main frame in the T cycle?Platform virtual machineActive
Degree,WhereinS represents definite value, 1≤w≤s；Represent
Represent m platform main frame in the T cycle?Platform virtual machineEnliven labelling,
Wherein, at represents the liveness threshold value of virtual machine；
Described virtual machine (vm) migration strategy is to carry out as follows:
Step 1, define the T cycle overload migrate army unit be classified as L(T)；It is defined on m platform main frame in the T cycleUsed
Overload migrate queue beInitialize T=1；Initialize m=1；
If step 2 m ＞ M sets up, then jump to step 7；Otherwise, initialize overload and migrate queueFor sky；Mark virtual
Machine setIn the m platform main frame enlivening labelling and T cycle of all virtual machinesEnliven labellingNo
Identical virtual machine；And by the virtual machine that marks from virtual machine setIn move out, put into overload migrate queue be
In；Thus update m platform main frame in the described T cycleThe overload used migrates queueAnd the T cycle m
Platform main frameMiddle virtual machine set
If the m platform main frame in T cycle of step 3The CPU that in virtual machine set after middle renewal, all virtual machines use
The summation of computing capability more than or equal to pcm×utmSet up, then jump to step 4；Otherwise jump to step 6；
Step 4, will update after virtual machine set in all virtual machines be ranked up according to load order from small to large, it is thus achieved that
Orderly virtual machine set after renewal
Step 5, orderly virtual machine set after described renewalIn successively virtual machine is moved into update after overload move
Move in queue, until the orderly virtual machine set after updatingIn CPU computing capability total of use of all virtual machines
With less than or equal to pcm×utmTill；Thus again update m platform main frame in the T cycleThe overload used migrates team
RowAnd the T cycle m platform main frameMiddle virtual machine set
Step 6, m+1 is made to be assigned to m；Jump to step 2；
Step 7, make T cycle overload migrate army unit be classified asDescribed the T cycle
Overload migrate total queue L(T)In virtual machine quantity be designated asThe overload in described T cycle is migrated army unit
Row L(T)In virtual machine be ranked up according to its CPU computing capability used order from big to small, thus after obtaining sequence
T cycle overload migrate total queue After representing sequence
The overload in T cycle migrates total queue L′(T)In i-th(T)Platform virtual machine；I-th(T)Platform virtual machineThe CPU used
Computing capability is designated asI-th(T)Platform virtual machineActive sign be designated as1≤i(T)≤I(T)；
Step 8, make i(T)=1；
If step 9 i (T)＞ I(T)Set up, then it represents that in the T cycle, the All hosts in host complexes does not all transship, by T
The host complexes all not transshipped in the individual cycle is designated asThe T cycle
In m platform do not transship main frame and be designated asIn the T cycle
M platform does not transships main frameIn virtual machine set be designated as
AndAnd jump to step 12；Otherwise, by M platform master
All hosts in machine PM is ranked up from big to small according to its CPU computing capability used, thus obtains in the T cycle i-th(T)M platform main frame after minor sort
Represent in the T cycle i-th(T)M platform main frame after minor sortIn m platform main frame；And
Step 10, make m=1；
Step 11, to sequence after T cycle overload migration total queue L '(T)In i-th(T)Platform virtual machineJudgement is
No satisfiedAndAs met condition, then by i-th(T)Platform is virtual
MachinePut into i-th(T)M platform main frame after minor sortIn, and make i(T)+ 1 is assigned to i(T)After, jump to step
9；As being unsatisfactory for condition, then, after making m+1 be assigned to m, jump to step 11；
Step 12, define the T cycle underloading migrate army unit be classified as L*(T), it is defined on m platform main frame in the T cycleMade
Underloading migrate queue beInitialize m=1；
If step 13 m ＞ M sets up, then jump to step 14；Otherwise, initialize underloading and migrate queueFor sky；If the T week
Interim m platform does not transships main frameIn virtual machine setIn the computing capability of CPU that uses of all virtual machines
Summation is less than or equal to ltm×pcm, then willIn virtual machine all migrate into underloading migrate queueIn, thus update
The T cycle m platform main frameThe underloading used migrates queueMake m+1 be assigned to m, jump to step 13；
Step 14, make T cycle underloading migrate army unit be classified asDescribed T
The underloading in cycle migrates total queue L*(T)In virtual machine quantity be designated as I*(T)；The underloading in described T cycle is migrated total queue
L*(T)In virtual machine be ranked up according to its CPU computing capability used order from big to small, thus after obtaining sequence
The total queue of underloading in T cycle Represent the after sequence
The underloading in T cycle migrates total queue L '*(T)In i-th*(T)Platform virtual machine；I-th*(T)Platform virtual machineUsed
CPU computing capability is designated asI-th*(T)Platform virtual machineActive sign be designated as1≤i*(T)≤I*(T)；
Step 13, make i*(T)=1；
If step 14 i*(T)＞ I*(T)Set up, then it represents that the most non-underloading of the All hosts in host complexes in the T cycle, and jump
Go to step 17；Otherwise, the All hosts in M platform main frame PM is arranged from big to small according to its CPU computing capability used
Sequence, thus obtain in the T cycle i-th*(T)M platform main frame after minor sort Represent the T week
Interim i-th*(T)M platform main frame after minor sortIn m platform main frame；And
Step 15, make m=1；
Step 16, to sequence after T cycle underloading migration total queue L '*(T)In i-th*(T)Platform virtual machineJudge
Whether meetAndAs met condition, then by i-th*(T)
Platform virtual machinePut into i-th*(T)M platform main frame after minor sortIn, and make i*(T)+ 1 is assigned to i*(T)After, jump
Go to step 14；As being unsatisfactory for condition, then, after making m+1 be assigned to m, jump to step 16；
Step 17, make T+1 be assigned to T, and jump to step 2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610291246.8A CN105930202B (en) | 2016-04-29 | 2016-04-29 | A kind of virtual machine migration method of three threshold values |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610291246.8A CN105930202B (en) | 2016-04-29 | 2016-04-29 | A kind of virtual machine migration method of three threshold values |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105930202A true CN105930202A (en) | 2016-09-07 |
| CN105930202B CN105930202B (en) | 2019-03-08 |
Family
ID=56835198
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610291246.8A Active CN105930202B (en) | 2016-04-29 | 2016-04-29 | A kind of virtual machine migration method of three threshold values |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105930202B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106897137A (en) * | 2017-04-19 | 2017-06-27 | 南京大学 | A kind of physical machine based on live migration of virtual machine and virtual machine mapping conversion method |
| CN107423109A (en) * | 2017-05-24 | 2017-12-01 | 兰雨晴 | Virtual machine energy-saving scheduling method based on anonymous stochastic variable |
| CN108388471A (en) * | 2018-01-31 | 2018-08-10 | 山东汇贸电子口岸有限公司 | A kind of management method constraining empty machine migration based on double threshold |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104102523A (en) * | 2013-04-03 | 2014-10-15 | 华为技术有限公司 | Method for migrating virtual machine and resource scheduling platform |
| US9141420B2 (en) * | 2010-11-04 | 2015-09-22 | Alcatel Lucent | Overload control in a cloud computing environment |
| CN105159751A (en) * | 2015-09-17 | 2015-12-16 | 河海大学常州校区 | Energy-efficient virtual machine migration method in cloud data center |
| US20160116966A1 (en) * | 2008-11-20 | 2016-04-28 | International Business Machines Corporation | Method and apparatus for power-efficiency management in a virtualized cluster system |
-
2016
- 2016-04-29 CN CN201610291246.8A patent/CN105930202B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160116966A1 (en) * | 2008-11-20 | 2016-04-28 | International Business Machines Corporation | Method and apparatus for power-efficiency management in a virtualized cluster system |
| US9141420B2 (en) * | 2010-11-04 | 2015-09-22 | Alcatel Lucent | Overload control in a cloud computing environment |
| CN104102523A (en) * | 2013-04-03 | 2014-10-15 | 华为技术有限公司 | Method for migrating virtual machine and resource scheduling platform |
| CN105159751A (en) * | 2015-09-17 | 2015-12-16 | 河海大学常州校区 | Energy-efficient virtual machine migration method in cloud data center |
Non-Patent Citations (1)
| Title |
|---|
| 周文煜 等: "基于虚拟机迁移的虚拟机集群资源调度", 《华中科技大学学报(自然科学版)》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106897137A (en) * | 2017-04-19 | 2017-06-27 | 南京大学 | A kind of physical machine based on live migration of virtual machine and virtual machine mapping conversion method |
| CN106897137B (en) * | 2017-04-19 | 2020-05-19 | 南京大学 | Physical machine and virtual machine mapping conversion method based on virtual machine live migration |
| CN107423109A (en) * | 2017-05-24 | 2017-12-01 | 兰雨晴 | Virtual machine energy-saving scheduling method based on anonymous stochastic variable |
| CN107423109B (en) * | 2017-05-24 | 2020-05-01 | 兰雨晴 | Anonymous random variable-based virtual machine energy-saving scheduling method |
| CN108388471A (en) * | 2018-01-31 | 2018-08-10 | 山东汇贸电子口岸有限公司 | A kind of management method constraining empty machine migration based on double threshold |
| CN108388471B (en) * | 2018-01-31 | 2021-02-09 | 浪潮云信息技术股份公司 | Management method based on double-threshold constraint virtual machine migration |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105930202B (en) | 2019-03-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ibrahim et al. | Leen: Locality/fairness-aware key partitioning for mapreduce in the cloud | |
| Fernández-Cerero et al. | SCORE: Simulator for cloud optimization of resources and energy consumption | |
| Xie et al. | Improving mapreduce performance through data placement in heterogeneous hadoop clusters | |
| Chen et al. | Improving MapReduce performance using smart speculative execution strategy | |
| Lee et al. | A dynamic data placement strategy for hadoop in heterogeneous environments | |
| Rao et al. | Performance issues of heterogeneous hadoop clusters in cloud computing | |
| Xu et al. | Adaptive task scheduling strategy based on dynamic workload adjustment for heterogeneous Hadoop clusters | |
| Ibrahim et al. | Handling partitioning skew in mapreduce using leen | |
| Shabestari et al. | A taxonomy of software-based and hardware-based approaches for energy efficiency management in the Hadoop | |
| You et al. | A load-aware scheduler for MapReduce framework in heterogeneous cloud environments | |
| Diaz et al. | Energy-aware fast scheduling heuristics in heterogeneous computing systems | |
| Tang et al. | CPU–GPU utilization aware energy-efficient scheduling algorithm on heterogeneous computing systems | |
| Song et al. | Modulo based data placement algorithm for energy consumption optimization of MapReduce system | |
| CN103885811B (en) | Method, system and device that dummy machine system total system is migrated online | |
| Feng et al. | Energy efficiency for MapReduce workloads: An in-depth study | |
| Fan et al. | Intelligent resource scheduling based on locality principle in data center networks | |
| Fernández-Cerero et al. | Energy policies for data-center monolithic schedulers | |
| Yan et al. | Cost-efficient consolidating service for Aliyun’s cloud-scale computing | |
| Hu et al. | Improved heuristic job scheduling method to enhance throughput for big data analytics | |
| Han et al. | Marble: A multi-gpu aware job scheduler for deep learning on hpc systems | |
| Marszałkowski et al. | Time and energy performance of parallel systems with hierarchical memory | |
| CN105930202B (en) | A kind of virtual machine migration method of three threshold values | |
| Li et al. | Energy-aware task scheduling optimization with deep reinforcement learning for large-scale heterogeneous systems | |
| Bawankule et al. | Historical data based approach for straggler avoidance in a heterogeneous Hadoop cluster | |
| Miao et al. | Efficient flow-based scheduling for geo-distributed simulation tasks in collaborative edge and cloud environments |
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 |