CN106648868A - Scheduling algorithm for providing continuous service by cloud system in power failure - Google Patents

Scheduling algorithm for providing continuous service by cloud system in power failure Download PDF

Info

Publication number
CN106648868A
CN106648868A CN201611190700.7A CN201611190700A CN106648868A CN 106648868 A CN106648868 A CN 106648868A CN 201611190700 A CN201611190700 A CN 201611190700A CN 106648868 A CN106648868 A CN 106648868A
Authority
CN
China
Prior art keywords
virtual machine
formula
machine
physical machine
income
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.)
Pending
Application number
CN201611190700.7A
Other languages
Chinese (zh)
Inventor
吴巍炜
何秀军
汪建平
罗军舟
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.)
Southeast University
Original Assignee
Southeast 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 Southeast University filed Critical Southeast University
Priority to CN201611190700.7A priority Critical patent/CN106648868A/en
Publication of CN106648868A publication Critical patent/CN106648868A/en
Pending legal-status Critical Current

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
    • G06F9/46—Multiprogramming arrangements
    • G06F9/48—Program initiating; Program switching, e.g. by interrupt
    • G06F9/4806—Task transfer initiation or dispatching
    • G06F9/4843—Task transfer initiation or dispatching by program, e.g. task dispatcher, supervisor, operating system
    • G06F9/485—Task life-cycle, e.g. stopping, restarting, resuming execution
    • G06F9/4856—Task life-cycle, e.g. stopping, restarting, resuming execution resumption being on a different machine, e.g. task migration, virtual machine migration
    • 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
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

The invention discloses a scheduling algorithm for providing continuous service by a cloud system in a power failure. The scheduling algorithm comprises the following steps: (1) initializing a formula (as shown in the specification); (2) when a formula (as shown in the specification) and a formula (as shown in the specification), calculating a formula (as shown in the specification), if a formula (as shown in the specification), executing a formula (as shown in the specification), activating a new PM to operate VMs in Ak, setting Sk=Sk-1UAK, k=k+1, U=U-t, and judging whether remaining energy is enough to continue cycling to activate a new PM or end or not; (3) calculating a formula (as shown in the specification), setting t to be greater than or equal to 0 and less than or equal to T, setting a formula (as shown in the specification), and if the gain the Sk is greater than a formula (as shown in the specification), returning the gain of the Sk; or else, returning the gain in a formula (as shown in the specification). The invention provides a solution which can approximate an optimal solution on average performance aiming at the problem of utilizing limited standby energy source in the situation of the power failure; the use efficiency of each unit power-on time when a physical machine is activated can be ensured, so that the gain of an operated virtual machine on the physical machine can be maximized; therefore, the use efficiency of the energy resource is ensured.

Description

Cloud system provides the dispatching algorithm of continuous service during a kind of power-off
Technical field
The present invention relates to cloud system technical field, especially cloud system provides the scheduling calculation of continuous service during a kind of power-off Method.
Background technology
In cloud system, keep service continuation it is critical that, power-off be one it is most common to continuation also most Serious threat.In order to improve the recovery capability of cloud during power-off, cloud service supplier would generally be promptly electric in data center deployment Power supply such as battery and generator, it is also possible to service is copied into other data centers when power-off occurs.But above-mentioned side Method all existing defects, the supply of energy is limited, and each service is in the service time of impacted data center's demand It is different, need efficient cloud service continuation dispatching algorithm to improve service revenue.
The content of the invention
The technical problem to be solved is, there is provided cloud system provides the scheduling of continuous service and calculates during a kind of power-off Method, it can be ensured that service efficiency during activation physical machine per unit conduction time, maximizes the virtual machine operated in physical machine Income.
To solve above-mentioned technical problem, cloud system provides the dispatching algorithm of continuous service when the present invention provides a kind of power-off, Comprise the steps：
(1) initializeK=1,
(2) whenAndWhen calculateIfThen carry out Ak=REVENUEBUNDLE (V, (J Sk-1)≤t, activate new physical machine PM to run AkIn void Plan machine VMs, makes Sk=Sk-1∪Ak, k=k+1,Judge whether that dump energy is sufficient to continue cycling through activation new Platform PM or end；
(3) calculateOrder If SkIn income ratioGreatly, then S is returnedkIncome；Otherwise, returnIn income；Wherein, J is before data center's power-off The set of the virtual machine of operation；V is the number of resources of single physical machine PM；sjCpu resources required for jth platform virtual machine are big It is little；djFor the deadline of jth platform virtual machine；J≤tTo meet d in virtual machine set JjThe set of≤t；pjFor jth platform virtual machine The operation before deadline completes brought profit；T is the time of data center's service restoration；I is physics in data center The set of machine；U is operational emergency service total amount after power-off in data center；AkIt is to be allocated in kth wheel iteration Virtual machine subset；SkFor virtual machine set allocated in front k wheels, i.e. Sk=∪1≤i≤kAi；In for algorithm, for satisfaction is given In determining the remaining unassigned virtual machine subset under deadline, meet the given resource capacity upper limit and make obtained income Maximized subclass；For in algorithm, in the unit intervalThe deadline of middle maximum revenue；It is in capacity V Distribute physical machine collection under lower single physical machineWhen getable maximum return；It is under capacity V Distribute physical machine collection under single physical machineWhen reach the physical machine set of maximum return.
Beneficial effects of the present invention are：For the problem that limited stand-by power source how is utilized under powering-off state, there is provided one The individual solution that optimal solution can be approached in average behavior, it can be ensured that making per unit conduction time during activation physical machine With efficiency, the income of the virtual machine operated in physical machine is maximized, so as to ensure the service efficiency of the energy.
Description of the drawings
The structural representation of cloud system when Fig. 1 is the power-off of the present invention.
Fig. 2 is the assessment mode schematic diagram the time required to continuous service.
Fig. 3 is the algorithm flow schematic diagram of the present invention.
Fig. 4 (a) is that the virtual machine quantity in obtainable physical machine with conduction time in the case of limited of the present invention is different right In the schematic diagram that financial value affects.
Fig. 4 (b) be the present invention different physical machines be availability status when income change schematic diagram.
Specific embodiment
As shown in figure 1, during a kind of power-off cloud system structure, local data center possesses physical machine, physical machine and urgent Non-firm power (UPS etc.), remote data center (public cloud etc.) provides standby virtual machine service.When disaster or powering-off state occur When, local data center is estimated to local resource number and virtual machine continuation required time, income, and operation cloud system continues Service dispatch algorithm, by migrating or being redirected to remote data center service continuity is realized.Wherein, root the time required to service It is estimated according to different situations.
As shown in Fig. 2 under powering-off state, virtual machine maintains to be commented by following different situations the time required to continuation Estimate：Determine the time required to locally being migrated by virtual machine the time required to when selecting to have locally executed with tasks carrying；When need weight When being directed to remote data center, required time is determined by the time of redirection if the existing corresponding virtual machine in distal end, and such as Fruit distal end without corresponding virtual machine, then by distinguishing newly-built virtual machine, building virtual machine and migrating data, the transplanting seed of virtual machine three The time required to situation determines.
As shown in figure 3, cloud system provides the dispatching algorithm of continuous service during a kind of power-off, comprise the steps：
(1) initializeK=1,
(2) whenAndWhen calculateIfThen carry out Ak=REVENUEBUNDLE (V, (J Sk-1)≤t, activate new physical machine PM to run AkIn void Plan machine VMs, makes Sk=Sk-1∪Ak, k=k+1,Judge whether that dump energy is sufficient to continue cycling through activation new Platform PM or end；
(3) calculateOrder If SkIn income ratioGreatly, then S is returnedkIncome；Otherwise, returnIn income；Wherein, J is before data center's power-off The set of the virtual machine of operation；V is the number of resources of single physical machine PM；sjCpu resources required for jth platform virtual machine are big It is little；djFor the deadline of jth platform virtual machine；J≤tTo meet d in virtual machine set JjThe set of≤t；pjFor jth platform virtual machine The operation before deadline completes brought profit；T is the time of data center's service restoration；I is physics in data center The set of machine；U is operational emergency service total amount after power-off in data center；AkIt is to be allocated in kth wheel iteration Virtual machine subset；SkFor virtual machine set allocated in front k wheels, i.e. Sk=U1≤i≤kAi；In for algorithm, for satisfaction is given In determining the remaining unassigned virtual machine subset under deadline, meet the given resource capacity upper limit and make obtained income most The subclass of bigization；For in algorithm, in the unit intervalThe deadline of middle maximum revenue；It is under capacity V Distribute physical machine collection under single physical machineWhen getable maximum return；It is to place an order in capacity V Distribute physical machine collection under individual physical machineWhen reach the physical machine set of maximum return.
The conduction time of t unit is suppose there is, [0, t) select multiple virtual machines to be V in single capacity in interval from J Physical machine on run, definitionIt is the set of the virtual machine that final term meets dj≤t, only final term is little Can run in such a physical machine in the virtual machine of t.Therefore it may only be necessary to consider J≤t.
From the set for having obtainedThe subset of a virtual machine is found, the set meets total revenue in selected virtual machine Total resources demand is maximized under the restrictive condition less than V.A virtual machine set has been obtainedWith capacity V, program is made procedureThe subset of return demand, whileFor maximum return.DefinitionIt is the maximum return that front j virtual machine is only used under the premise of v capacity.IncomeWhen j-th virtual machine Without it is selected when be equal toOr be equal to when j-th virtual machine is selectedThus, recursive function is：
By compilingPerformThe time of compiling rev (n, V, J) is complicated Degree is O (nV).Once need, can further by by it zoom to δ V remove value V time complexity dependence come Income at least (1- δ) rev (n, V, J) is reached, is used under conditions of error range is δRun time.
In order to solve this crucial subproblem, it is only necessary to caller REVENUEBUNDLE (V, J≤t), J here≤tIt is Deadline meets djThe virtual machine set of≤t.
The CSC of a large amount of physical machines is solved the problems, such as on the basis of such scheme.One physical machine of activation is multiple to run Virtual machine, so efficiently can meet using resource capacity and first the service efficiency of conduction time, then meet logical Virtual machine is iteratively activated under conditions of electric time-constrain.
With the help of having REVENUEBUNDLE algorithms, it is known that interval [0, t), calculate maximum return rev (V, J≤t)。 First virtual machine is issued to maximum in all possible t ∈ [0, T] conditionNext, this algorithm iteration ground Find remaining virtual machine.
The specific design of this algorithm has in BRP and represents.This algorithm iteration ground selects a subset of virtual machine, and is They activate a new physical machine.If AkIt is to distribute in iteration k and Sk=∪1≤i≤kAiOn virtual machine subset.In iteration k In, algorithm caller REVENUEBUNDLE (V, (and J Sk-1)≤t) testing each time t ∈ [1, T] to find when one BetweenAnd set AkFor the virtual machine that t meets rev (V, (J sk-1)≤t).It is activated One new physical machine consumes t time quantums to obtain income distributing the virtual machine of AkThis calculation Method iteratively finds ensuing set Sk+1And be all chosen in all of virtual machine or all of time increased set Sk+1After reached and terminate in the case of capacity U.When such case occurs, this algorithm is returned bigger SkWithBetween income Set.NowIt is the accessible maximum receipts in the case where having J virtual machine input, U time quantum, a physical machine Benefit.
Because the optimal solution for calculating cloud service persistent problem CSC is NP-hard, we claim dispatching algorithm of the present invention Be CSC-SCHEDULE execution and simple greed, the upper bound of optimal solution be compared.Optimal solution is expressed as OPT UB, and this is By the way that ILP (integer programming) to be loosened into what the LP (linear programming) of xij ∈ [0,1] was obtained.The assessment of inventive algorithm is as follows It is shown.
It is different that Fig. 4 (a) illustrates the virtual machine quantity in the case of being limited in obtainable physical machine and conduction time For the impact of financial value.In this case, the quantity n ∈ [50,500] of virtual machine, quantity m=100 of physical machine is powered Time U=5000.
As a rule, these three incomes for planting algorithm all increase with the growth of the quantity of virtual machine.When virtual machine number When amount is less than 200, the difference of three a variety of algorithms is inappreciable.After this, the income of inventive algorithm keep it is stable because Bottleneck is in for conduction time.On average, inventive algorithm is more better than greedy algorithm, and income is at most OPT UB's 0.41 times.
Fig. 4 (b) elaborates the change of the income when different physical machines is availability status.Physical machine quantity [50, 200] change between, virtual machine number is fixed as 500.On average, inventive algorithm is more better than greedy algorithm, and income is extremely forced Nearly optimal solution.
Although the present invention is illustrated with regard to preferred embodiment and has been described, it is understood by those skilled in the art that Without departing from scope defined by the claims of the present invention, variations and modifications can be carried out to the present invention.

Claims (1)

1. cloud system provides the dispatching algorithm of continuous service during a kind of power-off, it is characterised in that comprise the steps：
(1) initializeK=1,
(2) whenAndWhen calculateIfThen Carry out Ak=REVENUEBUNDLE (V, (J Sk-1)≤t, activate new physical machine PM to run AkIn virtual machine VMs, order Sk=Sk-1∪Ak, k=k+1,Judge whether that dump energy is sufficient to continue cycling through the new PM of activation or end；
(3) calculateOrder
If SkIn income ratioGreatly, then S is returnedkIncome；Otherwise, ReturnIn income；Wherein, J is the set of the virtual machine run before data center's power-off；V is the resource of single physical machine PM Number；sjCpu resource sizes required for jth platform virtual machine；djFor the deadline of jth platform virtual machine；J≤tFor virtual machine collection Close in J and meet djThe set of≤t；pjThe profit for completing to bring is run before deadline by jth platform virtual machine；T is in data The time of heart service restoration；I is the set of physical machine in data center；U is operational urgent after power-off in data center Power supply total amount；AkIt is the virtual machine subset being allocated in kth wheel iteration；SkAllocated virtual machine set in taking turns for front k, That is Sk=∪1≤i≤kAi；In for algorithm, for satisfaction is given in the remaining unassigned virtual machine subset under deadline, Meet the given resource capacity upper limit and make the subclass of obtained maximum revenue；For in algorithm, in the unit intervalMiddle income Maximized deadline；It is the distribution physical machine collection under single physical machine under capacity VWhen getable maximum receive Benefit；It is the distribution physical machine collection under single physical machine under capacity VWhen reach the thing of maximum return Reason machine set.
CN201611190700.7A 2016-12-21 2016-12-21 Scheduling algorithm for providing continuous service by cloud system in power failure Pending CN106648868A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201611190700.7A CN106648868A (en) 2016-12-21 2016-12-21 Scheduling algorithm for providing continuous service by cloud system in power failure

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201611190700.7A CN106648868A (en) 2016-12-21 2016-12-21 Scheduling algorithm for providing continuous service by cloud system in power failure

Publications (1)

Publication Number Publication Date
CN106648868A true CN106648868A (en) 2017-05-10

Family

ID=58834441

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201611190700.7A Pending CN106648868A (en) 2016-12-21 2016-12-21 Scheduling algorithm for providing continuous service by cloud system in power failure

Country Status (1)

Country Link
CN (1) CN106648868A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109614198A (en) * 2018-11-26 2019-04-12 东南大学 A kind of virtual machine under electricity price dynamic change environment integrates dispatching algorithm

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109614198A (en) * 2018-11-26 2019-04-12 东南大学 A kind of virtual machine under electricity price dynamic change environment integrates dispatching algorithm

Similar Documents

Publication Publication Date Title
CN109324875B (en) Data center server power consumption management and optimization method based on reinforcement learning
Wolke et al. More than bin packing: Dynamic resource allocation strategies in cloud data centers
US11126242B2 (en) Time varying power management within datacenters
CN102103516B (en) Virtual-CPU based frequency and voltage scaling
US8402140B2 (en) Methods and apparatus for coordinated energy management in virtualized data centers
CN103885831B (en) The system of selection of virtual machine host machine and device
Li et al. Opportunistic scheduling in clouds partially powered by green energy
US20110307110A1 (en) Management of a virtual power infrastructure
CN106020934A (en) Optimized deploying method based on virtual cluster online migration
CN104854563A (en) Automated profiling of resource usage
CN104657215A (en) Virtualization energy-saving system in Cloud computing
CA2799985A1 (en) Leveraging smart-meters for initiating application migration across clouds for performance and power-expenditure trade-offs
Sampaio et al. Towards high-available and energy-efficient virtual computing environments in the cloud
CN105373432B (en) A kind of cloud computing resource scheduling method based on virtual resource status predication
Khan et al. Energy, performance and cost efficient cloud datacentres: A survey
CN103823718A (en) Resource allocation method oriented to green cloud computing
CA3189144A1 (en) Power aware scheduling
Yang et al. Online learning-based server provisioning for electricity cost reduction in data center
CN103888371A (en) Method and device for virtual network mapping
CN107203255A (en) Power-economizing method and device are migrated in a kind of network function virtualized environment
CN102541622B (en) Method for placing load-related virtual machine
Lin et al. Novel resource allocation model and algorithms for cloud computing
CN107479949A (en) Low energy consumption cloud computing method for scheduling task
Chamas et al. Two-phase virtual machine placement algorithms for cloud computing: An experimental evaluation under uncertainty
CN106648868A (en) Scheduling algorithm for providing continuous service by cloud system in power failure

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20170510

RJ01 Rejection of invention patent application after publication