CN106055425A - Method for cloud disaster recovery data backup based on game theory - Google Patents

Abstract

A method for cloud disaster recovery data backup based on a game theory relates to the technical field of data backup disaster recovery, and solves the problems in an existing data backup method that the resource cost is high, the storage cost is high, and influences of behaviors of source end cloud suppliers on a target end cloud supplier are neglected. The method for optimal data backup in a cloud disaster recovery environment based on the game theory provided by the invention is characterized in that interaction between the source end cloud supplier and the target end cloud supplier is simulated into a storage resource pricing model; and storage resource leading courses among backup nodes of the target end cloud supplier are modeled into a gaming course which pursues maximum benefits. The invention also provides a method for quantitative computation of optimal solutions of a storage resource quantity and a storage resource price, so that benefits of both gaming parties, namely the source end cloud supplier and the target end cloud supplier can be maximized simultaneously.

Description

Based on game theoretic cloud disaster tolerance data back up method

Technical field

The present invention relates to data backup disaster tolerance technology field, be specifically related to a kind of based on the optimum under game theoretic cloud environment Data back up method.

Background technology

Along with the development of information technology, the value of data is continuously increased, and people gradually strengthen for the dependence of network data, Especially finance or official's data, even the machine of delaying of the data degradation of fraction or short time all may be to the people Lives and properties cause harm greatly.Therefore, Data Disaster-Tolerance Technology receives and studies widely and pay close attention to.People would generally be not Same geographical position Backup Data copy is to reach data reliability requirement.But, traditional data store arranging method resource Spend huge.And the low cost of cloud storage, on-demand purchase and dynamic make it progressively become the optimum selection of data backup.

In recent years, cloud storage is widely used in data disaster tolerance with the pricing model of a kind of on-demand purchase.Working as Before data back up method in, lay particular emphasis on the parameter paying close attention to destination node more, such as storage cost and data recovery times etc. and right Memory capacity, minimum bandwidth and copy amount have carried out strict restriction, and have ignored the behavior of source cloud provider to purpose The impact of Duan Yun provider.

Game theory is the decision-making when behavior generation direct interaction of research decision-maker and this decision-making equal Weighing apparatus problem, has struggle or the mathematical theory of competition character phenomenon and method.Game theory considers prediction behavior individual in game And agenda, and study their optimisation strategy.In biology, economics, international relations, computer science, political science, army Thing strategy and other a lot of subjects are all widely used.

Summary of the invention

The present invention solves that available data backup method exists that resource expenditure is relatively big and carrying cost high, have ignored source simultaneously The behavior of the Duan Yun provider problem on the impact of destination cloud provider, it is provided that a kind of standby based on game theoretic cloud disaster tolerance data Part method.

Based on game theoretic cloud disaster tolerance data back up method, the method is realized by following steps:

Data are backed up on i source cloud provider SCPi by step one, user respectively, described source cloud provider SCP By the user data backup that self stores in destination cloud service provider DCP, described destination cloud service provider DCP bag Purpose memory node D_nodej containing j storage price change；I is the number of source cloud provider, holds cloud to provide for the purpose of j The number of the memory node of business；

Step 2, utility function U of calculating source cloud provider SCP_scpiUtility function with destination cloud provider DCP U_dcp；Specifically it is expressed as with following formula:

Formula one,

In formula, U_scpiRepresent the utility function of i-th source cloud provider SCPi, B_ijFor SCPi in destination cloud provider Interests acquired after storing resource on the jth purpose memory node D_nodej of DCP, described B_ijIt is expressed as with formula two:

Formula two,

In formula, b_iFor positive parameter, it is used for distinguishing each source cloud provider, t_jFor the positive parameter more than 1, it is used for distinguishing mesh The different purpose memory nodes of Duan Yun provider；x_ijStrategy for source cloud provider；C_ijFor i-th source cloud provider SCPi rents the cost that storage resource is spent from the jth purpose memory node D_nodej of destination cloud provider DCP, It is expressed as with formula three:

Formula three, C_ij=p_j·x_ij

In formula, p_jFor the purpose of the strategy of Duan Yun provider memory node, it may be assumed that j the purpose of destination cloud provider DCP is deposited The unit price of the storage resource on storage node D_nodej；

Described N_ijFor Network Load Balance, it is expressed as with formula four:

Formula four,

In formula, L_jRepresent that the maximum load on the jth purpose memory node D_nodej of destination cloud provider DCP is equal Weighing apparatus；

Formula two, formula three, formula four are substituted into formula one, it is thus achieved that formula five:

Formula five,

In formula, m is the positive integer more than j, and n is the positive integer more than i；

The utility function formula six of described destination cloud provider DCP is expressed as:

Formula six,

In formula, U_dcpThe utility function of the destination cloud provider DCP by being chosen, B_jDuan Yun provider DCP for the purpose of ' Jth purpose memory node D_nodej storage resource is leased to the income acquired in source cloud provider, with formula seven table Show:

Formula seven,

Described C_jFor the purpose of ', the consuming cost of the jth purpose memory node D_nodej of Duan Yun provider DCP, uses formula Eight are expressed as:

Formula eight,

In formula, p_j' represent the storage resource list on the jth purpose memory node D_nodej of destination cloud provider DCP Position consuming cost, described p_j' ＜ p_j；

By formula seven, formula eight substitutes into formula six, it is thus achieved that formula nine:

Formula nine,

Formula five and formula nine is used to calculate utility function U obtaining described source cloud provider SCP_scpiInitial value and institute State utility function U of destination cloud provider DCP_dcpInitial value；

Step 3, employing iterative algorithm realize Nash Equilibrium；

Particularly as follows: use iterative algorithm to calculate the tactful x of source cloud provider_ij；It is expressed as with formula ten:

Formula ten,

In formula,x_ij(τ₁) it is τ₁I-th source cloud after secondary iteration Provider SCPi plan is stored in the resource quantity of the jth purpose memory node D_nodej of destination cloud provider DCP, τ₁ Represent the iterations of source cloud provider, x_ij(τ₁+ 1) x is represented_ij(τ₁) next iteration result；δ is i-th source cloud Provider SCPi carries out step factor during resource quantity iteration；

Step 4, source cloud provider SCP revise self storage resource quantity, j the storage of destination cloud provider DCP Node D_nodej then constantly adjusts the list of storage resource according to the change of i source cloud provider SCPi storage resource quantity Position price p_j, method of adjustment formula 11 is expressed as:

Formula 11,

In formula,τ₂For the iterations of source cloud provider, p_j(τ₂) it is τ₂Destination cloud after secondary iteration The storage resource units price of the jth memory node D_nodej of provider DCP, p_j(τ₂+ 1) p is represented_j(τ₂) change next time For result；When for the purpose of θ, the jth memory node D_nodej of Duan Yun provider DSP carries out resource storage cell price iteration Step factor；

Step 5, the tactful x of the source cloud provider obtained according to step 3_ijThe destination cloud obtained with step 4 provides The tactful p of business's memory node_j, use formula nine to calculate utility function U of destination cloud provider DCP_dcpNew iteration knot Really, and judge U_dcpWhether it is maximum, if it is not, then return step 4；If it is, execution step 6；

Step 6, calculate utility function U of i-th source cloud provider SCPi according to formula five_scpi, and judge U_scpiIt is No for maximum, perform step 3 if it is not, then return, if it is, source cloud provider SCP and destination cloud provider DCP Reach Nash Equilibrium, it is thus achieved that optimum xij and pj value, it is achieved final Nash Equilibrium.

Beneficial effects of the present invention: of the present invention based on optimal data backup side under game theoretic cloud disaster tolerance environment Method, is modeled as a storage resource pricing model by the interaction between source cloud provider and destination cloud provider, and Storage resource lease process model building between destination cloud provider backup node is become a game seeking maximum return Journey.The present invention gives storage resource quantity and the method for storage resource price optimal solution quantitative Analysis simultaneously so that game is double Side, i.e. source cloud provider and destination cloud provider interests reach maximum simultaneously.

Accompanying drawing explanation

Fig. 1 is theory diagram based on game theoretic cloud disaster tolerance data back up method of the present invention；

Fig. 2 is flow chart based on game theoretic cloud disaster tolerance data back up method of the present invention；

Fig. 3 is of the present invention based in game theoretic cloud disaster tolerance data back up method three of destination cloud provider Memory node storage resource quantity change schematic diagram；

Fig. 4 is of the present invention based in game theoretic cloud disaster tolerance data back up method three of destination cloud provider Memory node storage resource price change schematic diagram；

Fig. 5 be of the present invention based on five source cloud providers in game theoretic cloud disaster tolerance data back up method respectively It is stored in the resource quantity distribution schematic diagram of three memory nodes of destination cloud provider；

Fig. 6 is of the present invention based on five source cloud provider effectiveness in game theoretic cloud disaster tolerance data back up method Function variation tendency schematic diagram；

Fig. 7 is of the present invention based on the effect of source cloud provider SCP1 in game theoretic cloud disaster tolerance data back up method With function variation tendency schematic diagram；

Fig. 8 is of the present invention to deposit based on destination cloud provider three in game theoretic cloud disaster tolerance data back up method Storage node utility function variation tendency schematic diagram；

Fig. 9 is of the present invention to deposit based on destination cloud provider three in game theoretic cloud disaster tolerance data back up method Storage node total utility function variation tendency schematic diagram.

Detailed description of the invention

Detailed description of the invention one, combine Fig. 1 and Fig. 2 and present embodiment is described, based on game theoretic cloud disaster tolerance data backup Method, the method is realized by following steps:

Data are backed up on different source cloud provider SCPi by A, user respectively, i=1, and 2 ... n, n represent available Source cloud provider quantity；For the safety and integrity demand of data and the consideration to economic factor, source cloud carries The user data backup self need to being had for business SCP to another cloud service provider, i.e. destination cloud provider DCP.Mesh Duan Yun provider DCP comprise the variable purpose memory node D_nodej of multiple storage price (j=1,2 ..., m, m represent mesh The available memory node quantity that comprised of Duan Yun provider DCP).Purpose memory node D_nodej is according to source cloud provider The size of the data volume that SCP is backed up dynamically adjusts storage price, constantly adjusts resource price and the number of resources of Backup Data Amount makes common interest maximize.

B, dynamic by between the purpose memory node D_nodej of source cloud provider SCP and destination cloud provider DCP Interaction is modeled as storing resource pricing model, and detailed process is as follows:

Source cloud provider SCP rents storage resource on the purpose memory node D_nodej of destination cloud provider DCP Time, it is desirable to obtain maximum interests: include the unit storage price lower than D_nodej present bidding, and higher network Load balancing.

The utility function formula one calculating described source cloud provider SCP is expressed as:

Formula one,

U in formula_scpiRepresent the utility function of i-th source cloud provider SCPi.B_ijRepresent that SCPi wishes at destination cloud Interests acquired after storing resource on the jth purpose memory node D_nodej of provider DCP, i.e. it is standby that SCPi completes data The expense that after Fen, user is paid, is expressed as with formula two:

Formula two,

B in formula_iIt is a positive parameter, in order to distinguish each source cloud provider.t_jIt is a positive parameter more than 1, uses To distinguish the different purpose memory nodes of destination cloud provider.x_ijRepresent the strategy of source cloud provider, i.e. i-th source cloud The storage resource that provider SCPi wants to rent on the jth purpose memory node D_nodej of destination cloud provider DCP Quantity.

C_ijRepresent the i-th source cloud provider SCPi jth purpose memory node D_ from destination cloud provider DCP Rent the cost that storage resource is spent on nodej, be expressed as with formula three:

Formula three, C_ij=p_j·x_ij(i=1,2,3,4,5 j=1,2,3)

P in formula_jRepresent the strategy of purpose memory node, i.e. the jth purpose memory node D_ of destination cloud provider DCP Storage resource units price on nodej.

N in formula one_ijRepresent Network Load Balance, formula four represent:

Formula four,

L in formula_jRepresent that the maximum load on the jth purpose memory node D_nodej of destination cloud provider DCP is equal Weighing apparatus, i.e. maximum storage capacity.

Formula two, formula three, formula four substitute into formula one can get formula five and be expressed as:

Formula five,

In present embodiment, the utility function formula six calculating described destination cloud provider DCP is expressed as:

Formula six,

U in formula_dcpThe utility function of the destination cloud provider DCP that expression is chosen, i.e. the maximum return that DCP is obtained, Including hiring out storage resource to the interests acquired in source cloud provider SCP and the cost spent.B_j' represent that destination cloud carries Storage resource is leased to the income acquired in source cloud provider by the jth purpose memory node D_nodej for business DCP, uses Formula seven represents:

Formula seven,

C in formula six_j' represent that the consumption of the jth purpose memory node D_nodej of destination cloud provider DCP becomes This, including maintenance cost, power consumption.Represent with formula eight:

Formula eight,

P in formula_j' represent the storage resource list on the jth purpose memory node D_nodej of destination cloud provider DCP Position consuming cost, p_j' ＜ p_j。

By formula seven, formula eight substitutes into formula six can obtain formula nine, as follows:

Formula nine

Utility function U of described source cloud provider SCP is calculated respectively with formula five and formula nine_scpiInitial value and described Utility function U of destination cloud provider DCP_dcpInitial value.

C, use iterative algorithm realize Nash Equilibrium.

τ₁Representing the iterations of source cloud provider, i-th source cloud provider SCPi uses the method for formula ten repeatedly Generation calculates wants to be stored in the resource quantity of the jth purpose memory node D_nodej of destination cloud provider DCP.

Formula ten,

In formulax_ij(τ₁) represent τ₁I-th source after secondary iteration Cloud provider SCPi plan is stored in the resource quantity of the jth purpose memory node D_nodej of destination cloud provider DCP, x_ij(τ₁+ 1) x is represented_ij(τ₁) next iteration result.δ ＞ 0, is a constant, represents that i-th source cloud provider SCPi enters Step factor during row resource quantity iteration.

D, while source cloud provider SCP constantly revises self storage resource quantity, destination cloud provider each Memory node D_nodej then constantly adjusts according to the change of all source cloud provider SCPi storage resource quantity and himself deposits The unit price of storage resource.Method of adjustment is carried out with the form iteration of formula 11.

Formula 11,

In formulaτ₂Represent the iterations of source cloud provider.

p_j(τ₂) represent τ₂The storage money of the jth memory node D_nodej of destination cloud provider DCP after secondary iteration Source unit price, p_j(τ₂+ 1) p is represented_j(τ₂) next iteration result.θ ＞ 0, is a constant, represents destination cloud provider The jth memory node D_nodej of DSP carries out step factor during resource storage cell price iteration.

E, it is stored in destination cloud provider according to the source cloud provider SCPi plan calculated in above-mentioned steps C Resource quantity x of the jth purpose memory node D_nodej of DCP_ijAnd the destination cloud provider calculated in step D Storage resource units price p of the jth memory node D_nodej of DCP_j, use method described in formula nine to calculate destination cloud Utility function U of provider DCP_dcpNew iteration result.Judge U_dcpWhether it is maximum, if it is, continue executing with step Rapid F, returns if not and performs step D；

F, calculate utility function U of i-th source cloud provider SCPi according to formula five_scpi, and judge U_scpiIt is whether Big value, performs step C if it is not, then return；If it is, be final Nash Equilibrium so that game participant both sides: source The benefit of cloud provider SCP and destination cloud provider DCP.

The data of xij size are backed up to depositing of destination cloud provider with price pj by G, source cloud provider SCPi respectively In storage node D_nodej.

Detailed description of the invention two, combining Fig. 1 to Fig. 9 present embodiment is described, present embodiment is detailed description of the invention one Described embodiment based on game theoretic cloud disaster tolerance data back up method:

Data are backed up to different source cloud provider SCP1 by a, user respectively, on SCP2, SCP3, SCP4, SCP5.In conjunction with Fig. 1, each source cloud provider SCP1, the user data backup that self need to be had by SCP2, SCP3, SCP4, SCP5 is to another Cloud service provider, i.e. destination cloud provider DCP.DCP comprises three variable purpose memory node D_ of storage price Node1, D_node2 and D_node3.Purpose memory node D_node1, D_node2, D_node3 are according to source cloud provider The size of the data volume that SCP1, SCP2, SCP3, SCP4, SCP5 are backed up dynamically adjusts storage price.

B, by source cloud provider SCP1, the purpose storage of SCP2, SCP3, SCP4, SCP5 and destination cloud provider DCP Node D_node1, the dynamic interaction process model building between D_node2, D_node3 is storage resource pricing model, detailed process As follows:

One, the utility function of described source cloud provider SCP is calculated:

$\begin{array}{cc}{U}_{scpi}=\underset{j=1}{\overset{3}{\Σ}}(B_{ij}-C_{ij}-N_{ij})& \left(\begin{array}{cc}i=1,2,3,4,5& j=1,2,3\end{array}\right)\end{array}---\left(1\right)$

U in formula_scpiRepresent the utility function of i-th source cloud provider SCPi.B_ijRepresent that SCPi wishes at destination cloud Interests acquired after storing resource on the jth purpose memory node D_nodej of provider DCP, i.e. it is standby that SCPi completes data The expense that after Fen, user is paid:

$\begin{array}{cc}{B}_{ij}=b_i\·{(1+x_{ij})}^{\frac{1}{t_j}}& \left(\begin{array}{cc}i=1,2,3,4,5& j=1,2,3\end{array}\right)\end{array}---\left(2\right)$

B in formula (2)_iIt is a positive parameter, in order to distinguish each source cloud provider, b in the present embodiment₁=2.5, b₂=2.6, b₃=2.7, b₄=2.8, b₅=2.9.t_jIt is a positive parameter more than 1, in order to distinguish destination cloud provider Different purpose memory nodes, t in the present embodiment₁=4, t₂=3, t₃=2.x_ijRepresent the strategy of source cloud provider, i.e. i-th Individual source cloud provider SCPi wants that rents on the jth purpose memory node D_nodej of destination cloud provider DCP to deposit The quantity of storage resource, initial value is 0.

C_ijRepresent the i-th source cloud provider SCPi jth purpose memory node D_ from destination cloud provider DCP Rent on nodej and store the cost that resource is spent:

C_ij=p_j·x_ij(i=1,2,3,4,5 j=1,2,3) (3)

P in formula (3)_jRepresent the strategy of purpose memory node, i.e. the jth purpose storage joint of destination cloud provider DCP Storage resource units price on some D_nodej, the most respectively p₁=0.1, p₂=0.2, p₃=0.3.

N in formula (1)_ijExpression Network Load Balance:

$\begin{array}{cc}{N}_{ij}=\frac{1}{L_j-{\mathrm{\Σ}}_{i=1}^5{x}_{ij}}& \left(\begin{array}{cc}i=1,2,3,4,5& j=1,2,3\end{array}\right)\end{array}---\left(4\right)$

L_jRepresent the maximum load equilibrium on the jth purpose memory node D_nodej of destination cloud provider DCP, i.e. Maximum storage capacity, the most respectively L₁=35, L₂=20, L₃=11.

Formula (2), formula (3), formula (4) are substituted into formula (1) and can obtain:

$\begin{array}{cc}{U}_{scpi}=\underset{j=1}{\overset{3}{\Σ}}(b_i\·{(1+x_{ij})}^{\frac{1}{t_j}}-p_j\·x_{ij}-\frac{1}{L_j-{\mathrm{\Σ}}_{i=1}^5{x}_{ij}})& \left(\begin{array}{cc}i=1,2,3,4,5& j=1,2,3\end{array}\right)\end{array}---\left(5\right)$

Two, the utility function of described destination cloud provider DCP is calculated:

$\begin{array}{cc}{U}_{dcp}=\underset{j=1}{\overset{3}{\Σ}}({B_j}^{\′}-{C_j}^{\′})& (j=1,2,3)\end{array}---\left(6\right)$

U in formula (6)_dcpRepresent the utility function of the destination cloud provider DCP chosen, i.e. the maximum receipts that DCP is obtained Benefit, including hiring out storage resource to the interests acquired in source cloud provider SCP and the cost spent.

B_j' represent that storage resource is leased to source by the jth purpose memory node D_nodej of destination cloud provider DCP Income acquired in Duan Yun provider:

$\begin{array}{cc}{B_j}^{\′}=\underset{i=1}{\overset{5}{\Σ}}{p}_j\·x_{ij}& \left(\begin{array}{cc}i=1,2,3,4,5& j=1,2,3\end{array}\right)\end{array}---\left(7\right)$

C in formula (6)_j' represent the consuming cost of jth purpose memory node D_nodej of destination cloud provider DCP, Including maintenance cost, power consumption, as shown in formula (8):

$\begin{array}{cc}{C_j}^{\′}=\underset{i=1}{\overset{5}{\Σ}}{p_j}^{\′}\·x_{ij}& \left(\begin{array}{cc}i=1,2,3,4,5& j=1,2,3\end{array}\right)\end{array}---\left(8\right)$

P in formula (8)_j' represent the storage resource on the jth purpose memory node D_nodej of destination cloud provider DCP Unit consumption cost, p_j' ＜ p_j, p in the present embodiment_j'=0.7p_j。

By formula (7), formula (8) substitutes into formula (6) and can obtain:

$\begin{array}{cc}{U}_{dcp}=\underset{j}{\overset{3}{\Σ}}\underset{i=1}{\overset{5}{\Σ}}(p_j\·x_{ij}-{p_j}^{\′}\·x_{ij})=\underset{j}{\overset{3}{\Σ}}\underset{i=1}{\overset{5}{\Σ}}\[\left(p_j-{p_j}^{\′}\right)\·x_{ij}\]& \left(\begin{array}{cc}i=1,2,3,4,5& j=1,2,3\end{array}\right)\end{array}---\left(9\right)$

Utility function U of described source cloud provider SCP is calculated respectively by formula (5) and formula (9)_scpiInitial value and described mesh Utility function U of Duan Yun provider DCP_dcpInitial value.

C, use iterative algorithm realize Nash Equilibrium.τ₁Representing the iterations of source cloud provider, i-th source cloud carries Want to be stored in the jth purpose memory node of destination cloud provider DCP with formula (10) method iterative computation for business SCPi The resource quantity of D_nodej.

$\begin{array}{cc}{x}_{ij}({\mathrm{\τ}}_1+1)=x_{ij}\left({\mathrm{\τ}}_1\right)+\mathrm{\δ}\·\[\frac{\∂U_{scpi}}{\∂x_{ij}}\]& \left(\begin{array}{cc}i=1,2,3,4,5& j=1,2,3\end{array}\right)\end{array}---\left(10\right)$

In formula

x_ij(τ₁) represent τ₁After secondary iteration, i-th source cloud provider SCPi plan is stored in destination cloud provider The resource quantity of the jth purpose memory node D_nodej of DCP, x_ij(τ₁+ 1) x is represented_ij(τ₁) next iteration result.δ ＞ 0, is a constant, and expression i-th source cloud provider SCPi carries out step factor during resource quantity iteration, at the present embodiment Middle δ=0.1.

D, while self storage resource quantity is constantly revised by source cloud provider, each storage of destination cloud provider Node D_nodej (j=1,2,3) is then according to the change of all sources cloud provider SCPi (i=1,2,3,4,5) storage resource quantity Change the unit price constantly adjusting himself storage resource.

$\begin{array}{cc}{p}_j({\mathrm{\τ}}_2+1)=p_j\left({\mathrm{\τ}}_2\right)+\mathrm{\θ}\·\[\frac{\∂U_{dcp}}{\∂p_j}\]& (j=1,2,3)\end{array}---\left(11\right)$

In formula (11)τ₂Represent the iterations of source cloud provider.

p_j(τ₂) represent τ₂The storage money of the jth memory node D_nodej of destination cloud provider DCP after secondary iteration Source unit price, p_j(τ₂+ 1) p is represented_j(τ₂) next iteration result.θ ＞ 0, is a constant, represents destination cloud provider The jth memory node D_nodej of DSP carries out step factor during resource storage cell price iteration, in the present embodiment θ= 0.001。

E, it is stored in destination cloud provider according to the source cloud provider SCPi plan calculated in above-mentioned steps three Resource quantity x of the jth purpose memory node D_nodej of DCP_ij, and the destination cloud provider calculated in step 4 Storage resource units price p of the jth memory node D_nodej of DCP_j, use method described in formula nine to calculate destination cloud Utility function U of provider DCP_dcpNew iteration result.Judge U_dcpWhether it is maximum, returns step d if not, if It is then to continue executing with step f.

F, it is stored in destination cloud provider according to the source cloud provider SCPi plan calculated in above-mentioned steps c Resource quantity x of the jth purpose memory node D_nodej of DCP_ijAnd the destination cloud provider calculated in step d Storage resource units price p of the jth memory node D_nodej of DCP_j, calculate i-th source cloud provider according to formula five Utility function U of SCPi_scpi, and judge U_scpiWhether it is maximum, if it is not, then return step c, if it is, combine Fig. 3 To Fig. 9, U_SCPiAnd U_dcpReaching maximum, game both sides reach Nash Equilibrium, it is thus achieved that optimum xij and pj value, are final receiving Assorted equilibrium so that game participant both sides: the benefit of source cloud provider SCP and destination cloud provider DCP.

The data of xij size are backed up by g, source cloud provider SCP1, SCP2, SCP3, SCP4, SCP5 respectively with price pj To memory node D_node1, D_node2, D_node3 of destination cloud provider.

Claims (3)

1. based on game theoretic cloud disaster tolerance data back up method, it is characterized in that, the method is realized by following steps:

Data are backed up on i source cloud provider SCPi by step one, user respectively, and described source cloud provider SCP will be from The user data backup of body storage is in destination cloud service provider DCP, and described destination cloud service provider DCP comprises j The purpose memory node D_nodej of individual storage price change；I is the number of source cloud provider, Duan Yun provider for the purpose of j The number of memory node；

Step 2, utility function U of calculating source cloud provider SCP_scpiUtility function U with destination cloud provider DCP_dcp； Specifically it is expressed as with following formula:

Formula one,

In formula, U_scpiRepresent the utility function of i-th source cloud provider SCPi, B_ijFor SCPi at destination cloud provider DCP Jth purpose memory node D_nodej on store interests acquired after resource, described B_ijIt is expressed as with formula two:

Formula two,

In formula, b_iFor positive parameter, it is used for distinguishing each source cloud provider, t_jFor the positive parameter more than 1, it is used for distinguishing destination The different purpose memory nodes of cloud provider；x_ijStrategy for source cloud provider；C_ijFor i-th source cloud provider SCPi The cost that storage resource is spent is rented, by public affairs from the jth purpose memory node D_nodej of destination cloud provider DCP Formula three is expressed as:

Formula three, C_ij=p_j·x_ij

In formula, p_jFor the purpose of the strategy of Duan Yun provider memory node, it may be assumed that j the purpose storage joint of destination cloud provider DCP The unit price of the storage resource on some D_nodej；

Described N_ijFor Network Load Balance, it is expressed as with formula four:

Formula four,

In formula, L_jRepresent the maximum load equilibrium on the jth purpose memory node D_nodej of destination cloud provider DCP；

Formula two, formula three, formula four are substituted into formula one, it is thus achieved that formula five:

Formula five,

In formula, m is the positive integer more than j, and n is the positive integer more than i；

The utility function formula six of described destination cloud provider DCP is expressed as:

Formula six,

In formula, U_dcpThe utility function of the destination cloud provider DCP by being chosen, B_jThe jth of Duan Yun provider DCP for the purpose of ' Storage resource is leased to the income acquired in source cloud provider by individual purpose memory node D_nodej, represents with formula seven:

Formula seven,

Described C_jThe consuming cost of the jth purpose memory node D_nodej of Duan Yun provider DCP for the purpose of ', with formula eight table It is shown as:

Formula eight,

In formula, p_j' represent that the storage resource units on the jth purpose memory node D_nodej of destination cloud provider DCP disappears Consumption cost, described p_j' ＜ p_j；

By formula seven, formula eight substitutes into formula six, it is thus achieved that formula nine:

Formula nine,

Formula five and formula nine is used to calculate utility function U obtaining described source cloud provider SCP_scpiInitial value and described mesh Utility function U of Duan Yun provider DCP_dcpInitial value；

Step 3, employing iterative algorithm realize Nash Equilibrium；

Particularly as follows: use iterative algorithm to calculate the tactful x of source cloud provider_ij；It is expressed as with formula ten:

Formula ten,

In formula,x_ij(τ₁) it is τ₁I-th source cloud provider after secondary iteration SCPi plan is stored in the resource quantity of the jth purpose memory node D_nodej of destination cloud provider DCP, τ₁Expression source The iterations of Duan Yun provider, x_ij(τ₁+ 1) x is represented_ij(τ₁) next iteration result；δ is i-th source cloud provider SCPi carries out step factor during resource quantity iteration；

Step 4, source cloud provider SCP revise self storage resource quantity, j the memory node of destination cloud provider DCP D_nodej then constantly adjusts the unit valency of storage resource according to the change of i source cloud provider SCPi storage resource quantity Lattice p_j, method of adjustment formula 11 is expressed as:

Formula 11,

In formula,τ₂For the iterations of source cloud provider, p_j(τ₂) it is τ₂After secondary iteration, destination cloud provides The storage resource units price of the jth memory node D_nodej of business DCP, p_j(τ₂+ 1) p is represented_j(τ₂) next iteration knot Really；Step-length when the jth memory node D_nodej of Duan Yun provider DSP carries out resource storage cell price iteration for the purpose of θ The factor；

Step 5, the tactful x of the source cloud provider obtained according to step 3_ijThe destination cloud provider obtained with step 4 deposits The tactful p of storage node_j, use formula nine to calculate utility function U of destination cloud provider DCP_dcpNew iteration result, and Judge U_dcpWhether it is maximum, if it is not, then return step 4；If it is, execution step 6；

Step 6, calculate utility function U of i-th source cloud provider SCPi according to formula five_scpi, and judge U_scpiIt is whether Maximum, performs step 3 if it is not, then return, if it is, source cloud provider SCP reaches with destination cloud provider DCP Nash Equilibrium, it is thus achieved that optimum xij value and pj value, it is achieved final Nash Equilibrium.

The most according to claim 1 based on game theoretic cloud disaster tolerance data back up method, it is characterised in that described purpose is deposited The size of the data volume that storage node D_nodej is backed up according to source cloud provider SCP dynamically adjusts storage price, by not The disconnected resource price adjusting Backup Data and resource quantity make source cloud provider SCP and destination cloud service provider DCP Benefit.

The most according to claim 1 and 2 based on game theoretic cloud disaster tolerance data back up method, it is characterised in that also to include Step 7, described source cloud provider SCP respectively by the data of the tactful xij of source cloud provider according to destination cloud provider The tactful pj of memory node backs up in the memory node D_nodej of destination cloud provider.