CN106304308A - A kind of multi-service deposit system medium cloud business energy optimization dispatching method - Google Patents

Abstract

nullThe present invention relates to a kind of multi-service deposit system medium cloud business energy optimization dispatching method，Establish one can optimize simultaneously multiple cloud business energy consumption and can maximum system throughput frequency spectrum resource distribution double-goal optimal model，Introduce cloud business and upload the higher limit of energy，The minimum target that cloud business is uploaded energy is rewritten as the energy restrictive condition less than certain threshold value，Biobjective scheduling problem is made to become single-object problem，Then by the method for backward iteration，Obtain optimum energy consumption and relation between the channel set that current time slots is assigned to，And by this relation，Throughput of system will be optimized and optimize in cloud traffic energy the two target unification to time scale，The last distribution being carried out channel again by designed 01 integer programming algorithms or Lagrange duality algorithm，Maximum system throughput on the premise of meeting cloud traffic energy consumption requirement.

Description

A kind of multi-service deposit system medium cloud business energy optimization dispatching method

Technical field

The present invention relates to a kind of multi-service deposit system medium cloud business energy optimization dispatching method, belong to wireless cloud and calculate skill Art and Radio Resource optimisation technique field.

Background technology

The swift and violent growth of internet traffic and constantly bringing forth new ideas of application, promoted the development of mobile cloud computing technology.Wireless Application data are uploaded to high in the clouds by wireless network and process by cloud business, reduce the computing capability to mobile terminal and want Ask, improve Business Processing efficiency simultaneously.But, compared to flourish wireless application, the battery capacity of mobile terminal is not How a foot always bottleneck that cannot break through, therefore, reduce energy expenditure when terminal uploads cloud business datum and gradually become For people's focus of attention.

It is exactly basically through-put power accumulation in time that terminal uploads the energy expenditure of cloud business datum.Transmission Power is relevant with transfer rate and channel status.When uploading data total amount and being constant, the biggest then through-put power of transfer rate is the biggest, The transmission time is the fewest.Therefore, the optimization to energy expenditure is the scheduling to transfer rate actually.Existing research is along above-mentioned think of Many effort have been done in terms of reducing cloud traffic energy consumption in road.Successively propose under single channel and multichannel scene, optimize Cloud business upload energy, obtain the rate scheduling scheme of optimum, and find cloud business upload energy along with the increasing of the number of channel Add and reduce.

But, existing all documents are only for having the situation of single cloud business in system, and real system medium cloud business Uploading rate except obeying in addition to " energetic optimum " this scheduling strategy, the spectral band that cloud business is occupied certainly will be limited by Width, and the bandwidth that business is occupied is inevitable the most relevant with other business in system.Therefore, cloud traffic scheduling strategy is placed on many Consider under the scene that business coexists just to tally with the actual situation, there is presently no the research of this respect.

Summary of the invention

The technical problem to be solved is to provide a kind of in consideration general service QoS and the situation of throughput of system Under, set up resource distribution and cloud service rate dispatches double-goal optimal model, respectively obtain cloud business by optimized algorithm optimum Rate scheduling strategy and the multi-service of channel optimal distribution strategy deposit system medium cloud business energy optimization dispatching method.

The present invention is to solve above-mentioned technical problem by the following technical solutions: the present invention devises a kind of multi-service and deposits System medium cloud business energy optimization dispatching method, in the system that general service and cloud business coexist, it is achieved the power consumption of cloud business is excellent Change scheduling, comprise the steps:

Step A. is respectively directed to each general service and each cloud business, uses shannon formula to obtain transfer rate and transmission Relation between power, wherein, channel gain obeys independent same distribution, and then obtains each service transmission rate and transmission respectively Relation between power, subsequently into step B；

Step B. is respectively directed to each cloud business, according to the relation between service transmission rate and through-put power, it is thus achieved that base In service transmission rate, bandwidth, the cloud business energy consumption model of channel gain, and then obtain the cloud business energy of each cloud business respectively Consumption model, subsequently into step C；

Step C. is respectively directed to each cloud business, for each working time slot corresponding to cloud business, first, according to cloud industry The cloud business energy consumption model of business, uses dynamic programming method, it is thus achieved that cloud business its last working time slot corresponding chronologically Cloud business time-slot energy consumption cost function, then, use contrary recurrence method, according to the channel gain in work at present time slot, Obtain the cloud business corresponding cloud business time-slot energy consumption cost function of each working time slot before it chronologically successively；And then obtain The cloud business time-slot energy consumption cost function of each cloud business its each working time slot the most corresponding, subsequently into step D；

Step D. is respectively directed to each cloud business, for the cloud business time-slot energy consumption of each working time slot corresponding to cloud business Cost function, it is thus achieved that the optimal rate of cloud business respectively each working time slot corresponding, and then obtain corresponding to cloud business Little overall situation energy consumption model, thus, it is thus achieved that the minimum overall situation energy consumption model corresponding to each cloud business, subsequently into step E1；

Step E1. is according to current time slots channel quantity to be allocated, it is thus achieved that for current time slots each cloud business to be accessed, The all channel assignment scheme of each general service, subsequently into step E2；

Step E2. is respectively directed to each channel assignment scheme of current time slots, according to the minimum corresponding to each cloud business Overall situation energy consumption model, it is thus achieved that under channel assignment scheme, current time slots accessed the minimum overall energy consumption of each cloud business, obtain simultaneously Obtain under channel assignment scheme, current time slots is accessed each general service transfer rate sum, i.e. system current time-slot throughput； And then obtaining under each channel assignment scheme, current time slots is accessed the minimum overall situation energy consumption of each cloud business, and system is worked as Front time-slot throughput, subsequently into step E3；

Step E3., for all channel assignment scheme of current time slots, is got rid of and be there is minimum overall situation energy corresponding to cloud business Consumption is higher than the channel assignment scheme of default cloud business energy consumption higher limit, and in residue channel assignment scheme, selecting system is current Channel assignment scheme corresponding to time-slot throughput maximum, as current time slots optimal channel assignment scheme, subsequently into step Rapid E4；

Step E4. uses current time slots optimal channel assignment scheme, and the minimum overall situation energy corresponding to each cloud business Consumption model realization current time slots cloud business energy optimization scheduling.

As a preferred technical solution of the present invention: described step A, to step B, specifically includes foundation such as drag:

$\begin{array}{cc}\underset{M_t,K_{m,t}}{max}& \underset{m\∈M_t}{\Σ}{R}_{m,t}(K_{m,t},g_{m,k,t})\end{array}$

$\begin{array}{cc}\underset{R_{i,t},K_t}{\min }& E_i(R_{i,t},K_i,\widehat{g_{i,t}})\end{array}$

s.t.

$R_{m,t}(K_{m,t},g_{m,k,t})\≥R_m^0,\∀m\∈M_t$

$\mathrm{\Δ}t\underset{t={\mathrm{\ΔT}}_i}{\overset{{\mathrm{\ΔT}}_i+T_i}{\Σ}}{R}_{i,t}=L_i$

$0\≤P_{i,t}\≤P_{i,max},\∀i,t$

Wherein,For the throughput of system of time slot t,For time slot t choose general Logical collection of services,General service m (m ∈ M is distributed to for time slot t base station_t) channel set, g_m,k,tFor time slot t Distribute to the channel gain of the channel k of general service m；Data slot internal diabetes is uploaded whole for cloud business i The mobile phone energy consumption of consumption, R_i,tFor cloud business i in the speed of time slot t, Δ t is slot time,Divide for time slot t base station The channel set of dispensing cloud business i,Represent the flat expression average of each channel gain, K_iForMiddle letter The number in road, g_i,k,tFor allocated time slot t to the channel gain of the channel k of cloud business i；P_i,tTransmission merit for time slot t cloud business i Rate；Show that channel can not duplicate allocation；K represents that intrasystem channel number, corresponding channel set are K={1 ... k ... K}；M represents the number of general service in system, and corresponding services sets is M={1 ... m ... M}；In system Number I of cloud business, corresponding services sets is I=={1 ... i ... I}, cloud business i needs the data volume uploaded to be L_i, upload The time restriction of data is T_i, the moment starting to upload data is t=Δ T_i。

As a preferred technical solution of the present invention: described step C, to step D, specifically includes and operates as follows:

The energy optimization model of cloud business i is as follows,

$\begin{array}{cc}\underset{R_{i,t}}{min}& \mathrm{\Δ}t\×\underset{t=1}{\overset{T_i}{\Σ}}E\[\frac{(2^{\frac{S_t}{{\mathrm{\ΔtK}}_{i}B}}-1)\×K_i{N}_{0}B}{\widehat{g_t}}\]\end{array}$

s.t.

$\underset{t=1}{\overset{T_i}{\Σ}}{S}_t=L_i$

$S_t\≥0,\∀t$

Optimized model is rewritten as by value function,

$J_t(L_t,\widehat{g_t})=\left\{\begin{array}{cc}\min (\frac{(2^{\frac{S_t}{{\mathrm{\ΔtK}}_{i}B}}-1)K_i{N}_{0}B}{\widehat{g_t}}+E\[J_{t-1}(L_{t-1},\widehat{g_{t-1}})\]),& t=T,\mathrm{...},2\\ \frac{(2^{\frac{L_1}{{\mathrm{\ΔtK}}_{i}B}}-1)K_i{N}_{0}B}{\widehat{g_1}},& t=1\end{array}\right.$

Wherein, S_tFor decision content, refer to decision-making concrete in each stage, the data volume that i.e. this stage is to be sent；L_tFor shape State variable, remaining data volume (including this stage) in referring to each stage；For target function, it is to weigh a decision-making The quantitative index of process, herein refers to energy consumption minimum index；

Utilizing mathematical induction, finally trying to achieve cloud business i at the optimal rate of time slot t is,

The minimum overall situation energy consumption model of cloud business i is,

${E_i}^{*}\left(K_i\right)=\mathrm{\Δ}t(T_i+1)\×2^{\frac{L_i}{(T_i+1)\×T\×K_{i}B}}{G}_i(v_{i,{\mathrm{\ΔT}}_i},...,v_{i,T_i+{\mathrm{\ΔT}}_i})-\mathrm{\Δ}t(T_i+1)v_{i,T_i+{\mathrm{\ΔT}}_i}$

Wherein,

$G_i(v_{i,t},...v_{i,{\mathrm{\ΔT}}_i+T_i})={\left(\underset{x=t}{\overset{{\mathrm{\ΔT}}_i+T_i}{\Π}}{v}_{i,x}\right)}^{\frac{1}{{\mathrm{\ΔT}}_i+T_i-t+1}}.$

As a preferred technical solution of the present invention: described step E1, to step E4, specifically includes and utilizes 0-1 paced beat Draw the channel assignment scheme that Algorithm for Solving makes throughput of system maximum as follows:

Optimization problem is converted into Zero-one integer programming problem, i.e.

$\begin{array}{cc}\underset{x}{max}& e^{T}X\end{array}$

s.t.

[A₁,...,A_N] Χ=1_K

$\left[\begin{array}{ccc}{1}_C^T& \mathrm{...}& 0_C^T\\ .& .& .\\ .& .& .\\ .& .& .\\ 0_C^T& \mathrm{...}& 1_C^T\end{array}\right]X=1_N$

Wherein, Χ=[Χ₁,...,Χ_N]^TBe size be the decision-making column vector of NC, Χ=[Χ₁,...,Χ_N]^T, Χ_n=[x_n,1,...,x_n,C]^T, x_n,j∈ { 0,1}, x_n,jRepresent during for " 1 " that business n uses corresponding the dividing of jth row in allocation matrix Formula case, on the contrary represent and do not use；Represent the allocative decision that a business is possible Number；E be a size be the weight matrix of N × C, its element e_n,jExpression business n uses corresponding the dividing of jth row in allocation matrix Percentage contribution to optimization aim during formula case, i.e.

$\begin{array}{cc}{e}_{i,j}=\left\{\begin{array}{cc}0,& K_i=K_j\\ -1000000,& else\end{array}\right.& \∀i\∈\[1,\mathrm{...},I_t\],\∀j\end{array}$

$\begin{array}{cc}{e}_{i,j}=\left\{\begin{array}{cc}0,& {E_{i,j}}^{*}\≤{\mathrm{\ϵ}}_i\\ -1000000,& else\end{array}\right.& \∀i\∈\[I_t+1,\mathrm{...},I_t^{first}\],\∀j\end{array}$

$\begin{array}{cc}{e}_{m,j}=\left\{\begin{array}{cc}{R}_{m,j,t},& R_{m,j,t}\≥R_m^0\\ 0,& R_{m,j,t}=0\\ -1000000,& else\end{array}\right.& \∀m\∈\[I_t^{first}+1,...,N\],\∀j\end{array}$

A_nBe size be the channel allocation matrix being made up of element 0,1 of K × C, " 1 " represents corresponding channel distribution To this business, " 0 " represents and does not distributes, such as, when having K=3 channel, the most each business has the distribution side that C=7 kind is possible Case, the channel allocation matrix of business n is:

$A_n=\left[\begin{array}{ccccccc}0& 1& 0& 0& 1& 0& 1\\ 0& 0& 1& 0& 1& 1& 1\\ 0& 0& 0& 1& 0& 1& 1\end{array}\right],\∀n\∈N$

Use the method for exhaustion to try to achieve the optimal solution of this optimization problem, i.e. current time slots optimal channel assignment scheme, wherein, K table Showing that intrasystem channel number is K, corresponding channel set is K={1 ... k ... K}；M represents the number of general service in system, Corresponding services sets is M={1 ... m ... M}；Number I of cloud business in system, corresponding services sets is I={1 ... i ... I}, cloud business i needs the data volume uploaded to be L_i, the time restriction uploading data is T_i, the moment starting to upload data is t= ΔT_i。

A kind of multi-service of the present invention deposit system medium cloud business energy optimization dispatching method use above technical scheme Compared with prior art, have following technical effect that multi-service that the present invention designs deposit system medium cloud business energy optimization are adjusted Degree method, it is considered to general service QoS and throughput of system, establishes one and can optimize multiple cloud business energy consumption and can simultaneously The double-goal optimal model of the frequency spectrum resource distribution of maximum system throughput, introduces cloud business and uploads the higher limit of energy, will Cloud business is uploaded the minimum target of energy and is rewritten as the energy restrictive condition less than certain threshold value so that biobjective scheduling problem becomes For single-object problem, then by the method for backward iteration, obtain the letter that optimum energy consumption is assigned to current time slots Relation between road collection, and by this relation, throughput of system will be optimized and optimization cloud traffic energy the two target is unified In a time scale, carried out the distribution of channel the most again by designed Zero-one integer programming algorithm, meet cloud business energy Amount consumes maximum system throughput on the premise of requiring.

Corresponding with above-mentioned, the technical problem to be solved is to provide one and is considering general service QoS and be In the case of system handling capacity, set up resource distribution and cloud service rate dispatches double-goal optimal model, by optimized algorithm respectively Obtain cloud business optimal rate scheduling strategy and the multi-service of channel optimal distribution strategy deposit system medium cloud business energy optimization Dispatching method.

The present invention is to solve above-mentioned technical problem by the following technical solutions: the present invention devises a kind of multi-service and deposits System medium cloud business energy optimization dispatching method, in the system that general service and cloud business coexist, it is achieved the power consumption of cloud business is excellent Changing scheduling, wherein, preset cloud business energy consumption higher limit, described cloud business energy optimization dispatching method comprises the steps:

Step A. is respectively directed to each general service and each cloud business, uses shannon formula to obtain transfer rate and transmission Relation between power, wherein, channel gain obeys independent same distribution, and then obtains each service transmission rate and transmission respectively Relation between power, subsequently into step B；

Step B. is respectively directed to each cloud business, according to the relation between service transmission rate and through-put power, it is thus achieved that base In service transmission rate, bandwidth, the cloud business energy consumption model of channel gain, and then obtain the cloud business energy of each cloud business respectively Consumption model, subsequently into step C；

Step C. is respectively directed to each cloud business, for each working time slot corresponding to cloud business, first, according to cloud industry The cloud business energy consumption model of business, uses dynamic programming method, it is thus achieved that cloud business its last working time slot corresponding chronologically Cloud business time-slot energy consumption cost function, then, use contrary recurrence method, according to the channel gain in work at present time slot, Obtain the cloud business corresponding cloud business time-slot energy consumption cost function of each working time slot before it chronologically successively；And then obtain The cloud business time-slot energy consumption cost function of each cloud business its each working time slot the most corresponding, subsequently into step D；

Step D. is respectively directed to each cloud business, for the cloud business time-slot energy consumption of each working time slot corresponding to cloud business Cost function, it is thus achieved that the optimal rate of cloud business respectively each working time slot corresponding, and then obtain corresponding to cloud business Little overall situation energy consumption model, thus, it is thus achieved that the minimum overall situation energy consumption model corresponding to each cloud business, subsequently into step F1；

Step F1. will be rewritten as phase based on default cloud business energy consumption higher limit, system maximum time-slot throughput optimization problem The Lagrange duality function answered, and enter step F2；

Step F2. illustrates the dual problem of former optimization problem by Lagrange duality algorithm table, and former optimization problem is with right Even problem has identical solution, obtains last solution by solving dual problem, subsequently into step F3；

Step F3. utilizes greedy algorithm to carry out the channel distribution of current time slots, and distribution principle is: a channel should be distributed to Enable to the business that Lagrangian increment is maximum, subsequently into step F4；

Step F4. utilizes two way classification to solve the antithesis coefficient of optimum, and then obtains current time slots optimal channel assignment scheme, Subsequently into step F5；

Step F5. uses current time slots optimal channel assignment scheme, and the minimum overall situation energy corresponding to each cloud business Consumption model realization current time slots cloud business energy optimization scheduling.

As a preferred technical solution of the present invention: described step A, to step B, specifically includes foundation such as drag:

$\begin{array}{cc}\underset{M_t,K_{m,t}}{max}& \underset{m\∈M_t}{\Σ}{R}_{m,t}(K_{m,t},g_{m,k,t})\end{array}$

$\begin{array}{cc}\underset{R_{i,t},K_i}{\min }& E_i(R_{i,t},K_i,\widehat{g_{i,t}})\end{array}$

s.t.

$R_{m,t}(K_{m,t},g_{m,k,t})\≥R_m^0,\∀m\∈M_t$

$\mathrm{\Δ}t\underset{t={\mathrm{\ΔT}}_i}{\overset{{\mathrm{\ΔT}}_i+T_i}{\Σ}}{R}_{i,t}=L_i$

$0\≤P_{i,t}\≤P_{i,max},\∀i,t$

Wherein,For the throughput of system of time slot t,For time slot t choose general Logical collection of services,General service m (m ∈ M is distributed to for time slot t base station_t) channel set, g_m,k,tFor time slot t Distribute to the channel gain of the channel k of general service m；Data slot internal diabetes is uploaded whole for cloud business i The mobile phone energy consumption of consumption, R_i,tFor cloud business i in the speed of time slot t, Δ t is slot time,Divide for time slot t base station The channel set of dispensing cloud business i,Represent the flat expression average of each channel gain, K_iForMiddle letter The number in road, g_i,k,tFor allocated time slot t to the channel gain of the channel k of cloud business i；P_i,tTransmission merit for time slot t cloud business i Rate；Show that channel can not duplicate allocation；K represents that intrasystem channel number, corresponding channel set are K={1 ... k ... K}；M represents the number of general service in system, and corresponding services sets is M={1 ... m ... M}；In system Number I of cloud business, corresponding services sets is I={1 ... i ... I}, cloud business i needs the data volume uploaded to be L_i, upload number According to time restriction be T_i, the moment starting to upload data is t=Δ T_i。

As a preferred technical solution of the present invention: described step C, to step D, specifically includes and operates as follows:

The energy optimization model of cloud business i is as follows,

$\begin{array}{cc}\underset{R_{i,t}}{min}& \mathrm{\Δ}t\×\underset{t=1}{\overset{T_i}{\Σ}}E\[\frac{(2^{\frac{S_t}{{\mathrm{\ΔtK}}_{i}B}}-1)\×K_i{N}_{0}B}{\widehat{g_t}}\]\end{array}$

s.t.

$\underset{t=1}{\overset{T_i}{\Σ}}{S}_t=L_i$

$S_t\≥0,\∀t$

Optimized model is rewritten as by value function,

$J_t(L_t,\widehat{g_t})=\left\{\begin{array}{cc}\min (\frac{(2^{\frac{S_t}{{\mathrm{\ΔtK}}_{i}B}}-1)K_i{N}_{0}B}{\widehat{g_t}}+E\[J_{t-1}(L_{t-1},\widehat{g_{t-1}})\]),& t=T,\mathrm{...},2\\ \frac{(2^{\frac{L_1}{{\mathrm{\ΔtK}}_{i}B}}-1)K_i{N}_{0}B}{\widehat{g_1}},& t=1\end{array}\right.$

Wherein, S_tFor decision content, refer to decision-making concrete in each stage, the data volume that i.e. this stage is to be sent；L_tFor shape State variable, remaining data volume (including this stage) in referring to each stage；For target function, it is to weigh a decision-making The quantitative index of process, herein refers to energy consumption minimum index；

Utilizing mathematical induction, finally trying to achieve cloud business i at the optimal rate of time slot t is,

The minimum overall situation energy consumption model of cloud business i is,

${E_i}^{*}\left(K_i\right)=\mathrm{\Δ}t(T_i+1)\×2^{\frac{L_i}{(T_i+1)\×T\×K_{i}B}}{G}_i(v_{i,{\mathrm{\ΔT}}_i},...,v_{i,T_i+{\mathrm{\ΔT}}_i})-\mathrm{\Δ}t(T_i+1)v_{i,T_i+{\mathrm{\ΔT}}_i}$

Wherein,

$G_i(v_{i,t},...v_{i,{\mathrm{\ΔT}}_i+T_i})={\left(\underset{x=t}{\overset{{\mathrm{\ΔT}}_i+T_i}{\Π}}{v}_{i,x}\right)}^{\frac{1}{{\mathrm{\ΔT}}_i+T_i-t+1}}.$

As a preferred technical solution of the present invention: described step F1, to step F4, specifically includes as follows:

Corresponding glug will be rewritten as based on default cloud business energy consumption higher limit, system maximum time-slot throughput optimization problem Bright day, function was:

$\begin{array}{c}L(M_t,K_{m,t},{\mathrm{\α}}_t,{\mathrm{\β}}_t)=\\ \underset{m\∈M_t}{\Σ}{R}_{m,t}(K_{m,t},g_{m,k,t})+\underset{i\∈I_t^{first}}{\Σ}{\mathrm{\α}}_{i,t}\×({\mathrm{\ϵ}}_i-{E_i}^{*}(K_i\left)\right)+\underset{m\∈M_t}{\Σ}{\mathrm{\β}}_{m,t}\×(R_{m,t}-R_m^0)\end{array}$

Its dual function is,

$f({\mathrm{\α}}_t,{\mathrm{\β}}_t)=\underset{M_t,K_{m,t}}{max}L(M_t,K_{m,t},{\mathrm{\α}}_t,{\mathrm{\β}}_t)$

Corresponding dual problem is

$\begin{array}{cc}\underset{{\mathrm{\α}}_t,{\mathrm{\β}}_t}{min}& f({\mathrm{\α}}_t,{\mathrm{\β}}_t)\end{array}$

s.t.

α_t,β_t≥0

Solve this dual problem, carry out the distribution of channel first with greedy algorithm, then utilize two way classification to solve optimum Antithesis coefficient, concretely comprise the following steps:

I. initialize

Ii. make

Iii. for arbitrary business n, travel through its arbitrary channel k that can be assigned to, all business are carried out as above

Operation, obtains business n and is assigned to L (M during channel k_t,K_m,t,α_t,β_t) increment size, be set to Δ w_n,k, meet

$\begin{array}{cc}{\mathrm{\Δw}}_{i,k}={\mathrm{\α}}_{i,t}\×({\mathrm{\ϵ}}_i-{E_i}^{*}(K_i\left)\right)-{\mathrm{\α}}_{i,t}\×({\mathrm{\ϵ}}_i-{E_i}^{*}(K_i\∪k\left)\right)& \∀i\∈\[1,...,I_t^{first}\]\end{array}$

$\begin{array}{c}{\mathrm{\Δw}}_{m,k}=R_{m,t}(K_m\∪k,g_{m,k,t})+{\mathrm{\β}}_{m,t}\×\left(R_{m,t}\right(K_m\∪k,g_{m,k,t})-R_m^0)\\ \begin{array}{cc}-R_{m,t}(K_m,g_{m,k,t})-{\mathrm{\β}}_{m,t}\×\left(R_{m,t}\right(K_m,g_{m,k,t})-R_m^0)& \∀m\∈\[I_t^{first}+1,\mathrm{...},N\]\end{array}\end{array}$

Find so that Δ w_n,kMaximum (n^*,k^*), by corresponding k^*Distribute to n^*；

Iv. step iii is repeated until all of channel distributes；

V., under channel assignment scheme obtained above, ε is calculated_i-E_i ^*(K_i) value, if ε_i-E_i ^*(K_i) >=0, then correspondingOtherwiseCalculateValue, ifThen correspondingOtherwise

Repeat step ii-v, until forAnd forWherein, δ is our constant of arranging precision for control algolithm, and δ is the least, algorithm Degree of accuracy is the highest, and wherein, K represents that intrasystem channel number is K, and corresponding channel set is K={1 ... k ... K}；M represents system The number of general service in system, corresponding services sets is M={1 ... m ... M}；Number I of cloud business in system, corresponding industry Business integrates as I={1 ... i ... I}, cloud business i needs the data volume uploaded to be L_i, the time restriction uploading data is T_i, start on The moment passing data is t=Δ T_i。

A kind of multi-service of the present invention deposit system medium cloud business energy optimization dispatching method use above technical scheme Compared with prior art, have following technical effect that multi-service that the present invention designs deposit system medium cloud business energy optimization are adjusted Degree method, it is considered to general service QoS and throughput of system, establishes one and can optimize multiple cloud business energy consumption and can simultaneously The double-goal optimal model of the frequency spectrum resource distribution of maximum system throughput, introduces cloud business and uploads the higher limit of energy, will Cloud business is uploaded the minimum target of energy and is rewritten as the energy restrictive condition less than certain threshold value so that biobjective scheduling problem becomes For single-object problem, then by the method for backward iteration, obtain the letter that optimum energy consumption is assigned to current time slots Relation between road collection, and by this relation, throughput of system will be optimized and optimization cloud traffic energy the two target is unified In a time scale, carried out the distribution of channel the most again by designed Lagrange duality algorithm, meet cloud business Maximum system throughput on the premise of energy expenditure requirement.

Accompanying drawing explanation

Fig. 1 is multi-service designed by the present invention the schematic flow sheet of deposit system medium cloud business energy optimization dispatching method；

Fig. 2 be under Zero-one integer programming algorithm and Lagrange duality algorithm throughput of system with transmission time slot variation diagram；

Fig. 3 is that under Zero-one integer programming algorithm and Lagrange duality algorithm, cloud business consumes energy with variation diagram deadline；

Fig. 4 is that under Zero-one integer programming algorithm and Lagrange duality algorithm, cloud business consumes energy with data volume variation diagram；

Fig. 5 is computation complexity comparison diagram under Zero-one integer programming algorithm and Lagrange duality algorithm.

Detailed description of the invention

Below in conjunction with Figure of description, the detailed description of the invention of the present invention is described in further detail.

Cloud business refers to data are uploaded to the class business that high in the clouds carries out processing, and its qos requirement is in certain cut-off Complete the data transmission of corresponding data amount with minimum energy expenditure in time；It is all kinds of that general service is in addition to outside cloud business The general designation of business, its qos requirement is that real time rate should be not less than certain rate requirement.

The present invention devises a kind of multi-service deposit system medium cloud business energy optimization dispatching method, at general service and cloud In the system that business coexists, it is achieved cloud business power consumption Optimized Operation, in actual application, specifically include the following two kinds embodiment, its In, as it is shown in figure 1, the first embodiment comprises the steps:

Step A. is respectively directed to each general service and each cloud business, uses shannon formula to obtain transfer rate and transmission Relation between power, wherein, channel gain obeys independent same distribution, and then obtains each service transmission rate and transmission respectively Relation between power, subsequently into step B.

Step B. is respectively directed to each cloud business, according to the relation between service transmission rate and through-put power, it is thus achieved that base In service transmission rate, bandwidth, the cloud business energy consumption model of channel gain, and then obtain the cloud business energy of each cloud business respectively Consumption model, subsequently into step C.

Above-mentioned steps A, to step B, specifically includes foundation such as drag:

$\begin{array}{cc}\underset{M_t,K_{m,t}}{max}& \underset{m\∈M_t}{\Σ}{R}_{m,t}(K_{m,t},g_{m,k,t})\end{array}$

$\begin{array}{cc}\underset{R_{i,t},K_t}{\min }& E_i(R_{i,t},K_i,\widehat{g_{i,t}})\end{array}$

s.t.

$R_{m,t}(K_{m,t},g_{m,k,t})\≥R_m^0,\∀m\∈M_t$

$\mathrm{\Δ}t\underset{t={\mathrm{\ΔT}}_i}{\overset{{\mathrm{\ΔT}}_i+T_i}{\Σ}}{R}_{i,t}=L_i$

$0\≤P_{i,t}\≤P_{i,max},\∀i,t$

Wherein,For the throughput of system of time slot t,For time slot t choose general Logical collection of services,General service m (m ∈ M is distributed to for time slot t base station_t) channel set, g_m,k,tFor time slot t Distribute to the channel gain of the channel k of general service m；Data slot internal diabetes is uploaded whole for cloud business i The mobile phone energy consumption of consumption, R_{i t}For cloud business i in the speed of time slot t, Δ t is slot time,Divide for time slot t base station The channel set of dispensing cloud business i,Represent the flat expression average of each channel gain, K_iForMiddle letter The number in road, g_i,k,tFor allocated time slot t to the channel gain of the channel k of cloud business i；P_i,tTransmission merit for time slot t cloud business i Rate；Show that channel can not duplicate allocation；K represents that intrasystem channel number, corresponding channel set are K={1 ... k ... K}；M represents the number of general service in system, and corresponding services sets is M={1 ... m ... M}；In system Number I of cloud business, corresponding services sets is I={1 ... i ... I}, cloud business i needs the data volume uploaded to be L_i, upload number According to time restriction be T_i, the moment starting to upload data is t=Δ T_i。

Step C. is respectively directed to each cloud business, for each working time slot corresponding to cloud business, first, according to cloud industry The cloud business energy consumption model of business, uses dynamic programming method, it is thus achieved that cloud business its last working time slot corresponding chronologically Cloud business time-slot energy consumption cost function, then, use contrary recurrence method, according to the channel gain in work at present time slot, Obtain the cloud business corresponding cloud business time-slot energy consumption cost function of each working time slot before it chronologically successively；And then obtain The cloud business time-slot energy consumption cost function of each cloud business its each working time slot the most corresponding, subsequently into step D.

Step D. is respectively directed to each cloud business, for the cloud business time-slot energy consumption of each working time slot corresponding to cloud business Cost function, it is thus achieved that the optimal rate of cloud business respectively each working time slot corresponding, and then obtain corresponding to cloud business Little overall situation energy consumption model, thus, it is thus achieved that the minimum overall situation energy consumption model corresponding to each cloud business, subsequently into step E1.

Above-mentioned steps C, to step D, specifically includes and operates as follows:

The energy optimization model of cloud business i is as follows:

$\begin{array}{cc}\underset{R_{i,t}}{min}& \mathrm{\Δ}t\×\underset{t=1}{\overset{T_i}{\Σ}}E\[\frac{(2^{\frac{S_t}{{\mathrm{\ΔtK}}_{i}B}}-1)\×K_i{N}_{0}B}{\widehat{g_t}}\]\end{array}$

s.t.

$\underset{t=1}{\overset{T_i}{\Σ}}{S}_t=L_i$

$S_t\≥0,\∀t$

Optimized model is rewritten as by value function:

$J_t(L_t,\widehat{g_t})=\left\{\begin{array}{cc}\min (\frac{(2^{\frac{S_t}{{\mathrm{\ΔtK}}_{i}B}}-1)K_i{N}_{0}B}{\widehat{g_t}}+E\[J_{t-1}(L_{t-1},\widehat{g_{t-1}})\]),& t=T,\mathrm{...},2\\ \frac{(2^{\frac{L_1}{{\mathrm{\ΔtK}}_{i}B}}-1)K_i{N}_{0}B}{\widehat{g_1}},& t=1\end{array}\right.$

Wherein, S_tFor decision content, refer to decision-making concrete in each stage, the data volume that i.e. this stage is to be sent；L_tFor shape State variable, remaining data volume (including this stage) in referring to each stage；For target function, it is to weigh a decision-making The quantitative index of process, herein refers to energy consumption minimum index.

Utilizing mathematical induction, finally trying to achieve cloud business i at the optimal rate of time slot t is,

The minimum overall situation energy consumption model of cloud business i is,

${E_i}^{*}\left(K_i\right)=\mathrm{\Δ}t(T_i+1)\×2^{\frac{L_i}{(T_i+1)\×T\×K_{i}B}}{G}_i(v_{i,{\mathrm{\ΔT}}_i},...,v_{i,T_i+{\mathrm{\ΔT}}_i})-\mathrm{\Δ}t(T_i+1)v_{i,T_i+{\mathrm{\ΔT}}_i}$

Wherein,

$G_i(v_{i,t},...v_{i,{\mathrm{\ΔT}}_i+T_i})={\left(\underset{x=t}{\overset{{\mathrm{\ΔT}}_i+T_i}{\Π}}{v}_{i,x}\right)}^{\frac{1}{{\mathrm{\ΔT}}_i+T_i-t+1}}.$

Step E1. is according to current time slots channel quantity to be allocated, it is thus achieved that for current time slots each cloud business to be accessed, The all channel assignment scheme of each general service, subsequently into step E2.

Step E2. is respectively directed to each channel assignment scheme of current time slots, according to the minimum corresponding to each cloud business Overall situation energy consumption model, it is thus achieved that under channel assignment scheme, current time slots accessed the minimum overall energy consumption of each cloud business, obtain simultaneously Obtain under channel assignment scheme, current time slots is accessed each general service transfer rate sum, i.e. system current time-slot throughput； And then obtaining under each channel assignment scheme, current time slots is accessed the minimum overall situation energy consumption of each cloud business, and system is worked as Front time-slot throughput, subsequently into step E3.

Step E3., for all channel assignment scheme of current time slots, is got rid of and be there is minimum overall situation energy corresponding to cloud business Consumption is higher than the channel assignment scheme of default cloud business energy consumption higher limit, and in residue channel assignment scheme, selecting system is current Channel assignment scheme corresponding to time-slot throughput maximum, as current time slots optimal channel assignment scheme, subsequently into step Rapid E4.

Step E4. uses current time slots optimal channel assignment scheme, and the minimum overall situation energy corresponding to each cloud business Consumption model realization current time slots cloud business energy optimization scheduling.

Above-mentioned steps E1, to step E4, specifically includes and utilizes Zero-one integer programming Algorithm for Solving to make throughput of system maximum Channel assignment scheme as follows:

Optimization problem is converted into Zero-one integer programming problem, i.e.

$\begin{array}{cc}\underset{x}{max}& e^{T}X\end{array}$

s.t.

[A₁,...,A_N] Χ=1_K

$\left[\begin{array}{ccc}{1}_C^T& \mathrm{...}& 0_C^T\\ .& .& .\\ .& .& .\\ .& .& .\\ 0_C^T& \mathrm{...}& 1_C^T\end{array}\right]X=1_N$

Wherein, Χ=[Χ₁,...,Χ_N]^TBe size be the decision-making column vector of NC, Χ=[Χ₁,...,Χ_N]^T, Χ_n=[x_n,1,...,x_n,C]^T, x_n,j∈ { 0,1}, x_n,jRepresent during for " 1 " that business n uses corresponding the dividing of jth row in allocation matrix Formula case, on the contrary represent and do not use；Represent the allocative decision that a business is possible Number；E be a size be the weight matrix of N × C, its element e_n,jExpression business n uses corresponding the dividing of jth row in allocation matrix Percentage contribution to optimization aim during formula case, i.e.

$\begin{array}{cc}{e}_{i,j}=\left\{\begin{array}{cc}0,& K_i=K_j\\ -1000000,& else\end{array}\right.& \∀i\∈\[1,\mathrm{...},I_t\],\∀j\end{array}$

$\begin{array}{cc}{e}_{i,j}=\left\{\begin{array}{cc}0,& {E_{i,j}}^{*}\≤{\mathrm{\ϵ}}_i\\ -1000000,& else\end{array}\right.& \∀i\∈\[I_t+1,\mathrm{...},I_t^{first}\],\∀j\end{array}$

$\begin{array}{cc}{e}_{m,j}=\left\{\begin{array}{cc}{R}_{m,j,t},& R_{m,j,t}\≥R_m^0\\ 0,& R_{m,j,t}=0\\ -1000000,& else\end{array}\right.& \∀m\∈\[I_t^{first}+1,...,N\],\∀j\end{array}$

A_nBe size be the channel allocation matrix being made up of element 0,1 of K × C, " 1 " represents corresponding channel distribution To this business, " 0 " represents and does not distributes, such as, when having K=3 channel, the most each business has the distribution side that C=7 kind is possible Case, the channel allocation matrix of business n is:

$A_n=\left[\begin{array}{ccccccc}0& 1& 0& 0& 1& 0& 1\\ 0& 0& 1& 0& 1& 1& 1\\ 0& 0& 0& 1& 0& 1& 1\end{array}\right],\∀n\∈N$

Use the method for exhaustion to try to achieve the optimal solution of this optimization problem, i.e. current time slots optimal channel assignment scheme, wherein, K table Showing that intrasystem channel number is K, corresponding channel set is K={1 ... k ... K}；M represents the number of general service in system, Corresponding services sets is M={1 ... m ... M}；Number I of cloud business in system, corresponding services sets is I={1 ... i ... I}, cloud business i needs the data volume uploaded to be L_i, the time restriction uploading data is T_i, the moment starting to upload data is t= ΔT_i。

Also shown in FIG. 1, the second embodiment, preset cloud business energy consumption higher limit, described cloud business energy optimization is adjusted Degree method, specifically includes following steps:

Step A. is respectively directed to each general service and each cloud business, uses shannon formula to obtain transfer rate and transmission Relation between power, wherein, channel gain obeys independent same distribution, and then obtains each service transmission rate and transmission respectively Relation between power, subsequently into step B.

Step B. is respectively directed to each cloud business, according to the relation between service transmission rate and through-put power, it is thus achieved that base In service transmission rate, bandwidth, the cloud business energy consumption model of channel gain, and then obtain the cloud business energy of each cloud business respectively Consumption model, subsequently into step C.

Wherein, step A, to step B, specifically includes foundation such as drag:

$\begin{array}{cc}\underset{M_t,K_{m,t}}{max}& \underset{m\∈M_t}{\Σ}{R}_{m,t}(K_{m,t},g_{m,k,t})\end{array}$

$\begin{array}{cc}\underset{R_{i,t},K_t}{\min }& E_i(R_{i,t},K_i,\widehat{g_{i,t}})\end{array}$

s.t.

$R_{m,t}(K_{m,t},g_{m,k,t})\≥R_m^0,\∀m\∈M_t$

$\mathrm{\Δ}t\underset{t={\mathrm{\ΔT}}_i}{\overset{{\mathrm{\ΔT}}_i+T_i}{\Σ}}{R}_{i,t}=L_i$

$0\≤P_{i,t}\≤P_{i,max},\∀i,t$

Wherein,For the throughput of system of time slot t,For time slot t choose general Logical collection of services,General service m (m ∈ M is distributed to for time slot t base station_t) channel set, g_m,k,tFor time slot t Distribute to the channel gain of the channel k of general service m；Data slot internal diabetes is uploaded whole for cloud business i The mobile phone energy consumption of consumption, R_i,tFor cloud business i in the speed of time slot t, Δ t is slot time,Divide for time slot t base station The channel set of dispensing cloud business i,Represent the flat expression average of each channel gain, K_iForMiddle letter The number in road, g_i,k,tFor allocated time slot t to the channel gain of the channel k of cloud business i；P_i,tTransmission merit for time slot t cloud business i Rate；Show that channel can not duplicate allocation；K represents that intrasystem channel number, corresponding channel set are K={1 ... k ... K}；M represents the number of general service in system, and corresponding services sets is M={1 ... m ... M}；In system Number I of cloud business, corresponding services sets is I={1 ... i ... I}, cloud business i needs the data volume uploaded to be L_i, upload number According to time restriction be T_i, the moment starting to upload data is t=Δ T_i。

Step C. is respectively directed to each cloud business, for each working time slot corresponding to cloud business, first, according to cloud industry The cloud business energy consumption model of business, uses dynamic programming method, it is thus achieved that cloud business its last working time slot corresponding chronologically Cloud business time-slot energy consumption cost function, then, use contrary recurrence method, according to the channel gain in work at present time slot, Obtain the cloud business corresponding cloud business time-slot energy consumption cost function of each working time slot before it chronologically successively；And then obtain The cloud business time-slot energy consumption cost function of each cloud business its each working time slot the most corresponding, subsequently into step D.

Step D. is respectively directed to each cloud business, for the cloud business time-slot energy consumption of each working time slot corresponding to cloud business Cost function, it is thus achieved that the optimal rate of cloud business respectively each working time slot corresponding, and then obtain corresponding to cloud business Little overall situation energy consumption model, thus, it is thus achieved that the minimum overall situation energy consumption model corresponding to each cloud business, subsequently into step F1.

Wherein, step C, to step D, specifically includes and operates as follows:

The energy optimization model of cloud business i is as follows,

$\begin{array}{cc}\underset{R_{i,t}}{min}& \mathrm{\Δ}t\×\underset{t=1}{\overset{T_i}{\Σ}}E\[\frac{(2^{\frac{S_t}{{\mathrm{\ΔtK}}_{i}B}}-1)\×K_i{N}_{0}B}{\widehat{g_t}}\]\end{array}$

s.t.

$\underset{t=1}{\overset{T_i}{\Σ}}{S}_t=L_i$

$S_t\≥0,\∀t$

Optimized model is rewritten as by value function,

$J_t(L_t,\widehat{g_t})=\left\{\begin{array}{cc}\min (\frac{(2^{\frac{S_t}{{\mathrm{\ΔtK}}_{i}B}}-1)K_i{N}_{0}B}{\widehat{g_t}}+E\[J_{t-1}(L_{t-1},\widehat{g_{t-1}})\]),& t=T,\mathrm{...},2\\ \frac{(2^{\frac{L_1}{{\mathrm{\ΔtK}}_{i}B}}-1)K_i{N}_{0}B}{\widehat{g_1}},& t=1\end{array}\right.$

Wherein, S_tFor decision content, refer to decision-making concrete in each stage, the data volume that i.e. this stage is to be sent；L_tFor shape State variable, remaining data volume (including this stage) in referring to each stage；For target function, it is to weigh a decision-making The quantitative index of process, herein refers to energy consumption minimum index；

Utilizing mathematical induction, finally trying to achieve cloud business i at the optimal rate of time slot t is,

The minimum overall situation energy consumption model of cloud business i is,

${E_i}^{*}\left(K_i\right)=\mathrm{\Δ}t(T_i+1)\×2^{\frac{L_i}{(T_i+1)\×T\×K_{i}B}}{G}_i(v_{i,{\mathrm{\ΔT}}_i},...,v_{i,T_i+{\mathrm{\ΔT}}_i})-\mathrm{\Δ}t(T_i+1)v_{i,T_i+{\mathrm{\ΔT}}_i}$

Wherein,

$G_i(v_{i,t},...v_{i,{\mathrm{\ΔT}}_i+T_i})={\left(\underset{x=t}{\overset{{\mathrm{\ΔT}}_i+T_i}{\Π}}{v}_{i,x}\right)}^{\frac{1}{{\mathrm{\ΔT}}_i+T_i-t+1}}.$

Step F1. will be rewritten as phase based on default cloud business energy consumption higher limit, system maximum time-slot throughput optimization problem The Lagrange duality function answered, and enter step F2.

Step F2. illustrates the dual problem of former optimization problem by Lagrange duality algorithm table, and former optimization problem is with right Even problem has identical solution, obtains last solution by solving dual problem, subsequently into step F3.

Step F3. utilizes greedy algorithm to carry out the channel distribution of current time slots, and distribution principle is: a channel should be distributed to Enable to the business that Lagrangian increment is maximum, subsequently into step F4.

Step F4. utilizes two way classification to solve the antithesis coefficient of optimum, and then obtains current time slots optimal channel assignment scheme, Subsequently into step F5.

Step F5. uses current time slots optimal channel assignment scheme, and the minimum overall situation energy corresponding to each cloud business Consumption model realization current time slots cloud business energy optimization scheduling.

Wherein, step F1, to step F4, specifically includes as follows:

Corresponding glug will be rewritten as based on default cloud business energy consumption higher limit, system maximum time-slot throughput optimization problem Bright day, function was:

$\begin{array}{c}L(M_t,K_{m,t},{\mathrm{\α}}_t,{\mathrm{\β}}_t)=\\ \underset{m\∈M_t}{\Σ}{R}_{m,t}(K_{m,t},g_{m,k,t})+\underset{i\∈I_t^{first}}{\Σ}{\mathrm{\α}}_{i,t}\×({\mathrm{\ϵ}}_i-{E_i}^{*}(K_i\left)\right)+\underset{m\∈M_t}{\Σ}{\mathrm{\β}}_{m,t}\×(R_{m,t}-R_m^0)\end{array}$

Its dual function is,

$f({\mathrm{\α}}_t,{\mathrm{\β}}_t)=\underset{M_t,K_{m,t}}{max}L(M_t,K_{m,t},{\mathrm{\α}}_t,{\mathrm{\β}}_t)$

Corresponding dual problem is

$\begin{array}{cc}\underset{{\mathrm{\α}}_t,{\mathrm{\β}}_t}{min}& f({\mathrm{\α}}_t,{\mathrm{\β}}_t)\end{array}$

s.t.

α_t,β_t≥0

Solve this dual problem, carry out the distribution of channel first with greedy algorithm, then utilize two way classification to solve optimum Antithesis coefficient, concretely comprise the following steps:

Vi. initialize

Vii. make

Viii. for arbitrary business n, travel through its arbitrary channel k that can be assigned to, all business are as above grasped Make, obtain business n and be assigned to L (M during channel k_t,K_m,t,α_t,β_t) increment size, be set to Δ w_n,k, meet

$\begin{array}{cc}{\mathrm{\Δw}}_{i,k}={\mathrm{\α}}_{i,t}\×({\mathrm{\ϵ}}_i-{E_i}^{*}(K_i\left)\right)-{\mathrm{\α}}_{i,t}\×({\mathrm{\ϵ}}_i-{E_i}^{*}(K_i\∪k\left)\right)& \∀i\∈\[1,...,I_t^{first}\]\end{array}$

$\begin{array}{c}{\mathrm{\Δw}}_{m,k}=R_{m,t}(K_m\∪k,g_{m,k,t})+{\mathrm{\β}}_{m,t}\×\left(R_{m,t}\right(K_m\∪k,g_{m,k,t})-R_m^0)\\ \begin{array}{cc}-R_{m,t}(K_m,g_{m,k,t})-{\mathrm{\β}}_{m,t}\×\left(R_{m,t}\right(K_m,g_{m,k,t})-R_m^0)& \∀m\∈\[I_t^{first}+1,\mathrm{...},N\]\end{array}\end{array}$

Find so that Δ w_n,kMaximum (n^*,k^*), by corresponding k^*Distribute to n^*；

Ix. step iii is repeated until all of channel distributes；

X., under channel assignment scheme obtained above, ε is calculated_i-E_i ^*(K_i) value, if ε_i-E_i ^*(K_i) >=0, then correspondingOtherwiseCalculateValue, ifThen correspondingOtherwise

Repeat step ii-v, until forAnd forWherein, δ is our constant of arranging precision for control algolithm, and δ is the least, algorithm Degree of accuracy is the highest, and wherein, K represents that intrasystem channel number is K, and corresponding channel set is K={1 ... k ... K}；M represents system The number of general service in system, corresponding services sets is M={1 ... m ... M}；Number I of cloud business in system, corresponding industry Business integrates as I={1 ... i ... I}, cloud business i needs the data volume uploaded to be L_i, the time restriction uploading data is T_i, start on The moment passing data is t=Δ T_i。

Channel last in multi-service designed by the present invention deposit system medium cloud business energy optimization dispatching method is divided For joining, specific design uses Zero-one integer programming algorithm and Lagrange duality algorithm to carry out the distribution of channel respectively, as Fig. 2, Shown in Fig. 3, Fig. 4, Fig. 5, it is seen that the different-effect that two algorithms are brought in actual applications.

The present invention design multi-service and deposit system medium cloud business energy optimization dispatching method, it is considered to general service QoS and Throughput of system, establish one can optimize simultaneously multiple cloud business energy consumption and can maximum system throughput frequency spectrum money The double-goal optimal model of source distribution, introduces cloud business and uploads the higher limit of energy, cloud business is uploaded the minimum target of energy It is rewritten as the energy restrictive condition less than certain threshold value so that biobjective scheduling problem becomes single-object problem, then leads to The method crossing backward iteration, obtains optimum energy consumption and relation between the channel set that current time slots is assigned to, and by this Individual relation, will optimize throughput of system and optimize in cloud traffic energy the two target unification to time scale, the most again Carried out the distribution of channel by designed Zero-one integer programming algorithm or Lagrange duality algorithm, disappear meeting cloud traffic energy Maximum system throughput on the premise of consumption requirement.

Above in conjunction with accompanying drawing, embodiments of the present invention are explained in detail, but the present invention is not limited to above-mentioned enforcement Mode, in the ken that those of ordinary skill in the art are possessed, it is also possible on the premise of without departing from present inventive concept Make a variety of changes.

Claims (6)

1. multi-service a deposit system medium cloud business energy optimization dispatching method, in the system that general service and cloud business coexist In, it is achieved cloud business power consumption Optimized Operation, it is characterised in that comprise the steps:

Step A. is respectively directed to each general service and each cloud business, uses shannon formula to obtain transfer rate and through-put power Between relation, wherein, channel gain obeys independent same distribution, and then obtains each service transmission rate and through-put power respectively Between relation, subsequently into step B；

Step B. is respectively directed to each cloud business, according to the relation between service transmission rate and through-put power, it is thus achieved that based on industry Business transfer rate, bandwidth, the cloud business energy consumption model of channel gain, and then obtain the cloud business energy consumption mould of each cloud business respectively Type, subsequently into step C；

Step C. is respectively directed to each cloud business, for each working time slot corresponding to cloud business, first, according to cloud business Cloud business energy consumption model, uses dynamic programming method, it is thus achieved that the cloud of cloud business its last working time slot corresponding chronologically Business time-slot energy consumption cost function, then, uses contrary recurrence method, according to the channel gain in work at present time slot, successively Obtain the cloud business corresponding cloud business time-slot energy consumption cost function of each working time slot before it chronologically；And then obtain each The cloud business time-slot energy consumption cost function of cloud business its each working time slot the most corresponding, subsequently into step D；

Step D. is respectively directed to each cloud business, and the cloud business time-slot energy consumption for each working time slot corresponding to cloud business is worth Function, it is thus achieved that the optimal rate of cloud business each working time slot the most corresponding, and then it is complete to obtain the minimum corresponding to cloud business Office's energy consumption model, thus, it is thus achieved that the minimum overall situation energy consumption model corresponding to each cloud business, subsequently into step E1；

Step E1. is according to current time slots channel quantity to be allocated, it is thus achieved that for current time slots each cloud business to be accessed, each The all channel assignment scheme of general service, subsequently into step E2；

Step E2. is respectively directed to each channel assignment scheme of current time slots, according to the minimum overall situation corresponding to each cloud business Energy consumption model, it is thus achieved that under channel assignment scheme, current time slots accessed the minimum overall situation energy consumption of each cloud business, obtain letter simultaneously Under road allocative decision, current time slots accessed each general service transfer rate sum, i.e. system current time-slot throughput；And then Obtaining under each channel assignment scheme, current time slots is accessed the minimum overall situation energy consumption of each cloud business, and when system is current Gap handling capacity, subsequently into step E3；

Step E3., for all channel assignment scheme of current time slots, is got rid of and be there is minimum overall situation energy consumption height corresponding to cloud business In the channel assignment scheme of default cloud business energy consumption higher limit, and in residue channel assignment scheme, selecting system current time slots Channel assignment scheme corresponding to handling capacity maximum, as current time slots optimal channel assignment scheme, subsequently into step E4； Step E4. uses current time slots optimal channel assignment scheme, and the minimum overall situation energy consumption model corresponding to each cloud business is real Existing current time slots cloud business energy optimization scheduling.

2. multi-service a deposit system medium cloud business energy optimization dispatching method, in the system that general service and cloud business coexist In, it is achieved cloud business power consumption Optimized Operation, it is characterised in that preset cloud business energy consumption higher limit, described cloud business energy optimization Dispatching method comprises the steps:

Step A. is respectively directed to each general service and each cloud business, uses shannon formula to obtain transfer rate and through-put power Between relation, wherein, channel gain obeys independent same distribution, and then obtains each service transmission rate and through-put power respectively Between relation, subsequently into step B；

Step B. is respectively directed to each cloud business, according to the relation between service transmission rate and through-put power, it is thus achieved that based on industry Business transfer rate, bandwidth, the cloud business energy consumption model of channel gain, and then obtain the cloud business energy consumption mould of each cloud business respectively Type, subsequently into step C；

Step C. is respectively directed to each cloud business, for each working time slot corresponding to cloud business, first, according to cloud business Cloud business energy consumption model, uses dynamic programming method, it is thus achieved that the cloud of cloud business its last working time slot corresponding chronologically Business time-slot energy consumption cost function, then, uses contrary recurrence method, according to the channel gain in work at present time slot, successively Obtain the cloud business corresponding cloud business time-slot energy consumption cost function of each working time slot before it chronologically；And then obtain each The cloud business time-slot energy consumption cost function of cloud business its each working time slot the most corresponding, subsequently into step D；

Step D. is respectively directed to each cloud business, and the cloud business time-slot energy consumption for each working time slot corresponding to cloud business is worth Function, it is thus achieved that the optimal rate of cloud business each working time slot the most corresponding, and then it is complete to obtain the minimum corresponding to cloud business Office's energy consumption model, thus, it is thus achieved that the minimum overall situation energy consumption model corresponding to each cloud business, subsequently into step F1；

Step F1. will be rewritten as based on default cloud business energy consumption higher limit, system maximum time-slot throughput optimization problem accordingly Lagrange duality function, and enter step F2；

By Lagrange duality algorithm table, step F2. illustrates that the dual problem of former optimization problem, former optimization problem are asked with antithesis Topic has identical solution, obtains last solution by solving dual problem, subsequently into step F3；

Step F3. utilizes greedy algorithm to carry out the channel distribution of current time slots, and distribution principle is: channel should be distributed to can Make the business that Lagrangian increment is maximum, subsequently into step F4；

Step F4. utilizes two way classification to solve the antithesis coefficient of optimum, and then obtains current time slots optimal channel assignment scheme, then Enter step F5；

Step F5. uses current time slots optimal channel assignment scheme, and the minimum overall situation energy consumption mould corresponding to each cloud business Type realizes the scheduling of current time slots cloud business energy optimization.

A kind of multi-service the most according to claim 1 or claim 2 deposit system medium cloud business energy optimization dispatching method, its feature exists In, described step A, to step B, specifically includes foundation such as drag:

$\begin{array}{cc}\underset{M_t,K_{m,t}}{max}& \underset{m\∈M_t}{\Σ}{R}_{m,t}(K_{m,t},g_{m,k,t})\end{array}$

$\begin{array}{cc}\underset{R_{i,t},K_i}{min}& E_i(R_{i,t},K_i,\widehat{g_{i,t}})\end{array}$

s.t.

$R_{m,t}(K_{m,t},g_{m,k,t})\≥R_m^0,\∀m\∈M_t$

$\mathrm{\Δ}t\underset{t={\mathrm{\ΔT}}_i}{\overset{{\mathrm{\ΔT}}_i+T_i}{\Σ}}{R}_{i,t}=L_i$

$0\≤P_{i,t}\≤P_{i,\max },\∀i,t$

Wherein,For the throughput of system of time slot t,The common industry chosen for time slot t Business set,General service m (m ∈ M is distributed to for time slot t base station_t) channel set, g_m,k,tFor allocated time slot t Channel gain to the channel k of general service m；For cloud business i at the whole data slot internal consumption uploaded Mobile phone energy consumption, R_i,tFor cloud business i in the speed of time slot t, Δ t is slot time,Distribute to for time slot t base station The channel set of cloud business i,Represent the flat expression average of each channel gain, K_iForMiddle channel Number, g_i,k,tFor allocated time slot t to the channel gain of the channel k of cloud business i；P_i,tThrough-put power for time slot t cloud business i；Show that channel can not duplicate allocation；K represents intrasystem channel number, and corresponding channel set is K= {1,…k,…K}；M represents the number of general service in system, and corresponding services sets is M={1 ... m ... M}；Cloud industry in system Number I of business, corresponding services sets isCloud business i needs the data volume uploaded to be L_i, upload data time Between be limited to T_i, the moment starting to upload data is t=Δ T_i。

A kind of multi-service deposit system medium cloud business energy optimization dispatching method, it is characterised in that Described step C, to step D, specifically includes and operates as follows:

The energy optimization model of cloud business i is as follows,

$\begin{array}{cc}\underset{R_{i,t}}{min}& \mathrm{\Δ}t\×\underset{t=1}{\overset{T_i}{\Σ}}E\[\frac{(2^{\frac{S_t}{{\mathrm{\ΔtK}}_{i}B}}-1)\×K_i{N}_{0}B}{\widehat{g_t}}\]\end{array}$

s.t.

$\underset{t=1}{\overset{T_i}{\Σ}}{S}_t=L_i$

$S_t\≥0,\∀t$

Optimized model is rewritten as by value function,

$J_t(L_t,\widehat{g_t})=\left\{\begin{array}{cc}\min \left(\frac{(2^{\frac{S_t}{{\mathrm{\ΔtK}}_{i}B}}-1)K_i{N}_{0}B}{\widehat{g_t}}+E\[J_{t-1}\left(L_{t-1},\widehat{g_{t-1}}\right)\]\right),& t=T,\mathrm{...},2\\ \frac{(2^{\frac{L_1}{{\mathrm{\ΔtK}}_{i}B}}-1)K_i{N}_{0}B}{\widehat{g_1}},& t=1\end{array}\right.$

Wherein, S_tFor decision content, refer to decision-making concrete in each stage, the data volume that i.e. this stage is to be sent；L_tBecome for state Amount, remaining data volume (including this stage) in referring to each stage；For target function, it is to weigh a decision making process Quantitative index, herein refer to energy consumption minimum index；

Utilizing mathematical induction, finally trying to achieve cloud business i at the optimal rate of time slot t is,

The minimum overall situation energy consumption model of cloud business i is,

${E_i}^{*}\left(K_i\right)=\mathrm{\Δ}t(T_i+1)\×2^{\frac{L_i}{(T_i+1)\×T\×K_{i}B}}{G}_i(v_{i,{\mathrm{\ΔT}}_i},\mathrm{...},v_{i,T_i+{\mathrm{\ΔT}}_i})-\mathrm{\Δ}t(T_i+1)v_{i,T_i+{\mathrm{\ΔT}}_i}$

Wherein,X=t ..., Δ T_i+T_i,

$G_i(v_{i,t},...v_{i,{\mathrm{\ΔT}}_i+T_i})={\left(\underset{x=t}{\overset{{\mathrm{\ΔT}}_i+T_i}{\Π}}{v}_{i,x}\right)}^{\frac{1}{{\mathrm{\ΔT}}_i+T_i-t+1}}.$

A kind of multi-service deposit system medium cloud business energy optimization dispatching method, it is characterised in that Described step E1, to step E4, specifically includes the channel utilizing Zero-one integer programming Algorithm for Solving to make throughput of system maximum and divides Formula case is as follows:

Optimization problem is converted into Zero-one integer programming problem, i.e.

$\begin{array}{cc}\underset{x}{max}& e^{T}X\end{array}$

s.t.

[A₁,...,A_N] Χ=1_K

$\left[\begin{array}{ccc}{1}_C^T& \mathrm{...}& 0_C^T\\ .& .& .\\ .& .& .\\ .& .& .\\ 0_C^T& \mathrm{...}& 1_C^T\end{array}\right]X=1_N$

Wherein, Χ=[Χ₁,...,Χ_N]^TBe size be the decision-making column vector of NC, Χ=[Χ₁,...,Χ_N]^T, Χ_n= [x_n,1,...,x_n,C]^T, x_n,j∈ { 0,1}, x_n,jRepresent during for " 1 " that business n uses the distribution side that in allocation matrix, jth row are corresponding Case, on the contrary represent and do not use；Represent the allocative decision number that a business is possible；e Be a size be the weight matrix of N × C, its element e_n,jExpression business n uses the distribution side that in allocation matrix, jth row are corresponding Percentage contribution to optimization aim during case, i.e.

$e_{i,j}=\left\{\begin{array}{cc}0,& K_i=K_j\\ -1000000,& else\end{array}\right.,\∀i\∈\[1,...,I_t\],\∀j$

$e_{i,j}=\left\{\begin{array}{cc}0,& {E_{i,j}}^{*}\≤{\mathrm{\ϵ}}_i\\ -1000000,& else\end{array}\right.,\∀i\∈\[I_t+1,...,I_t^{first}\],\∀j$

$e_{m,j}=\left\{\begin{array}{cc}{R}_{m,j,t},& R_{m,j,t}\≥R_m^0\\ 0,& R_{m,j,t}=0\\ -1000000,& else\end{array}\right.,\∀m\∈\[I_t^{first}+1,...,N\],\∀j$

A_nBe size be the channel allocation matrix being made up of element 0,1 of K × C, " 1 " represents that corresponding channel distributes to this Business, " 0 " represents and does not distributes, such as, when having K=3 channel, the most each business has the allocative decision that C=7 kind is possible, industry The channel allocation matrix of business n is:

$A_n=\left[\begin{array}{ccccccc}0& 1& 0& 0& 1& 0& 1\\ 0& 0& 1& 0& 1& 1& 1\\ 0& 0& 0& 1& 0& 1& 1\end{array}\right],\∀n\∈N$

Use the method for exhaustion to try to achieve the optimal solution of this optimization problem, i.e. current time slots optimal channel assignment scheme, wherein, K represents system Channel number in system is K, and corresponding channel set is K={1 ... k ... K}；M represents the number of general service in system, corresponding Services sets be M={1 ... m ... M}；Number I of cloud business in system, corresponding services sets isYun Ye Business i needs the data volume uploaded to be L_i, the time restriction uploading data is T_i, the moment starting to upload data is t=Δ T_i。

A kind of multi-service deposit system medium cloud business energy optimization dispatching method, it is characterised in that Described step F1, to step F4, specifically includes as follows:

Corresponding Lagrange will be rewritten as based on default cloud business energy consumption higher limit, system maximum time-slot throughput optimization problem Function is:

$\begin{array}{c}L(M_t,K_{m,t},{\mathrm{\α}}_t,{\mathrm{\β}}_t)=\\ \underset{m\∈M_t}{\Σ}{R}_{m,t}(K_{m,t},g_{m,k,t})+\underset{i\∈I_t^{first}}{\Σ}{\mathrm{\α}}_{i,t}\×({\mathrm{\ϵ}}_i-{E_i}^{*}(K_i\left)\right)+\underset{m\∈M_t}{\Σ}{\mathrm{\β}}_{m,t}\×(R_{m,t}-R_m^0)\end{array}$

Its dual function is,

$f({\mathrm{\α}}_t,{\mathrm{\β}}_t)=\underset{M_t,K_{m,t}}{max}L(M_t,K_{m,t},{\mathrm{\α}}_t,{\mathrm{\β}}_t)$

Corresponding dual problem is

$\begin{array}{cc}\underset{{\mathrm{\α}}_t,{\mathrm{\β}}_t}{min}& f({\mathrm{\α}}_t,{\mathrm{\β}}_t)\end{array}$

s.t.

α_t,β_t≥0

Solve this dual problem, carry out the distribution of channel first with greedy algorithm, then utilize two way classification to solve the right of optimum Even coefficient, concretely comprises the following steps:

I. initialize

Ii. make

Iii. for arbitrary business n, travel through its arbitrary channel k that can be assigned to, all business are operated as above, obtains Business n is assigned to L (M during channel k_t,K_m,t,α_t,β_t) increment size, be set to Δ w_n,k, meet

${\mathrm{\Δw}}_{i,k}={\mathrm{\α}}_{i,t}\×({\mathrm{\ϵ}}_i-{E_i}^{*}(K_i\left)\right)-{\mathrm{\α}}_{i,t}\×({\mathrm{\ϵ}}_i-{E_i}^{*}(K_i\∪k\left)\right),\∀i\∈\[1,...,I_t^{first}\]$

Find so that Δ w_n,kMaximum (n^*,k^*), by corresponding k^*Distribute to n^*；

Iv. step iii is repeated until all of channel distributes；

V., under channel assignment scheme obtained above, ε is calculated_i-E_i ^*(K_i) value, if ε_i-E_i ^*(K_i) >=0, then correspondingOtherwiseCalculateValue, ifThen correspondingOtherwise

Repeat step ii-v, until forAnd for Wherein, δ is our constant of arranging precision for control algolithm, and δ is the least, and algorithm degree of accuracy is the highest, and wherein, K represents system Interior channel number is K, and corresponding channel set is K={1 ... k ... K}；M represents the number of general service in system, corresponding Services sets is M={1 ... m ... M}；Number I of cloud business in system, corresponding services sets isCloud business i Needing the data volume uploaded is L_i, the time restriction uploading data is T_i, the moment starting to upload data is t=Δ T_i。