CN106304308B - A kind of multi-service and deposit system medium cloud business energy optimization dispatching method - Google Patents

Abstract

The present invention relates to a kind of multi-service and deposit system medium cloud business energy optimization dispatching methods, establish the double-goal optimal model that can optimize multiple cloud business energy consumptions simultaneously and be capable of the frequency spectrum resource distribution of maximum system throughput, introduce the upper limit value that cloud business uploads energy, the minimum target that cloud business uploads energy is rewritten as the restrictive condition that energy is less than certain threshold value, so that biobjective scheduling problem becomes single-object problem, then pass through the method for backward iteration, the relationship that optimum energy consumption is obtained between channel set that current time slots are assigned to, and pass through this relationship, it will be in optimization system handling capacity and optimization the two targets of cloud traffic energy unification a to time scale, the distribution of channel is finally carried out by designed Zero-one integer programming algorithm or Lagrange duality algorithm again, disappear meeting cloud traffic energy Maximum system throughput under the premise of consumption requires.

Description

An optimal scheduling method for cloud service energy consumption in a multi-service coexistence system

technical field

The invention relates to a cloud service energy consumption optimization scheduling method in a multi-service coexistence system, and belongs to the technical fields of wireless cloud computing technology and wireless resource optimization.

Background technique

The rapid growth of Internet traffic and the continuous innovation of applications have promoted the development of mobile cloud computing technology. The wireless cloud service uploads the application data to the cloud for processing through the wireless network, which reduces the computing power requirements of the mobile terminal and improves the service processing efficiency. However, compared with the booming wireless applications, the insufficient battery capacity of mobile terminals has always been a bottleneck that cannot be broken through. Therefore, how to reduce the energy consumption when terminals upload cloud service data has gradually become the focus of attention.

The energy consumption of the terminal uploading cloud service data is essentially the accumulation of transmission power over time. Transmission power is related to transmission rate and channel state. When the total amount of uploaded data remains unchanged, the greater the transmission rate, the greater the transmission power and the shorter the transmission time. Therefore, the optimization of energy consumption is actually the scheduling of the transmission rate. Existing researches have made many efforts to reduce the energy consumption of cloud services along the above lines. It has successively proposed to optimize the upload energy of cloud services in single-channel and multi-channel scenarios to obtain the optimal rate scheduling scheme, and found that the upload energy of cloud services decreases with the increase of the number of channels.

However, all existing literatures only focus on the case of a single cloud service in the system, and the upload rate of cloud services in the actual system is bound to be limited by the spectrum bandwidth occupied by cloud services, in addition to obeying the scheduling strategy of "energy optimal". The bandwidth occupied by a service must be related to other services in the system. Therefore, considering the cloud service scheduling strategy in the scenario of multi-service coexistence is in line with the actual situation, and there is no research in this area yet.

SUMMARY OF THE INVENTION

The technical problem to be solved by the present invention is to provide a dual-objective optimization model of resource allocation and cloud service rate scheduling under the consideration of common service QoS and system throughput, and obtain the cloud service optimal rate scheduling strategy and cloud service rate scheduling strategy through the optimization algorithm. An optimal scheduling method for cloud service energy consumption in a multi-service coexistence system with optimal channel allocation strategy.

In order to solve the above technical problems, the present invention adopts the following technical solutions: the present invention designs a method for optimizing the energy consumption of cloud services in a multi-service coexistence system, and realizes the optimal scheduling of energy consumption of cloud services in a system in which ordinary services and cloud services coexist, It includes the following steps:

Step A. Respectively for each common service and each cloud service, adopt Shannon's formula to obtain the relationship between transmission rate and transmission power, wherein, the channel gain obeys IID, and then obtains the relationship between each service transmission rate and transmission power respectively , and then go to step B;

Step B. For each cloud service, according to the relationship between the service transmission rate and transmission power, obtain a cloud service energy consumption model based on the service transmission rate, bandwidth, and channel gain, and then obtain the cloud service energy consumption model of each cloud service respectively. , and then enter step C;

Step C. For each cloud service, respectively, for each working time slot corresponding to the cloud service, first, according to the cloud service energy consumption model of the cloud service, adopt the dynamic programming method to obtain the last work time slot corresponding to the cloud service according to the time sequence. Then, using the reverse order recursion method, according to the channel gain in the current working time slot, the energy consumption value of the cloud service time slot corresponding to each previous working time slot of the cloud service according to the time sequence is obtained in turn. function; and then obtain the cloud service time slot energy consumption value function of each work time slot corresponding to each cloud service, and then enter step D;

Step D. For each cloud service, and for the cloud service time slot energy consumption value function of each work time slot corresponding to the cloud service, obtain the optimal rate of each work time slot corresponding to the cloud service, and then obtain the corresponding cloud service. Minimum global energy consumption model, thereby obtaining the minimum global energy consumption model corresponding to each cloud service, and then entering step E1;

Step E1. According to the number of channels to be allocated in the current time slot, obtain all channel allocation schemes to be accessed for each cloud service and each ordinary service for the current time slot, and then enter step E2;

Step E2. For each channel allocation scheme of the current time slot, according to the minimum global energy consumption model corresponding to each cloud service, obtain the minimum global energy consumption of each cloud service connected to the current time slot under the channel allocation scheme, and obtain at the same time. Under the channel allocation scheme, the sum of the transmission rates of the ordinary services connected to the current time slot is the throughput of the current time slot of the system; and then the minimum global energy consumption of each cloud service connected to the current time slot under each channel allocation scheme is obtained, and the current time slot throughput of the system, and then enter step E3;

Step E3. For all the channel allocation schemes of the current time slot, exclude the existence of the channel allocation scheme with the minimum global energy consumption corresponding to the cloud service higher than the preset cloud service energy consumption upper limit value, and in the remaining channel allocation scheme, select the current channel allocation scheme of the system. The channel allocation scheme corresponding to the maximum time slot throughput is taken as the optimal channel allocation scheme for the current time slot, and then enters step E4;

Step E4. Adopt the optimal channel allocation scheme for the current time slot and the minimum global energy consumption model corresponding to each cloud service to realize the optimal scheduling of the energy consumption of the cloud service in the current time slot.

As a preferred technical solution of the present invention: the step A to the step B specifically include establishing the following model:

s.t.

in, is the system throughput at time slot t, The common service set selected for time slot t, is the channel set allocated by the base station to the common service m (m∈M _t ) for the time slot t, and g _m,k,t is the channel gain of the channel k allocated to the common service m by the time slot t; is the mobile phone energy consumption consumed by cloud service i in the entire data upload time slot, R _i,t is the rate of cloud service i in time slot t, Δt is the time slot interval, is the channel set allocated by the base station to cloud service i for time slot t, Represents the mean value of each channel gain, and K _i is The number of channels in the middle, g _{i, k, t} is the channel gain of channel k allocated to cloud service i in time slot t; P _{i, t} is the transmission power of cloud service i in time slot t; Indicates that channels cannot be assigned repeatedly; K represents the number of channels in the system, and the corresponding channel set is K={1,...k,...K}; M represents the number of common services in the system, and the corresponding service set is M={ 1,...m,...M}; the number I of cloud services in the system, the corresponding service set is I=={1,...i,...I}, the amount of data to be uploaded by cloud service _i is Li, and the uploaded data The time limit is T _i , and the moment to start uploading data is t=ΔT _i .

As a preferred technical solution of the present invention: the step C to the step D specifically includes the following operations:

The energy consumption optimization model of cloud service i is as follows:

s.t.

The optimization model is rewritten using the value function as,

Among them, S _t is the amount of decision-making, which refers to the specific decision in each stage, that is, the amount of data to be sent in this stage; L _t is the state variable, which refers to the amount of data remaining in each stage (including this stage); is an index function, which is a quantitative index to measure a decision-making process, here refers to the minimum energy consumption index;

Using mathematical induction, the optimal rate of cloud service i in time slot t is finally obtained as,

The minimum global energy consumption model of cloud service i is,

in,

As a preferred technical solution of the present invention: the steps E1 to E4 specifically include using the 0-1 integer programming algorithm to solve the channel allocation scheme that maximizes the system throughput as follows:

The optimization problem is transformed into a 0-1 integer programming problem, i.e.,

s.t.

[A ₁ ,...,A _N ]Χ=1 _K

where Χ=[Χ ₁ ,...,Χ _N ] ^T is a decision column vector of size NC, Χ=[Χ ₁ ,...,Χ _N ] ^T , Χn=[x _{n ,1} , ...,x _n,C ] ^T , x _n,j ∈{0,1}, when x _n,j is "1", it means that business n adopts the allocation scheme corresponding to the jth column in the allocation matrix, otherwise it means not to use ; Represents the number of possible allocation schemes for a business; e is a weight matrix of size N×C, and its elements e _n,j represent the contribution of business n to the optimization goal when the allocation scheme corresponding to the jth column in the allocation matrix is used, that is

A _n is a channel allocation matrix composed of elements 0 and 1 with a size of K×C. "1" means that the corresponding channel is allocated to the service, and "0" means no allocation. For example, when there are K=3 channels in total, then Each service has C=7 possible allocation schemes, and the channel allocation matrix of service n is:

The optimal solution of the optimization problem is obtained by the exhaustive method, that is, the optimal channel allocation scheme for the current time slot, where K indicates that the number of channels in the system is K, and the corresponding channel set is K={1,...k, ...K}; M represents the number of common services in the system, and the corresponding service set is M={1,...m,...M}; the number of cloud services in the system I, the corresponding service set is I={1, ... _i ,...I}, the amount of data to be uploaded by cloud service _{i is Li, the time limit for uploading data is Ti, and the moment when data starts to be uploaded is t=ΔT i .}

Compared with the prior art, the method for optimal scheduling of cloud service energy consumption in a multi-service coexistence system according to the present invention adopts the above technical solution and has the following technical effects: , considering common service QoS and system throughput, a dual-objective optimization model for spectrum resource allocation that can simultaneously optimize the energy consumption of multiple cloud services and maximize system throughput is established. The minimum objective of uploading energy for cloud services is rewritten as a constraint that the energy is less than a certain threshold, so that the dual-objective optimization problem becomes a single-objective optimization problem, and then the optimal energy consumption and the channel set allocated to the current time slot are obtained through the reverse order iteration method. Through this relationship, the two goals of optimizing system throughput and optimizing cloud service energy are unified into one time scale, and finally the channel allocation is carried out through the designed 0-1 integer programming algorithm. Maximize system throughput under the premise of business energy consumption requirements.

Corresponding to the above, the technical problem to be solved by the present invention is to provide a dual-objective optimization model for resource allocation and cloud service rate scheduling under the consideration of common service QoS and system throughput, and obtain the optimal cloud service through the optimization algorithm. An optimal scheduling method for cloud service energy consumption in a multi-service coexistence system with optimal rate scheduling strategy and channel optimal allocation strategy.

In order to solve the above technical problems, the present invention adopts the following technical solutions: the present invention designs a method for optimizing the energy consumption of cloud services in a multi-service coexistence system, and realizes the optimal scheduling of energy consumption of cloud services in a system in which ordinary services and cloud services coexist, Wherein, the upper limit value of cloud service energy consumption is preset, and the cloud service energy consumption optimization scheduling method includes the following steps:

Step A. Respectively for each common service and each cloud service, adopt Shannon's formula to obtain the relationship between transmission rate and transmission power, wherein, the channel gain obeys IID, and then obtains the relationship between each service transmission rate and transmission power respectively , and then go to step B;

Step B. For each cloud service, according to the relationship between the service transmission rate and transmission power, obtain a cloud service energy consumption model based on the service transmission rate, bandwidth, and channel gain, and then obtain the cloud service energy consumption model of each cloud service respectively. , and then enter step C;

Step C. For each cloud service, respectively, for each working time slot corresponding to the cloud service, first, according to the cloud service energy consumption model of the cloud service, adopt the dynamic programming method to obtain the last work time slot corresponding to the cloud service according to the time sequence. Then, using the reverse order recursion method, according to the channel gain in the current working time slot, the energy consumption value of the cloud service time slot corresponding to each previous working time slot of the cloud service according to the time sequence is obtained in turn. function; and then obtain the cloud service time slot energy consumption value function of each work time slot corresponding to each cloud service, and then enter step D;

Step D. For each cloud service, and for the cloud service time slot energy consumption value function of each work time slot corresponding to the cloud service, obtain the optimal rate of each work time slot corresponding to the cloud service, and then obtain the corresponding cloud service. Minimum global energy consumption model, thereby obtaining the minimum global energy consumption model corresponding to each cloud service, and then entering step F1;

Step F1. Rewrite the system maximum time slot throughput optimization problem based on the preset cloud service energy consumption upper limit value into a corresponding Lagrangian dual function, and enter step F2;

Step F2. The dual problem of the original optimization problem is represented by the Lagrangian dual algorithm, the original optimization problem and the dual problem have the same solution, and the final solution is obtained by solving the dual problem, and then the step F3 is entered;

Step F3. Use the greedy algorithm to allocate the channel of the current time slot. The allocation principle is: a channel should be allocated to the service that can make the Lagrangian function increase the largest, and then enter Step F4;

Step F4. Use the bisection method to solve the optimal dual coefficient, and then obtain the optimal channel allocation scheme for the current time slot, and then enter step F5;

Step F5. Adopt the optimal channel allocation scheme for the current time slot and the minimum global energy consumption model corresponding to each cloud service to realize the optimal scheduling of the energy consumption of the cloud service in the current time slot.

As a preferred technical solution of the present invention: the step A to the step B specifically include establishing the following model:

s.t.

in, is the system throughput at time slot t, The common service set selected for time slot t, is the channel set allocated by the base station to the common service m (m∈M _t ) for the time slot t, and g _m,k,t is the channel gain of the channel k allocated to the common service m by the time slot t; is the mobile phone energy consumption consumed by cloud service i in the entire data upload time slot, R _i,t is the rate of cloud service i in time slot t, Δt is the time slot interval, is the channel set allocated by the base station to cloud service i for time slot t, Represents the mean value of each channel gain, and K _i is The number of channels in the middle, g _{i, k, t} is the channel gain of channel k allocated to cloud service i in time slot t; P _{i, t} is the transmission power of cloud service i in time slot t; Indicates that channels cannot be assigned repeatedly; K represents the number of channels in the system, and the corresponding channel set is K={1,...k,...K}; M represents the number of common services in the system, and the corresponding service set is M={ 1,...m,...M}; the number I of cloud services in the system, the corresponding service set is I={1,...i,...I}, the amount of data to be uploaded by cloud service _i is Li, and the number of uploaded data is Li. The time limit is T _i , and the time to start uploading data is t=ΔT _i .

As a preferred technical solution of the present invention: the step C to the step D specifically includes the following operations:

The energy consumption optimization model of cloud service i is as follows:

s.t.

The optimization model is rewritten using the value function as,

Among them, S _t is the amount of decision-making, which refers to the specific decision in each stage, that is, the amount of data to be sent in this stage; L _t is the state variable, which refers to the amount of data remaining in each stage (including this stage); is an index function, which is a quantitative index to measure a decision-making process, here refers to the minimum energy consumption index;

Using mathematical induction, the optimal rate of cloud service i in time slot t is finally obtained as,

The minimum global energy consumption model of cloud service i is,

in,

As a preferred technical solution of the present invention: the steps F1 to F4 specifically include the following:

Based on the preset upper limit value of cloud service energy consumption, the optimization problem of the maximum time slot throughput of the system is rewritten into the corresponding Lagrangian function as:

Its dual function is,

The corresponding dual problem is

s.t.

α _t ,β _t ≥0

To solve the dual problem, first use the greedy algorithm to allocate channels, and then use the bisection method to solve the optimal dual coefficient. The specific steps are:

i. Initialization

ii. Order

iii. For any service n, traverse any channel k that can be allocated to it, and perform the above for all services

Operation to obtain the incremental value of L(M _t , K _m,t , α _t , β _t ) when service n is allocated to channel k, set as Δw _n,k , satisfying

Find the (n ^* ,k ^* ) that maximizes Δwn _,k , and assign the corresponding k ^* to n ^* ;

iv. Repeat step iii until all channels are allocated;

v. Under the channel allocation scheme obtained above, calculate the value of ε _i -E _i ^* (K _i ), if ε _i -E _i ^* (K _i )≥0, then the corresponding otherwise calculate value, if then the corresponding otherwise

Repeat steps ii-v until for and for Among them, δ is the constant we set to control the accuracy of the algorithm, the smaller the δ, the higher the accuracy of the algorithm, where K represents the number of channels in the system is K, and the corresponding channel set is K={1,...k ,...K}; M represents the number of common services in the system, and the corresponding service set is M={1,...m,...M}; the number of cloud services in the system I, the corresponding service set is I={1 ,...i,...I}, the amount of data to be uploaded by cloud service _i is _Li , the time limit for uploading data is Ti, and the moment when data uploading starts is t=ΔT _i .

Compared with the prior art, the method for optimal scheduling of cloud service energy consumption in a multi-service coexistence system according to the present invention adopts the above technical solution and has the following technical effects: , considering common service QoS and system throughput, a dual-objective optimization model for spectrum resource allocation that can simultaneously optimize the energy consumption of multiple cloud services and maximize system throughput is established. The minimum objective of uploading energy for cloud services is rewritten as a constraint that the energy is less than a certain threshold, so that the dual-objective optimization problem becomes a single-objective optimization problem, and then the optimal energy consumption and the channel set allocated to the current time slot are obtained through the reverse order iteration method. Through this relationship, the two goals of optimizing system throughput and optimizing cloud service energy are unified into one time scale, and finally the channel is allocated through the designed Lagrangian dual algorithm. Maximize system throughput under the premise of business energy consumption requirements.

Description of drawings

Fig. 1 is the schematic flow chart of the cloud service energy consumption optimization scheduling method in the multi-service coexistence system designed by the present invention;

Figure 2 is a graph showing the variation of system throughput with transmission time slots under the 0-1 integer programming algorithm and the Lagrangian dual algorithm;

Figure 3 is a graph showing the change of cloud service energy consumption with the deadline under the 0-1 integer programming algorithm and the Lagrangian dual algorithm;

Figure 4 is a graph showing the change of cloud service energy consumption with data volume under the 0-1 integer programming algorithm and the Lagrangian dual algorithm;

Figure 5 is a comparison diagram of computational complexity under the 0-1 integer programming algorithm and the Lagrangian dual algorithm.

Detailed ways

The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

Cloud service refers to a type of service that uploads data to the cloud for processing, and its QoS requirements are to complete the data transmission of the corresponding amount of data with minimum energy consumption within a certain deadline; ordinary services are all types of services other than cloud services. A general term for services, and its QoS requirements are that the real-time rate should not be less than a certain rate requirement.

The present invention designs a cloud service energy consumption optimization scheduling method in a multi-service coexistence system. In a system where common services and cloud services coexist, the cloud service energy consumption optimization scheduling is realized. In practical applications, the following two embodiments are specifically included: Wherein, as shown in Figure 1, the first embodiment includes the following steps:

Step A. Respectively for each common service and each cloud service, adopt Shannon's formula to obtain the relationship between transmission rate and transmission power, wherein, the channel gain obeys IID, and then obtains the relationship between each service transmission rate and transmission power respectively , then go to step B.

Step B. For each cloud service, according to the relationship between the service transmission rate and transmission power, obtain a cloud service energy consumption model based on the service transmission rate, bandwidth, and channel gain, and then obtain the cloud service energy consumption model of each cloud service respectively. , then go to step C.

Above-mentioned step A to step B, specifically include establishing the following model:

s.t.

in, is the system throughput at time slot t, The common service set selected for time slot t, is the channel set allocated by the base station to the common service m (m∈M _t ) for the time slot t, and g _m,k,t is the channel gain of the channel k allocated to the common service m by the time slot t; is the mobile phone energy consumption consumed by cloud service i in the entire data upload time slot, R _{i t} is the rate of cloud service i in time slot t, Δt is the time slot interval, is the channel set allocated by the base station to cloud service i for time slot t, Represents the mean value of each channel gain, and K _i is The number of channels in the middle, g _{i, k, t} is the channel gain of channel k allocated to cloud service i in time slot t; P _{i, t} is the transmission power of cloud service i in time slot t; Indicates that channels cannot be assigned repeatedly; K represents the number of channels in the system, and the corresponding channel set is K={1,...k,...K}; M represents the number of common services in the system, and the corresponding service set is M={ 1,...m,...M}; the number I of cloud services in the system, the corresponding service set is I={1,...i,...I}, the amount of data to be uploaded by cloud service _i is Li, and the number of uploaded data is Li. The time limit is T _i , and the time to start uploading data is t=ΔT _i .

Step C. For each cloud service, respectively, for each working time slot corresponding to the cloud service, first, according to the cloud service energy consumption model of the cloud service, adopt the dynamic programming method to obtain the last work time slot corresponding to the cloud service according to the time sequence. Then, using the reverse order recursion method, according to the channel gain in the current working time slot, the energy consumption value of the cloud service time slot corresponding to each previous working time slot of the cloud service according to the time sequence is obtained in turn. function; and then obtain the cloud service time slot energy consumption value function of each work time slot corresponding to each cloud service, and then enter step D.

Step D. For each cloud service, and for the cloud service time slot energy consumption value function of each work time slot corresponding to the cloud service, obtain the optimal rate of each work time slot corresponding to the cloud service, and then obtain the corresponding cloud service. Minimum global energy consumption model, thereby obtaining the minimum global energy consumption model corresponding to each cloud service, and then entering step E1.

Above-mentioned steps C to step D, specifically include the following operations:

The energy consumption optimization model of cloud service i is as follows:

s.t.

The optimization model is rewritten using the value function as:

Among them, S _t is the amount of decision-making, which refers to the specific decision in each stage, that is, the amount of data to be sent in this stage; L _t is the state variable, which refers to the amount of data remaining in each stage (including this stage); is the index function, which is a quantitative index to measure a decision-making process, here refers to the minimum energy consumption index.

Using mathematical induction, the optimal rate of cloud service i in time slot t is finally obtained as,

The minimum global energy consumption model of cloud service i is,

in,

Step E1. According to the number of channels to be allocated in the current time slot, obtain all channel allocation schemes for each cloud service and each common service to be accessed in the current time slot, and then proceed to Step E2.

Step E2. For each channel allocation scheme of the current time slot, according to the minimum global energy consumption model corresponding to each cloud service, obtain the minimum global energy consumption of each cloud service connected to the current time slot under the channel allocation scheme, and obtain at the same time. Under the channel allocation scheme, the sum of the transmission rates of the ordinary services connected to the current time slot is the throughput of the current time slot of the system; and then the minimum global energy consumption of each cloud service connected to the current time slot under each channel allocation scheme is obtained, and the current time slot throughput of the system, and then enter step E3.

Step E3. For all the channel allocation schemes of the current time slot, exclude the existence of the channel allocation scheme with the minimum global energy consumption corresponding to the cloud service higher than the preset cloud service energy consumption upper limit value, and in the remaining channel allocation scheme, select the current channel allocation scheme of the system. The channel allocation scheme corresponding to the maximum time slot throughput is regarded as the optimal channel allocation scheme for the current time slot, and then the process goes to step E4.

Step E4. Adopt the optimal channel allocation scheme for the current time slot and the minimum global energy consumption model corresponding to each cloud service to realize the optimal scheduling of the energy consumption of the cloud service in the current time slot.

The above steps E1 to E4 specifically include using the 0-1 integer programming algorithm to solve the channel allocation scheme that maximizes the system throughput as follows:

The optimization problem is transformed into a 0-1 integer programming problem, i.e.,

s.t.

[A ₁ ,...,A _N ]Χ=1 _K

where Χ=[Χ ₁ ,...,Χ _N ] ^T is a decision column vector of size NC, Χ=[Χ ₁ ,...,Χ _N ] ^T , Χn=[x _{n ,1} , ...,x _n,C ] ^T , x _n,j ∈{0,1}, when x _n,j is "1", it means that business n adopts the allocation scheme corresponding to the jth column in the allocation matrix, otherwise it means not to use ; Represents the number of possible allocation schemes for a business; e is a weight matrix of size N×C, and its elements e _n,j represent the contribution of business n to the optimization goal when the allocation scheme corresponding to the jth column in the allocation matrix is used, that is

A _n is a channel allocation matrix composed of elements 0 and 1 with a size of K×C. "1" means that the corresponding channel is allocated to the service, and "0" means no allocation. For example, when there are K=3 channels in total, then Each service has C=7 possible allocation schemes, and the channel allocation matrix of service n is:

The optimal solution of the optimization problem is obtained by the exhaustive method, that is, the optimal channel allocation scheme for the current time slot, where K indicates that the number of channels in the system is K, and the corresponding channel set is K={1,...k, ...K}; M represents the number of common services in the system, and the corresponding service set is M={1,...m,...M}; the number of cloud services in the system I, the corresponding service set is I={1, ... _i ,...I}, the amount of data to be uploaded by cloud service _{i is Li, the time limit for uploading data is Ti, and the moment when data starts to be uploaded is t=ΔT i .}

Also as shown in FIG. 1 , in the second embodiment, the upper limit value of cloud service energy consumption is preset, and the cloud service energy consumption optimization scheduling method specifically includes the following steps:

Step A. Respectively for each common service and each cloud service, adopt Shannon's formula to obtain the relationship between transmission rate and transmission power, wherein, the channel gain obeys IID, and then obtains the relationship between each service transmission rate and transmission power respectively , then go to step B.

Step B. For each cloud service, according to the relationship between the service transmission rate and transmission power, obtain a cloud service energy consumption model based on the service transmission rate, bandwidth, and channel gain, and then obtain the cloud service energy consumption model of each cloud service respectively. , then go to step C.

Wherein, step A to step B, specifically includes establishing the following model:

s.t.

in, is the system throughput at time slot t, The common service set selected for time slot t, is the channel set allocated by the base station to the common service m (m∈M _t ) for the time slot t, and g _m,k,t is the channel gain of the channel k allocated to the common service m by the time slot t; is the mobile phone energy consumption consumed by cloud service i in the entire data upload time slot, R _i,t is the rate of cloud service i in time slot t, Δt is the time slot interval, is the channel set allocated by the base station to cloud service i for time slot t, Represents the mean value of each channel gain, and K _i is The number of channels in the middle, g _{i, k, t} is the channel gain of channel k allocated to cloud service i in time slot t; P _{i, t} is the transmission power of cloud service i in time slot t; Indicates that channels cannot be assigned repeatedly; K represents the number of channels in the system, and the corresponding channel set is K={1,...k,...K}; M represents the number of common services in the system, and the corresponding service set is M={ 1,...m,...M}; the number I of cloud services in the system, the corresponding service set is I={1,...i,...I}, the amount of data to be uploaded by cloud service _i is Li, and the number of uploaded data is Li. The time limit is T _i , and the time to start uploading data is t=ΔT _i .

Step C. For each cloud service, respectively, for each working time slot corresponding to the cloud service, first, according to the cloud service energy consumption model of the cloud service, adopt the dynamic programming method to obtain the last work time slot corresponding to the cloud service according to the time sequence. Then, using the reverse order recursion method, according to the channel gain in the current working time slot, the energy consumption value of the cloud service time slot corresponding to each previous working time slot of the cloud service according to the time sequence is obtained in turn. function; and then obtain the cloud service time slot energy consumption value function of each work time slot corresponding to each cloud service, and then enter step D.

Step D. For each cloud service, and for the cloud service time slot energy consumption value function of each work time slot corresponding to the cloud service, obtain the optimal rate of each work time slot corresponding to the cloud service, and then obtain the corresponding cloud service. Minimum global energy consumption model, thereby obtaining the minimum global energy consumption model corresponding to each cloud service, and then entering step F1.

Wherein, step C to step D, specifically include the following operations:

The energy consumption optimization model of cloud service i is as follows:

s.t.

The optimization model is rewritten using the value function as,

Among them, S _t is the amount of decision-making, which refers to the specific decision in each stage, that is, the amount of data to be sent in this stage; L _t is the state variable, which refers to the amount of data remaining in each stage (including this stage); is an index function, which is a quantitative index to measure a decision-making process, here refers to the minimum energy consumption index;

Using mathematical induction, the optimal rate of cloud service i in time slot t is finally obtained as,

The minimum global energy consumption model of cloud service i is,

in,

Step F1. Rewrite the system maximum timeslot throughput optimization problem based on the preset cloud service energy consumption upper limit value into a corresponding Lagrangian dual function, and enter step F2.

Step F2. The dual problem of the original optimization problem is represented by the Lagrangian dual algorithm. The original optimization problem and the dual problem have the same solution, and the final solution is obtained by solving the dual problem, and then the step F3 is entered.

Step F3. Use the greedy algorithm to allocate the channel of the current time slot. The allocation principle is: a channel should be allocated to the service that can make the Lagrangian function increase the most, and then go to Step F4.

Step F4. Use the bisection method to solve the optimal dual coefficient, and then obtain the optimal channel allocation scheme for the current time slot, and then go to Step F5.

Step F5. Adopt the optimal channel allocation scheme for the current time slot and the minimum global energy consumption model corresponding to each cloud service to realize the optimal scheduling of the energy consumption of the cloud service in the current time slot.

Wherein, steps F1 to F4 specifically include the following:

Based on the preset upper limit value of cloud service energy consumption, the optimization problem of the maximum time slot throughput of the system is rewritten into the corresponding Lagrangian function as:

Its dual function is,

The corresponding dual problem is

s.t.

α _t ,β _t ≥0

To solve the dual problem, first use the greedy algorithm to allocate channels, and then use the bisection method to solve the optimal dual coefficient. The specific steps are:

vi. Initialize

vii.

viii. For any service n, traverse any channel k that can be allocated to it, and perform the above operations for all services to obtain L(M _t ,K _m,t ,α _t ,β _t ) when service n is allocated to channel k The incremental value of , set as Δw _n,k , satisfying

Find the (n ^* ,k ^* ) that maximizes Δwn _,k , and assign the corresponding k ^* to n ^* ;

ix. Repeat step iii until all channels are allocated;

x. Under the channel allocation scheme obtained above, calculate the value of ε _i -E _i ^* (K _i ), if ε _i -E _i ^* (K _i )≥0, then the corresponding otherwise calculate value, if then the corresponding otherwise

Repeat steps ii-v until for and for Among them, δ is the constant we set to control the accuracy of the algorithm, the smaller the δ, the higher the accuracy of the algorithm, where K represents the number of channels in the system is K, and the corresponding channel set is K={1,...k ,...K}; M represents the number of common services in the system, and the corresponding service set is M={1,...m,...M}; the number of cloud services in the system I, the corresponding service set is I={1 ,...i,...I}, the amount of data to be uploaded by cloud service _i is _Li , the time limit for uploading data is Ti, and the moment when data uploading starts is t=ΔT _i .

For the allocation of the last channel in the cloud service energy consumption optimization scheduling method in the multi-service coexistence system designed by the present invention, the 0-1 integer programming algorithm and the Lagrangian dual algorithm are specifically designed to allocate channels, as shown in Figure 2 , Figure 3, Figure 4, Figure 5, it can be seen that the two algorithms have different effects in practical applications.

The cloud service energy consumption optimization scheduling method in the multi-service coexistence system designed by the present invention takes into account the common service QoS and system throughput, and establishes a spectrum resource allocation system that can simultaneously optimize the energy consumption of multiple cloud services and maximize the system throughput. The dual-objective optimization model introduces the upper limit of the uploading energy of cloud services, and rewrites the minimum objective of uploading energy of cloud services as the restriction condition that the energy is less than a certain threshold, so that the dual-objective optimization problem becomes a single-objective optimization problem. method to obtain the relationship between the optimal energy consumption and the channel set allocated to the current time slot, and through this relationship, the two goals of optimizing system throughput and optimizing cloud service energy are unified on a time scale, and finally through The designed 0-1 integer programming algorithm or Lagrangian dual algorithm is used to allocate channels and maximize the system throughput under the premise of meeting the energy consumption requirements of cloud services.

The embodiments of the present invention have been described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned embodiments, and can also be made within the scope of knowledge possessed by those of ordinary skill in the art without departing from the purpose of the present invention. Various changes.

Claims (6)

1. A method for optimal scheduling of cloud service energy consumption in a multi-service coexistence system, in a system where common services and cloud services coexist, realizes cloud service energy consumption optimal scheduling, it is characterized in that, comprises the following steps: Step A. Respectively for each common service and each cloud service, adopt Shannon's formula to obtain the relationship between transmission rate and transmission power, wherein, the channel gain obeys IID, and then obtains the relationship between each service transmission rate and transmission power respectively , and then go to step B; Step B. For each cloud service, according to the relationship between the service transmission rate and transmission power, obtain a cloud service energy consumption model based on the service transmission rate, bandwidth, and channel gain, and then obtain the cloud service energy consumption model of each cloud service respectively. , and then enter step C; Step C. For each cloud service, respectively, for each working time slot corresponding to the cloud service, first, according to the cloud service energy consumption model of the cloud service, adopt the dynamic programming method to obtain the last work time slot corresponding to the cloud service according to the time sequence. Then, using the reverse order recursion method, according to the channel gain in the current working time slot, the energy consumption value of the cloud service time slot corresponding to each previous working time slot of the cloud service according to the time sequence is obtained in turn. function; and then obtain the cloud service time slot energy consumption value function of each work time slot corresponding to each cloud service, and then enter step D; Step D. For each cloud service, and for the cloud service time slot energy consumption value function of each work time slot corresponding to the cloud service, obtain the optimal rate of each work time slot corresponding to the cloud service, and then obtain the corresponding cloud service. Minimum global energy consumption model, thereby obtaining the minimum global energy consumption model corresponding to each cloud service, and then entering step E1; Step E1. According to the number of channels to be allocated in the current time slot, obtain all channel allocation schemes to be accessed for each cloud service and each ordinary service for the current time slot, and then enter step E2; Step E2. For each channel allocation scheme of the current time slot, according to the minimum global energy consumption model corresponding to each cloud service, obtain the minimum global energy consumption of each cloud service connected to the current time slot under the channel allocation scheme, and obtain at the same time. Under the channel allocation scheme, the sum of the transmission rates of the ordinary services connected to the current time slot is the throughput of the current time slot of the system; and then the minimum global energy consumption of each cloud service connected to the current time slot under each channel allocation scheme is obtained, and the current time slot throughput of the system, and then enter step E3; Step E3. For all the channel allocation schemes of the current time slot, exclude the existence of the channel allocation scheme with the minimum global energy consumption corresponding to the cloud service higher than the preset cloud service energy consumption upper limit value, and in the remaining channel allocation scheme, select the current channel allocation scheme of the system. The channel allocation scheme corresponding to the maximum time slot throughput is taken as the optimal channel allocation scheme for the current time slot, and then enters step E4; Step E4. Adopt the optimal channel allocation scheme for the current time slot and the minimum global energy consumption model corresponding to each cloud service to realize the optimal scheduling of the energy consumption of the cloud service in the current time slot. 2. The method for optimal scheduling of cloud service energy consumption in a multi-service coexistence system according to claim 1, wherein the step E1 to the step E4 specifically include using a 0-1 integer programming algorithm to solve the problem so that the system throughput is maximized The channel allocation scheme is as follows: The optimization problem is transformed into a 0-1 integer programming problem, i.e., s.t. [A ₁ ,...,A _N ]X=1 _K Among them, N is the total number of services in the system, K is the total number of channels in the system; X=[X ₁ ,...,X _N ] ^T is a decision column vector of size NC, X=[X ₁ ,...,X _N ] ^T , X _n =[x _n,1 ,...,x _n,C ] ^T , when x _n,j ∈{0,1}, when x _n,j is "1" Indicates that service n adopts the allocation scheme corresponding to the jth column in the allocation matrix, otherwise it means not to use it. Since the channels allocated to the same service must be adjacent, if r adjacent channels are allocated to a service, there should be K- (r-1) allocation schemes, where K is the total number of channels in the system, so Indicates the total number of possible allocation schemes for a business; A _n is a channel allocation matrix composed of elements 0 and 1 with a size of K×C. "1" means that the corresponding channel is allocated to the service, and "0" means no allocation. For example, when there are K=3 channels in total, then Each service has C=7 possible allocation schemes, and the channel allocation matrix of service n is: e is a weight matrix of size N×C, and its elements e _n,j indicate that business n adopts the allocation scheme corresponding to the jth column in the allocation matrix contribution to the optimization objective when Among them, in the first expression, is the channel set allocated to cloud service i, It is the total number of cloud services that have been activated before time slot _t , this expression indicates that the channel set allocated to the activated cloud service should remain unchanged; in the second expression , ε _i is the preset maximum energy limit, is the total number of cloud services activated for the first time in time slot t, and E _i,j * represents the minimum energy consumption when the allocation scheme corresponding to the jth column in the allocation matrix is allocated to cloud service i, namely This expression indicates that the energy consumed by cloud service i should be less than the maximum energy limit; in the third expression, R _m,j,t represents the transmission when the allocation scheme corresponding to the jth column in the allocation matrix is allocated to the ordinary service m rate, that is This expression shows that if the transmission rate of the common service m is greater than its minimum rate limit, the target will be optimized. If the common service m is not selected, it will not affect the target. The exhaustive method is used to find the optimal solution of the optimization problem. solution, that is, the optimal channel allocation scheme for the current time slot. 3. A method for optimal scheduling of cloud service energy consumption in a multi-service coexisting system, in a system in which ordinary services and cloud services coexist, to realize optimal scheduling of cloud service energy consumption, characterized in that a cloud service energy consumption upper limit value is preset, The cloud service energy consumption optimization scheduling method includes the following steps: Step A. Respectively for each common service and each cloud service, adopt Shannon's formula to obtain the relationship between transmission rate and transmission power, wherein, the channel gain obeys IID, and then obtains the relationship between each service transmission rate and transmission power respectively , and then go to step B; Step B. For each cloud service, according to the relationship between the service transmission rate and transmission power, obtain a cloud service energy consumption model based on the service transmission rate, bandwidth, and channel gain, and then obtain the cloud service energy consumption model of each cloud service respectively. , and then enter step C; Step C. For each cloud service, respectively, for each working time slot corresponding to the cloud service, first, according to the cloud service energy consumption model of the cloud service, adopt the dynamic programming method to obtain the last work time slot corresponding to the cloud service according to the time sequence. Then, using the reverse order recursion method, according to the channel gain in the current working time slot, the energy consumption value of the cloud service time slot corresponding to each previous working time slot of the cloud service according to the time sequence is obtained in turn. function; and then obtain the cloud service time slot energy consumption value function of each work time slot corresponding to each cloud service, and then enter step D; Step D. For each cloud service, and for the cloud service time slot energy consumption value function of each work time slot corresponding to the cloud service, obtain the optimal rate of each work time slot corresponding to the cloud service, and then obtain the corresponding cloud service. Minimum global energy consumption model, thereby obtaining the minimum global energy consumption model corresponding to each cloud service, and then entering step F1; Step F1. Rewrite the system maximum time slot throughput optimization problem based on the preset cloud service energy consumption upper limit value into a corresponding Lagrangian dual function, and enter step F2; Step F2. The dual problem of the original optimization problem is represented by the Lagrangian dual algorithm, the original optimization problem and the dual problem have the same solution, and the final solution is obtained by solving the dual problem, and then the step F3 is entered; Step F3. Use the greedy algorithm to allocate the channel of the current time slot. The allocation principle is: a channel should be allocated to the service that can make the Lagrangian function increase the largest, and then enter Step F4; Step F4. Use the bisection method to solve the optimal dual coefficient, and then obtain the optimal channel allocation scheme for the current time slot, and then enter step F5; Step F5. Adopt the optimal channel allocation scheme for the current time slot and the minimum global energy consumption model corresponding to each cloud service to realize the optimal scheduling of the energy consumption of the cloud service in the current time slot. 4. The method for optimizing and scheduling cloud service energy consumption in a multi-service coexistence system according to claim 1 or 3, wherein the step A to the step B specifically includes establishing the following model: s.t. in, is the system throughput at time slot t, The common service set selected for time slot t, Allocate the base station for time slot t to normal service The channel set of , g _{m, k, t} is the channel gain of channel k allocated to ordinary service m by time slot t; is the mobile phone energy consumption consumed by cloud service i in the entire data upload time slot, R _i,t is the rate of cloud service i in time slot t, Δt is the time slot interval, is the channel set allocated by the base station to cloud service i for time slot t, Represents the mean value of each channel gain, and K _i is The number of channels in , g _{i, k, t} is the channel gain of channel k allocated to cloud service i by time slot t; is the preset minimum rate requirement of ordinary service m, P _i,t is the transmission power of cloud service i in time slot t, and P _i,max is the preset maximum transmission power requirement of cloud service i; Indicates that the channel cannot be repeatedly allocated; K represents the number of channels in the system, and the corresponding channel set is M represents the number of common services in the system, and the corresponding service set is The number I of cloud services in the system, and the corresponding service set is The amount of data to be uploaded by the cloud service _i is _Li , the time limit for uploading data is Ti, and the moment of starting to upload data is t=ΔT _i . 5. The method for optimizing and scheduling cloud service energy consumption in a multi-service coexistence system according to claim 4, wherein the steps C to D specifically include the following operations: The energy consumption optimization model of cloud service i is as follows: s.t. The optimization model is rewritten using the value function as, Among them, N ₀ is the noise power spectral density, B is the channel bandwidth; S _t is the decision quantity, which refers to the specific decision in the time slot t, that is, the amount of data to be sent in the time slot; L _t is the state variable, which refers to each The amount of data remaining in time slot t; is an index function, which is a quantitative index to measure a decision-making process, here refers to the minimum energy consumption index; Using mathematical induction, the optimal rate R _i,t * of cloud service i in time slot t is finally obtained as, The minimum global energy consumption model of cloud service i is, in, v _i,x represents the use of channel gain The statistical characteristics of the channel state of cloud service i obtained by the fractional moment of the reciprocal of ; is the geometric mean of v _i,x . 6. The method for optimizing and scheduling cloud service energy consumption in a multi-service coexistence system according to any one of claims 2 to 4, wherein the steps F1 to F4 specifically include the following: Based on the preset upper limit value of cloud service energy consumption, the optimization problem of the maximum time slot throughput of the system is rewritten into the corresponding Lagrangian function as: in, Both α _t and β _t are Lagrange multiplier vectors; Its dual function is, The corresponding dual problem is s.t. α _t ,β _t ≥0 To solve the dual problem, first use the greedy algorithm to allocate channels, and then use the bisection method to solve the optimal dual coefficient. The specific steps are as follows, where ε _i is the preset maximum energy limit, The total number of cloud services activated for the first time for time slot t: i. Initialization ii. Order iii. For any service n, traverse any channel k that can be allocated to it, and perform the above operations for all services, When service n is assigned to channel k The incremental value of , set as Δw _n,k , satisfying Find the (n ^* ,k ^* ) that maximizes Δwn _,k , and assign the corresponding k ^* to n ^* ; iv. Repeat step iii until all channels are allocated; v. Under the channel allocation scheme obtained above, calculate value, if corresponds to otherwise calculate value, if then the corresponding otherwise Repeat steps ii-v until for and for Among them, δ is the constant we set to control the accuracy of the algorithm. The smaller the δ, the higher the accuracy of the algorithm. Among them, K indicates that the number of channels in the system is K, and the corresponding channel set is M represents the number of common services in the system, and the corresponding service set is The number I of cloud services in the system, and the corresponding service set is The amount of data to be uploaded by the cloud service _i is _Li , the time limit for uploading data is Ti, and the moment of starting to upload data is t=ΔT _i .