A multi-objective resource scheduling method for cloud computing
By modeling resource scheduling problems into multi-objective optimization problems in cloud computing and adopting distributed estimation algorithms, the problems of service provider cost differences and maximum benefits are solved, and the benefits and efficiency improvement of resource scheduling are maximized.
Patent Information
- Application Number
- CN202111653724.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-12-30
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2041-12-30
AI Technical Summary
The existing cloud computing resource scheduling methods ignore the cost difference and profit maximization problems of service providers, and it is difficult to deal with the mixed types and irregular dimensions of decision variables, resulting in the inability to effectively maximize the profits of both parties.
Model the cloud computing resource scheduling problem as optimization problems with two goals and multiple constraints. A multi-objective optimization algorithm based on distributed estimation is adopted. By encoding resource scheduling scheme and objective function, a distributed estimation method is used to process different types of variables. Take the value and generate the optimal resource scheduling reference scheme.
It maximizes the benefits in the resource scheduling process in cloud computing scenarios, improves search efficiency, provides an allocation solution that is more suitable for users and service providers, and solves the problems of mixed type variables and unfixed dimensions.
Smart Images

Figure CN114330891B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of resource scheduling in cloud computing scenarios, and in particular to a multi-objective resource scheduling method for cloud computing. Background Art
[0002] In recent years, with the continuous development of computer networks and distributed computing technologies, cloud computing technology has been widely used in various fields and has achieved continuous success. Cloud computing technology combines scattered and different computing resources in the network to form a virtual computing resource center, and uniformly allocates and schedules them according to different users' different needs for computing power, bandwidth, storage, etc., thereby solving the limitations of fewer computing resources and longer computing time in traditional computer application scenarios, and thus greatly reducing users' expenses and time costs. The basic problem of cloud computing technology is the cloud resource scheduling problem. Usually, the computing resources in cloud computing technology come from different service providers. Therefore, how to uniformly manage and schedule the computing resources of the computing center according to different computing service needs of users and the computing resources provided by different service providers is an important task and a hot research direction in related fields. In actual scenarios, each service provider has different costs for the different computing resources it provides, and users have certain budget requirements for the computing resources they need. Service providers hope to obtain the maximum benefits, while users hope to pay as little as possible for the computing services they need. Therefore, how to reasonably allocate tasks to different service providers and maximize the benefits of both parties is one of the problems that need to be solved in resource scheduling in cloud computing technology. From this perspective, the problem of maximizing the benefits of resource scheduling in cloud computing technology is a multi-objective problem, and an effective way to solve multi-objective problems is to use a multi-objective evolutionary algorithm.
[0003] Multi-objective evolutionary optimization is a branch of evolutionary computing and swarm intelligence computing. Nowadays, it is often used to solve real multi-objective optimization problems in the real world. In the process of multi-objective evolutionary optimization, decision variables are used to represent the solution of the problem in the real world, and the objective function is used to represent the measurement index of the problem. New solutions are generated through crossover mutation, and solutions are selected through environmental selection, niche technology and other methods. Finally, after a certain number of iterations, the final solution set is obtained.
[0004] Most of the existing research on resource scheduling in cloud computing scenarios only considers implicit issues such as task-related or actual hardware energy consumption, while ignoring the more important benefit issues in the actual process. Most of the existing methods regard the cost of each type of computing resources provided by different service providers as a fixed value and price them the same. Although the scheduling scheme generated by this method is relatively simple and efficient, it ignores the cost differences of service providers and often cannot maximize the total benefit of service providers in actual practice. Moreover, due to the different service costs of each service provider, even if the service providers provide the same time and quality of service, the benefits they obtain vary. Secondly, most of the existing multi-objective optimization methods are only applicable to simple continuous or discrete decision variables, while the benefit maximization problem of resource scheduling in cloud computing scenarios has the characteristics of mixed types and unfixed number of dimensions in its decision variables. There is currently no more complete and systematic solution for this type of problem. In addition, in the process of multi-objective evolutionary optimization, the information of the intermediate population often implies some characteristics of the optimal solution. If this information can be used, it will have a beneficial effect on the entire search process. However, the multi-objective optimization methods currently widely used in cloud computing scenarios do not consider the use of this information in the optimization process. Summary of the invention
[0005] In view of the deficiencies in the prior art, the purpose of the present invention is to provide a multi-objective resource scheduling method for cloud computing, which models the benefit maximization problem of resource scheduling in a cloud computing scenario as an optimization problem with two objectives and multiple constraints for solution.
[0006] In order to achieve the above technical effects, the present invention proposes a multi-objective resource scheduling method for cloud computing, comprising:
[0007] Step 1: Set the encoding method of the resource scheduling scheme according to the computing resources to be allocated;
[0008] Step 2: Convert the computing resource scheduling problem into a mathematical model consisting of two objective optimizations and multiple constraint problems;
[0009] Step 3: Use a multi-objective optimization algorithm based on distributed estimation to solve the mathematical model and obtain a set of optimal resource scheduling reference solutions.
[0010] The step 1 comprises:
[0011] Step 1.1: Number all service providers, resource types, and users of computing resources to be allocated in the current cloud computing center;
[0012] Step 1.2: Count all orders submitted by users to be allocated computing resources;
[0013] Step 1.3: Determine the corresponding quotation information provided by each service provider for all computing resources required by the users to be allocated in the current cloud computing center;
[0014] Step 1.4: Set the encoding method of the resource scheduling plan.
[0015] The step 1.4 comprises:
[0016] Step 1.4.1: Set a tag list I=(I 1 , I 2 ,…,I u ,…,I C ), I u ∈{0, 1}, where I u Indicates whether all the requirements of user u are met. A value of 1 indicates that all requirements are met, and a value of 0 indicates that none of them are met. u = 1, 2, ..., C, where C indicates the total number of users to whom computing resources are to be allocated. When the requirements of a user are not fully met, the order of the user will not be executed.
[0017] Step 1.4.2: Set the service provider tag list X=(X 1 , X 2 , …, X u , …, X C ), where X u A list of service providers that provide all computing resources to user u; represents a list of service providers that provide the sth type of computing resources required by user u, s = 1, 2, 3, ..., L, where L represents the total number of types of all computing resources; E represents the total number of service providers; if service provider p provides the sth type of computing resources to user u, then If not provided, is 0; define vector Medium Element The number of values 1 is It indicates the maximum number of computing resources of the sth category that can be allocated to different service providers by user u;
[0018] Step 1.4.3: Set the quantity list of various computing resources provided by the service provider to all users Y = (Y 1 , Y 2 , …, Y u , …, Y C ), where Y u represents the amount of computing resources provided by the service provider to user u. represents the number of computing resources of the sth category provided by each service provider to user u, represents the number of computing resources of the sth category provided by service provider p to user u;
[0019] Step 1.4.4: Generate the code (I, X, Y) of each scheduling scheme according to steps 1.4.1 to 1.4.3.
[0020] The step 2 comprises:
[0021] Step 2.1: Establish the objective function Minimize F' to minimize the total cost of all users of computing resources to be allocated. U :
[0022]
[0023] Transform into the maximization function Maximize F U :
[0024]
[0025] Where U represents the set of numbers of users to be allocated computing resources, U = {1, 2, 3, ..., C}; v u represents the price paid by user u in the actual transaction process; b u represents the maximum price that user u can pay for the order. User u's order is B u =(d u , g u , b u ), d u represents the number of various computing resources required by user u, g u Indicates the maximum number of shares that each computing resource required by user u can be divided into,
[0026] Step 2.2: Establish the objective function MaximizeF that maximizes the total revenue of all service providers in the computing center P :
[0027]
[0028] Among them, r p represents the total revenue obtained by service provider p in the actual transaction process; P represents the set of service provider numbers, P = {1, 2, 3, ..., E}, E represents the total number of service providers, S represents the set of computing resource types, S = {1, 2, 3, ..., L}, L represents the total number of computing resource types, The lowest price quote for the computing resources of category s that the service provider can accept;
[0029] Step 2.3: Establish the constraints of the objective function:
[0030]
[0031]
[0032] In the formula, m p represents the number of various computing resources that service provider p can provide, It indicates the number of computing resources of type s that can be provided by service provider p.
[0033] The step 3 comprises:
[0034] Step 3.1: Generate N uniform reference vectors in the target space and form a vector matrix W, and find the distance vector W respectively. w The nearest T reference vectors form the adjacency vector matrix B w ; Among them, W w represents the w-th reference vector, w = 1, 2, ..., N;
[0035] Step 3.2: Randomly generate scheduling methods to form an initial population and encode them; the population contains N individuals, each of which represents a scheduling scheme, and for each scheduling scheme, calculate its objective function value in the mathematical model;
[0036] Step 3.3: Create an ideal point z, z = {z1, z2}, where z1 and z2 are the maximum values of all individuals in the current population on the two objective functions;
[0037] Step 3.4: Find the individual closest to each reference vector in the target space in the current population and establish an association;
[0038] Step 3.5: For each reference vector, generate a new solution using a distributed estimation-based method. If the fitness value of the new solution on the vector is greater than the solution currently associated with the vector, replace the solution.
[0039] Step 3.6: According to the updated population, find the maximum value of each target and update the ideal point z;
[0040] Step 3.7: Iterate steps 3.5 to 3.6 until the number of iterations of the loop is reached. The optimization calculation ends and the decision variables in the current population are the optimal solutions to the current problem.
[0041] The 3.5 includes:
[0042] Step 3.5.1: For the reference vector W w , in the adjacency matrix B wFind the T vectors contained in the file, and use the T individuals corresponding to these T vectors to form an archive. w ;
[0043] Step 3.5.2: Use distributed estimation to estimate the archive w Establish a variable value probability model for each dimension of all individual decision variables;
[0044] Step 3.5.3: Use the distributed estimation method according to the archive w The total number of individual dimensions in the dimension probability model is established;
[0045] Step 3.5.4: Sample the dimensional probability model and all variable dimension value probability models and generate new solutions;
[0046] Step 3.5.5: Determine whether the new solution satisfies the constraints;
[0047] Step 3.5.6: Iterate for each vector until all reference vectors have been executed.
[0048] The step 3.5.2 is specifically described as follows:
[0049] When the nth dimension is a discrete variable, according to the archive w The probability model is established based on the frequency of the values of the nth dimension of individuals;
[0050] When the nth dimension is a continuous variable, the archive is clustered using a fast clustering algorithm. w The values of the nth dimension of individuals in the dataset are clustered into different intervals of the same length, and a probability model for the continuous variable values is established based on the frequency of individuals appearing in each interval.
[0051] The step 3.5.5 comprises:
[0052] Step S1: If satisfied, calculate the fitness value of the new solution under the current vector and compare it with the fitness value of the solution corresponding to the current vector. If the fitness of the new solution under the current vector is greater than the fitness value of the solution corresponding to the vector, replace the solution with the new solution.
[0053] Step S2: If not, modify the new solution to make it satisfy the constraint and execute step S1.
[0054] The beneficial effects of the present invention are:
[0055] The present invention proposes a multi-objective resource scheduling method for cloud computing, which can handle a multi-objective evolutionary algorithm in which decision variables are composed of mixed-type variables and the variable dimensions are not fixed. The method can solve the problem of maximizing the benefits in the resource scheduling process under the cloud computing scenario. For different types of variables, two different distributed estimation methods are used to determine the value of the variable, making full use of the characteristics of the population in the evolutionary process, greatly increasing the search efficiency, and providing a set of solutions to decision makers, so that decision makers can obtain an allocation plan that is more suitable for the expectations of current users and service providers. BRIEF DESCRIPTION OF THE DRAWINGS
[0056] Figure 1 It is a flow chart of the multi-objective resource scheduling method for cloud computing in the present invention;
[0057] Figure 2 is a general flow chart of the multi-objective evolutionary algorithm in the present invention;
[0058] Figure 3 The effect diagram of the probability model established by the distributed estimation method for the continuous variable values in the decision variables in the present invention;
[0059] Figure 4 The effect diagram of the probability model established for the decision variable dimensions using the distributed estimation method in the present invention;
[0060] Figure 5 It is a schematic diagram of the optimization process of the scheduling method described in the present invention. DETAILED DESCRIPTION
[0061] The invention will be further described below with reference to the accompanying drawings and specific implementation examples.
[0062] In the process of design and production practice, we are faced with many optimization problems. For example, in the process of enterprise manufacturing, how to reduce the production time of goods, how to reduce the cost of goods, and improve product quality. Usually, there is more than one optimization goal to be dealt with, and multiple goals need to be optimized at the same time, but these optimization goals often conflict with each other. Improving the quality of one goal often harms other goals and cannot achieve the best at the same time. For example: optimizing the two goals of product production efficiency and product quality at the same time, but improving product production efficiency often reduces product quality, while improving product quality often reduces product production efficiency. This type of problem is called a multi-objective optimization problem (MOP).
[0063] Multi-objective evolutionary algorithms (MOEA) are a powerful tool for solving multi-objective optimization problems. In recent years, they have been widely used in resource scheduling, mechanical manufacturing, transportation, aerospace design and other fields and have achieved continuous success.
[0064] The multi-objective evolutionary algorithm is a population-based algorithm. Usually, the flowchart of the algorithm is as follows: Figure 2 As shown. First, generate an initial population P containing N individuals and calculate the target value of each individual, where each individual in the population represents a solution. After that, genetic operations such as crossover mutation are used on the individuals in the current population to generate the offspring population Q. The initial / parent population and the offspring population are merged and N individuals are selected from them through environmental selection (a method of comparing the pros and cons between individuals) to form a dominant population P′. The dominant population is used as the parent population and the above process is repeated until the maximum number of iterations is reached. Finally, the individuals in the dominant population are generated, which is the optimal solution found.
[0065] When designing a new MOEA, it is necessary to consider both the convergence and distribution of the solution. The convergence of the solution guarantees the quality of the solution, while the distribution of the solution can provide decision makers with more reference solutions with greater differences. In this embodiment, a multi-objective resource scheduling method for cloud computing is proposed. Figure 1 As shown, including:
[0066] Step 1: Setting the encoding method of the resource scheduling scheme according to the computing resources to be allocated; including:
[0067] Step 1.1: Number all service providers, resource types and users of computing resources to be allocated in the current cloud computing center; where the number set of service providers P = {1, 2, 3, ..., E}, where E represents the total number of service providers; the number set of computing resource types S = {1, 2, 3, ..., L}, where L represents the total number of computing resource types; the number set of users of computing resources to be allocated U = {1, 2, 3, ..., C}, where C represents the total number of users of computing resources to be allocated;
[0068] Step 1.2: Count all orders submitted by users for computing resources to be allocated; the number of various computing resources required by user u Indicates the number of computing resources of the sth type required by user u; the maximum number of shares that each computing resource required by user u can be divided into Indicates that the sth type of computing resources required by user u can be obtained by Different service providers provide services; user u's order B u =(d u , g u , bu ), b u represents the maximum price that user u can pay for the order;
[0069] Step 1.3: Determine the corresponding quotation information provided by each service provider for all the computing resources required by the users to be allocated in the current cloud computing center; the number of computing resources that service provider p can provide represents the quantity that service provider p can provide for the sth type of resource; the lowest price acceptable to service provider p for each type of computing resource Quotation O from service provider p p =(s p , a p );
[0070] Step 1.4: Set the encoding method of the resource scheduling scheme; including:
[0071] Step 1.4.1: Set a tag list I=(I 1 , I 2 ,…,I u ,…,I C ), I u ∈{0, 1}, where I u Indicates whether all the requirements of user u are met. A value of 1 indicates that all requirements are met, and a value of 0 indicates that none of them are met. u = 1, 2, ..., C, where C indicates the total number of users to whom computing resources are to be allocated. When the requirements of a user are not fully met, the order of the user will not be executed.
[0072] Step 1.4.2: Set the service provider tag list X=(X 1 , X 2 , …, X u , …, X C ), where X u A list of service providers that provide all computing resources to user u; represents a list of service providers that provide the sth type of computing resources required by user u, s = 1, 2, 3, ..., L, where L represents the total number of types of all computing resources; E represents the total number of service providers; if service provider p provides the sth type of computing resources to user u, then If not provided, is 0; define vector Medium Element The number of values 1 is It indicates the maximum number of computing resources of the sth category that can be allocated to different service providers by user u;
[0073] Step 1.4.3: Set the quantity list of various computing resources provided by the service provider to all users Y = (Y 1 , Y 2 , …, Y u , …, Y C ), where Y u represents the amount of computing resources provided by the service provider to user u. represents the number of computing resources of the sth category provided by each service provider to user u, represents the number of computing resources of the sth category provided by service provider p to user u;
[0074] To reduce redundancy, if If the value is 0, is empty, where the vector The number of non-empty elements in is
[0075] Step 1.4.4: Generate the code (I, X, Y) of each scheduling scheme according to steps 1.4.1 to 1.4.3;
[0076] This implementation case assumes that there are currently two users of computing resources to be allocated in the cloud computing center, and three service providers, which provide three different types of computing resources. Among them, the number set of service providers is P = {1, 2, 3}, the number set of computing resource types is S = {1, 2, 3}, and the number set of users of computing resources to be allocated is U = {1, 2}.
[0077] Count all the orders submitted by users to be allocated computing resources. Each order submitted by a user consists of three parts. The first part is the number of all computing resources required by user u, denoted by d u It indicates that, The second part is the maximum number of shares that each computing resource required by user u can be divided into, denoted by g u It indicates that, Indicates that the sth type of computing resources required by user u can be obtained by The third part is the maximum price that user u can pay for the order, denoted by b u From the above, we can see that the order of user u can be represented by symbol B u Indicates that B u =(d u , g u , b u ). Therefore, all users' orders can be represented by matrix B:
[0078]
[0079] Each service provider provides a corresponding quotation for all the computing resources required by the users to be allocated in the current cloud computing center. The quotation mainly consists of two parts. The first part is the amount of computing resources that service provider p can provide: this part is represented by symbol m p Indicates that represents the quantity of resources provided by service provider p for the sth type. The second part is the lowest price acceptable to service provider p for each type of computing resource, denoted by a p It indicates that, From the above, we can see that the quotation of service provider p can be represented by symbol O p Indicates that O p =(m p , a p ). Therefore, the quotations of all service providers can be represented by the matrix O:
[0080]
[0081] Set the encoding method of the resource scheduling scheme. Each scheduling scheme is encoded as: (I, X, Y), where I = (I 1 , I 2 ),I u is a flag indicating whether all the requirements of user u are met, 1 indicates all requirements are met, and 0 indicates not all requirements are met. X lists the service provider numbers that provide computing resources to all users, where X = (X 1 , X 2 ),X u A list of service providers that provide all computing resources to user u; A list of service providers that provide the s-th type of computing resources required by user u; If service provider p provides user u with computing resources of type s, then If not provided, is 0; define vector Medium Element The number of values 1 is represents the maximum number of computing resources of the sth type that user u is allowed to allocate to different service providers. Y lists the number of computing resources of each type provided by the service provider to all users, where Y = (Y 1 , Y 2 ), Y u represents the amount of computing resources provided by the service provider to user u. represents the number of computing resources of the sth category provided by each service provider to user u, It represents the quantity of computing resources of the sth category provided by service provider p to user u.
[0082] To reduce redundancy, if If the value is 0, is empty, where the vector The number of non-empty elements in is
[0083] Step 2: Convert the computing resource scheduling problem into a mathematical model consisting of two objective optimizations and multiple constraint problems. The objectives are to minimize the total expenditure of all users to be allocated in the cloud computing center and maximize the total revenue of all service providers. This includes:
[0084] Step 2.1: The total expenditure of the task to be minimized refers to the sum of the expenditures of all users of the computing resources to be allocated for their required computing resources. The objective function Minimize F' is established to minimize the sum of the expenditures of all users of the computing resources to be allocated for their required computing resources. U :
[0085]
[0086] Transform into the maximization function Maximize F U :
[0087]
[0088] Where U represents the set of numbers of users to be allocated computing resources, U = {1, 2, 3, ..., C}; v u represents the price paid by user u in the actual transaction process; b u represents the maximum price that user u can pay for the order. User u's order is B u =(d u , g u , b u ), d u represents the number of various computing resources required by user u, g u Indicates the maximum number of shares that each computing resource required by user u can be divided into, Indicates that the sth type of computing resources required by user u can be obtained by Provide services to different service providers;
[0089] Step 2.2: The total revenue of the maximization task refers to the sum of the revenues of all service providers in the computing center. The objective function Maximize F is established to maximize the sum of the revenues of all service providers in the computing center. P :
[0090]
[0091] Among them, r p represents the total revenue obtained by service provider p in the actual transaction process; P represents the set of service provider numbers, P = {1, 2, 3, ..., E}, E represents the total number of service providers, S represents the set of computing resource types, S = {1, 2, 3, ..., L}, L represents the total number of computing resource types, The lowest price quote for the computing resources of category s that the service provider can accept;
[0092] Step 2.3: The constraints refer to the variables in the above maximization formula needing to satisfy the continuous or discrete conditions, and establish the constraints of the objective function:
[0093]
[0094]
[0095] In the formula, m p represents the number of various computing resources that service provider p can provide, represents the number of computing resources of category s that can be provided by service provider p;
[0096] Step 3: Solve the mathematical model using a multi-objective optimization algorithm based on distributed estimation to obtain a set of optimal resource scheduling reference solutions; including:
[0097] Step 3.1: Generate N = 100 uniform reference vectors in the target space and form a vector matrix W, and find the distance vector W w The nearest T = 20 reference vectors form the adjacency vector matrix B w ; Among them, W w represents the w-th reference vector, w = 1, 2, ..., 100;
[0098] Step 3.2: Randomly generate scheduling methods to form an initial population and encode them; the population contains N = 100 individuals, each individual represents a scheduling scheme, and for each scheduling scheme, calculate its objective function value in the mathematical model;
[0099] Step 3.3: Create an ideal point z, z = {z1, z2}, where z1 and z2 are the maximum values of all individuals in the current population on the two objective functions;
[0100] Step 3.4: Find the individual closest to each reference vector in the target space in the current population and establish an association;
[0101] Step 3.5: For each reference vector, generate a new solution using a distributed estimation-based method. If the fitness value of the new solution on the vector is greater than the solution currently associated with the vector, replace the solution; including:
[0102] Step 3.5.1: For the reference vector W w , in the adjacency matrix B w Find the T = 20 vectors contained in the archive, and use the T = 20 individuals corresponding to these T = 20 vectors to form an archive w ;
[0103] Step 3.5.2: Use distributed estimation to estimate the archive w A variable value probability model is established for each dimension of all individual decision variables in; the specific expression is:
[0104] When the nth dimension is a discrete variable, according to the archive w The probability model is established based on the frequency of the values of the nth dimension of individuals;
[0105] When the nth dimension is a continuous variable, the density-based clustering algorithm DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is used to cluster the archive w The values of the nth dimension of individuals in the cluster are clustered into different intervals of the same length, and then a probability model of continuous variable values is established based on the frequency of individuals in each interval. The established probability model is as follows: Figure 3 shown.
[0106] Step 3.5.3: Use the distributed estimation method according to the archive w The total number of individual dimensions in the dimensional probability model is established; the established dimensional probability model is as follows Figure 4 .
[0107] Step 3.5.4: Sample the dimensional probability model and all variable dimensional value probability models and generate new solutions; for the dimensional probability model and discrete variable dimensional value probability model, directly use random sampling to take values. For the continuous variable dimensional value probability model, first randomly sample the interval, and then use uniform sampling to take values in the interval to generate new solutions;
[0108] Step 3.5.5: Determine whether the new solution satisfies the constraints; including:
[0109] Step S1: If satisfied, calculate the fitness value of the new solution under the current vector and compare it with the fitness value of the solution corresponding to the current vector. If the fitness of the new solution under the current vector is greater than the fitness value of the solution corresponding to the vector, replace the solution with the new solution.
[0110] Step S2: If not, modify the new solution to make it satisfy the constraint and execute step S1;
[0111] Step 3.5.6: Iterate for each vector until all reference vectors have been executed;
[0112] Step 3.6: According to the updated population, find the maximum value of each target and update the ideal point z;
[0113] Step 3.7: Iterate steps 3.5 to 3.6 until the number of iterations of the loop is reached. The optimization calculation ends and the decision variables in the current population are the optimal solutions to the current problem.
[0114] Figure 5 The optimization process of this scheduling method is shown. The dotted line is the uniform vector generated in step 3.1, and point P 1 …P 6 is the initial population individual generated in step 3.2, and point Z is the ideal point generated in step 6.3. Each individual in the population is associated with the uniform vector generated in step 3.1. For each vector in the target space (in terms of vector W 3 Take this vector as an example), find the 4 vectors closest to the vector (W 1 , W 2 , W 4 , W 5 ), the individuals associated with these vectors constitute the archive associated with the vector 3 (Step 3.5.1). Then, the archive is estimated using the distributed estimation method. 3 For each individual in the decision variable value probability model and dimensional probability model (step 3.5.2 to step 3.5.3). Sampling is performed on the established model to generate a new individual, which is represented by point P in the figure. 7 Indicates. The individual is compared with the old individual corresponding to the previous vector using the fitness evaluation function. If the fitness of the new individual in the current vector is greater than the fitness value of the old individual corresponding to the vector, the new individual replaces the old individual. The above operation is repeated many times until the maximum number of cycles or the maximum number of evaluations is reached. The individuals in the current population are the optimal solution to the cloud computing resource scheduling problem.
Claims
1. A multi-objective resource scheduling method for cloud computing, characterized in that: include: Step 1: Set the encoding method of the resource scheduling scheme according to the computing resources to be allocated; Step 2: Convert the computing resource scheduling problem into a mathematical model consisting of target optimization and constraint problems; Step 3: Use a multi-objective optimization algorithm based on distributed estimation to solve the mathematical model and obtain a set of optimal resource scheduling reference solutions. The step 1 comprises: Step 1.1: Number all service providers, resource types, and users of computing resources to be allocated in the current cloud computing center; Step 1.2: Count all orders submitted by users to be allocated computing resources; Step 1.3: Determine the corresponding quotation information provided by each service provider for all computing resources required by the users to be allocated in the current cloud computing center; Step 1.4: Set the encoding method of the resource scheduling plan; The step 2 comprises: Step 2.1: Establish the objective function Minimize F' to minimize the total cost of all users of computing resources to be allocated. U : Transform into the maximization function Maximize F U : Where U represents the set of numbers of users to be allocated computing resources, U = {1, 2, 3, ..., C}; v u represents the price paid by user u in the actual transaction process; b u represents the maximum price that user u can pay for the order. The order of user u is Bu=(d u ,q u , b u ), d u represents the number of various computing resources required by user u, g u Indicates the maximum number of shares that each computing resource required by user u can be divided into, Step 2.2: Establish the objective function Maximize F to maximize the total revenue of all service providers in the computing center P : Among them, r p represents the total revenue obtained by service provider p in the actual transaction process; P represents the set of service provider numbers, P = {1, 2, 3, ..., E}, E represents the total number of service providers, S represents the set of computing resource types, S = {1, 2, 3, ..., L}, L represents the total number of computing resource types, The lowest price quote for the s-th type of computing resources acceptable to the service provider; Step 2.3: Establish the constraints of the objective function: In the formula, m p represents the number of various computing resources that service provider p can provide, represents the number of computing resources of category s that can be provided by service provider p; The step 3 comprises: Step 3.1: Generate N uniform reference vectors in the target space and form a vector matrix W, and find the distance vector W respectively. w The nearest T reference vectors form the adjacency vector matrix B w , where W w represents the w-th reference vector, w = 1, 2, ..., N; Step 3.2: Randomly generate scheduling methods to form an initial population and encode them; the population contains N individuals, each of which represents a scheduling scheme, and for each scheduling scheme, calculate its objective function value in the mathematical model; Step 3.3: Create an ideal point z, z = {z1, z2}, where z1 and z2 are the maximum values of all individuals in the current population on the two objective functions; Step 3.4: Find the individual closest to each reference vector in the target space in the current population and establish an association; Step 3.5: For each reference vector, generate a new solution using a distributed estimation-based method. If the fitness value of the new solution on the vector is greater than the solution currently associated with the vector, replace the solution. Step 3.6: According to the updated population, find the maximum value of each target and update the ideal point z; Step 3.7: Iterate steps 3.5 to 3.6 until the number of iterations of the loop is reached. The optimization calculation ends and the decision variables in the current population are the optimal solutions to the current problem.
2. The multi-objective resource scheduling method for cloud computing according to claim 1, characterized in that: The step 1.4 comprises: Step 1.4.1: Set a tag list I = (I1, I2, ..., I u ,…,I C ), I u ∈{0, 1}, where I u Indicates whether all the requirements of user u are met. A value of 1 indicates that all requirements are met, and a value of 0 indicates that none of them are met. u = 1, 2, ..., C, where C indicates the total number of users to whom computing resources are to be allocated. When the requirements of a user are not fully met, the order of the user will not be executed. Step 1.4.2: Set a service provider tag list X = (X1, X2, ..., X u , …, X C ), where X u A list of service providers that provide all computing resources to user u; represents a list of service providers that provide the sth type of computing resources required by user u, s = 1, 2, 3, ..., L, where L represents the total number of types of all computing resources; E represents the total number of service providers; if service provider p provides the sth type of computing resources to user u, then If not provided, is 0; define vector Medium Element The number of values 1 is It indicates the maximum number of computing resources of the sth category that can be allocated to different service providers by user u; Step 1.4.3: Set the quantity list Y of various computing resources provided by the service provider to all users = (Y1, Y2, ..., Y u , …, Y C ), where Y u represents the amount of computing resources provided by the service provider to user u. represents the number of computing resources of the sth category provided by each service provider to user u, represents the number of computing resources of the sth category provided by service provider p to user u; Step 1.4.4: Generate the code (I, X, Y) of each scheduling scheme according to steps 1.4.1 to 1.4.
3.
3. The multi-objective resource scheduling method for cloud computing according to claim 1, characterized in that: The 3.5 includes: Step 3.5.1: For the reference vector W w , in the adjacency matrix B w Find the T vectors contained in the file, and use the T individuals corresponding to these T vectors to form an archive. w ; Step 3.5.2: Use distributed estimation to estimate the archive w Establish a variable value probability model for each dimension of all individual decision variables; Step 3.5.3: Use the distributed estimation method according to the archive w The total number of individual dimensions in the dimension probability model is established; Step 3.5.4: Sample the dimensional probability model and all variable dimension value probability models and generate new solutions; Step 3.5.5: Determine whether the new solution satisfies the constraints; Step 3.5.6: Iterate for each vector until all reference vectors have been executed.
4. The multi-objective resource scheduling method for cloud computing according to claim 3, characterized in that: The step 3.5.2 is specifically described as follows: When the nth dimension is a discrete variable, according to the archive w The probability model is established based on the frequency of the values of the nth dimension of individuals; When the nth dimension is a continuous variable, the archive is clustered using a fast clustering algorithm. w The values of the nth dimension of individuals in the dataset are clustered into different intervals of the same length, and a probability model for the continuous variable values is established based on the frequency of individuals appearing in each interval.
5. The multi-objective resource scheduling method for cloud computing according to claim 3, characterized in that: The step 3.5.5 comprises: Step S1: If satisfied, calculate the fitness value of the new solution under the current vector and compare it with the fitness value of the solution corresponding to the current vector. If the fitness of the new solution under the current vector is greater than the fitness value of the solution corresponding to the vector, replace the solution with the new solution. Step S2: If not, modify the new solution to make it satisfy the constraint and execute step S1.
