Heterogeneous cloud multi-target task scheduling method and system based on quantum collaborative optimization

Abstract

The invention provides a heterogeneous cloud multi-target task scheduling method and system based on quantum collaborative optimization, and the method comprises the steps: generating a DAG graph based on a user request, and constructing a chromosome code of charging mechanism collaboration; an initial population is randomly generated, quantum fitness evaluation is performed, non-dominated sorting is accelerated by applying a Grover search algorithm, a quantum-assisted crowding distance is introduced, and a to-be-evolved individual is obtained. Performing crossover and mutation operation on the sample by using a quantum genetic algorithm, and stopping when a convergence condition is reached; and finally, converting a Pareto solution set obtained by the quantum genetic algorithm into pheromones, initializing by using an ant colony algorithm and carrying out routing updating to complete scheduling and distribution. According to the invention, global search is carried out through the quantum genetic algorithm, and local optimization is carried out through the ant colony algorithm, so that the overall efficiency of the task scheduling method is improved; meanwhile, by improving a genetic algorithm chromosome coding mode and redefining a virtual machine cost function, the problem of billing mechanism difference in heterogeneous clouds is solved.

Description

Technical Field

[0001] The present invention relates to the field of cloud computing technology, and in particular to a heterogeneous cloud multi-objective task scheduling method and system based on quantum collaborative optimization. Background Art

[0002] Hybrid cloud integrates public cloud and private cloud through secure channels to achieve dynamic business allocation. At a time when the scale of cloud service users is increasing dramatically, ensuring service quality (QoS) has become the key, and task scheduling, as a core technology of cloud computing, needs to optimize the matching of tasks and resources within a limited time.

[0003] However, heterogeneous cloud architectures bring many challenges. In terms of billing mechanism, the fixed cost amortization of private clouds is different from the dynamic pricing strategy of public clouds. Cross-cloud scheduling needs to balance the two mutually constrained goals of resource utilization and economic cost. In terms of multi-objective optimization, not only the deadline must be met, but also multi-dimensional indicators such as virtual machine cost, energy consumption and load balancing must be taken into account. The existing scheduling algorithms have room for improvement in solution space search efficiency and Pareto frontier approximation accuracy.

[0004] At the same time, large-scale workflow scheduling faces severe computing challenges. As the scale of tasks increases, there are strict orders between tasks and cross-data center transmission requirements, coupled with multi-dimensional resource constraints such as virtual machine performance differences, network bandwidth fluctuations, and hybrid cloud billing model differences, the complexity of the problem has increased dramatically. Traditional optimization algorithms have low search efficiency and are prone to local optimality, which restricts the improvement of cloud computing performance.

[0005] In the field of multi-objective task scheduling in cloud computing, although traditional genetic algorithms are suitable for multi-objective high-dimensional scheduling optimization and have strong scalability, they have problems such as inefficient high-dimensional solution space when dealing with large-scale computing tasks. Ant colony algorithms have strong processing capabilities for complex problems, but they have problems such as lack of pheromones in the early stage and poor global retrieval capabilities. In the research of hybrid algorithms, some studies have solved the traveling salesman problem based on the quantum ant colony algorithm, but the problem of lack of pheromones in the ant colony algorithm itself has not been solved, and this method cannot adapt to dynamic task environments. Some studies have proposed a task scheduling strategy based on particle swarm and ant colony algorithms, but it cannot cope with heterogeneous cloud scenarios for large-scale computing. In addition, some studies have proposed a DCOH method, which aims to use genetic algorithms to solve task scheduling problems in hybrid cloud scenarios, but it does not take into account the complexity of task dependencies in hybrid cloud computing environments, so it cannot solve the difficulties in task scheduling in large-scale hybrid cloud computing scenarios. Some scholars have also proposed a dual-objective optimization task scheduling method, but it does not take into account the differences in billing in heterogeneous cloud environments. Therefore, existing methods find it difficult to balance global optimization and local development when dealing with task heterogeneity and resource dynamics. It is urgent to develop new multi-objective optimization scheduling algorithms to meet the challenges of heterogeneous cloud environments. Summary of the invention

[0006] In order to solve the above problems, the present invention proposes a heterogeneous cloud multi-objective task scheduling method and system based on quantum collaborative optimization, which improves the overall efficiency of the task scheduling method by performing global search through a quantum genetic algorithm and local optimization through an ant colony algorithm; at the same time, the problem of differences in billing mechanisms in heterogeneous clouds is solved by improving the genetic algorithm chromosome encoding method and redefining the virtual machine cost function.

[0007] In order to achieve the above object, the present invention adopts the following technical solution:

[0008] In a first aspect, the present invention provides a heterogeneous cloud multi-objective task scheduling method based on quantum collaborative optimization, comprising:

[0009] Generate a DAG graph based on user requests;

[0010] Constructing an objective function according to the DAG graph, and setting a chromosome code for coordination of a billing mechanism based on the objective function; the chromosome code includes task priority, virtual machine allocation, virtual machine type, and quantum rotation angle;

[0011] Randomly generate an initial population and evaluate the quantum fitness of the chromosomes in the population;

[0012] Individuals exceeding the fitness threshold are selected, and the Grover search algorithm is used to accelerate the non-dominated sorting to obtain the non-dominated solution set and perform level division; the quantum-assisted crowding distance is calculated, and the quantum rotation angles of individuals of different levels are adjusted according to the quantum-assisted crowding distance to obtain the individuals to be evolved;

[0013] The quantum genetic algorithm is used to perform quantum crossover and mutation operations on the evolving individuals. When the average fitness value defined based on the hypervolume meets the convergence condition, the quantum genetic algorithm is stopped.

[0014] The Pareto solution set obtained by the quantum genetic algorithm is converted into pheromones, and the ant colony algorithm is used to initialize the pheromones. Then the ants find the path according to the preset probability and update the pheromones to complete task scheduling and resource allocation.

[0015] In a second aspect, the present invention provides a heterogeneous cloud multi-objective task scheduling system based on quantum collaborative optimization, comprising:

[0016] A task parsing module is configured to generate a DAG graph based on user requests;

[0017] A chromosome encoding module is configured to construct an objective function according to the DAG graph, and to set a chromosome encoding for coordination of a billing mechanism based on the objective function; the chromosome encoding includes task priority, virtual machine allocation, virtual machine type, and quantum rotation angle;

[0018] The fitness evaluation module is configured to randomly generate an initial population and perform quantum fitness evaluation on the chromosomes in the population;

[0019] The non-dominated solution processing module is configured to select individuals that exceed the fitness threshold, use the Grover search algorithm to accelerate the non-dominated sorting, obtain the non-dominated solution set and perform level division; calculate the quantum-assisted crowding distance, adjust the quantum rotation angles of individuals of different levels according to the quantum-assisted crowding distance, and obtain the individuals to be evolved;

[0020] The quantum genetic evolution module is configured to use a quantum genetic algorithm to perform quantum crossover and mutation operations on the evolving individuals, and stop the quantum genetic algorithm when the average fitness value defined based on the hypervolume meets the convergence condition;

[0021] The ant colony scheduling optimization module is configured to convert the Pareto solution set obtained by the quantum genetic algorithm into pheromones, and initialize the pheromones using the ant colony algorithm. After that, the ants find paths according to the preset probability and update the pheromones to complete task scheduling and resource allocation.

[0022] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in a heterogeneous cloud multi-objective task scheduling method based on quantum collaborative optimization described in the first aspect.

[0023] In a fourth aspect, the present invention provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the heterogeneous cloud multi-objective task scheduling method based on quantum collaborative optimization described in the first aspect are implemented.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] 1. Aiming at the pain points of large differences in billing mechanisms and complex resource types in heterogeneous cloud environments, the present invention proposes a chromosome encoding mechanism, sets three fields of "task priority-virtual machine allocation-resource type" at the chromosome gene layer, realizes the coordinated optimization of billing strategy and resource scheduling, and breaks through the adaptation bottleneck of traditional optimization algorithms in hybrid billing scenarios.

[0026] 2. To address the problems of low computational efficiency and poor convergence in large-scale workflow scheduling in cloud computing environments, the present invention improves the traditional genetic-ant colony algorithm based on quantum computing, introduces the Grover quantum search algorithm to accelerate non-dominated sorting, and can effectively improve the efficiency of non-dominated sorting in classical evolutionary algorithms, achieving performance breakthroughs through parallel computing.

[0027] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their description are used to explain the present invention but do not constitute a limitation of the present invention.

[0029] Figure 1 A main flow chart of a heterogeneous cloud multi-objective task scheduling method based on quantum collaborative optimization provided by an embodiment of the present invention;

[0030] Figure 2 A detailed flow chart of a heterogeneous cloud multi-objective task scheduling method based on quantum collaborative optimization provided by an embodiment of the present invention;

[0031] Figure 3 A schematic diagram of a heterogeneous cloud multi-objective task scheduling method architecture based on quantum collaborative optimization provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0032] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0033] Explanation of terms

[0034] Quantum bit (Qubit): In classical computing, the basic storage unit of information is 0 and 1, called a bit. In quantum information, the basic storage unit of information is a quantum bit. A quantum bit may be in |0> or |1>, or it may be in a linear superposition of the two states. That is, |ψ>=α|0>+β|1>, where α and β represent the probability amplitudes of |0> and |1> respectively, |α| 2 and |β| 2 They represent the probability of the qubit being in |0> and |1>, respectively, satisfying |α| 2 +|β| 2 =1.

[0035] Embodiment 1

[0036] like Figure 1 As shown, this embodiment discloses a heterogeneous cloud multi-objective task scheduling method based on quantum collaborative optimization, comprising the following steps:

[0037] S1: Generate DAG graph based on user request;

[0038] S2: constructing an objective function according to the DAG graph, and setting a chromosome code for coordination of a billing mechanism based on the objective function; the chromosome code includes task priority, virtual machine allocation, virtual machine type, and quantum rotation angle;

[0039] S3: Randomly generate an initial population and evaluate the quantum fitness of the chromosomes in the population;

[0040] S4: Select individuals that exceed the fitness threshold, use the Grover search algorithm to accelerate the non-dominated sorting, obtain the non-dominated solution set and perform level division; calculate the quantum-assisted crowding distance, adjust the quantum rotation angles of individuals of different levels according to the quantum-assisted crowding distance, and obtain the individuals to be evolved;

[0041] S5: Use quantum genetic algorithm to perform quantum crossover and mutation operations on evolving individuals. When the average fitness value defined based on the hypervolume meets the convergence condition, stop the quantum genetic algorithm.

[0042] S6: The Pareto solution set obtained by the quantum genetic algorithm is converted into pheromones, and the ant colony algorithm is used to initialize the pheromones. After that, the ants find the path according to the preset probability and update the pheromones to complete task scheduling and resource allocation.

[0043] Next, combine Figure 2 , a heterogeneous cloud multi-objective task scheduling method based on quantum collaborative optimization disclosed in this embodiment is described in detail.

[0044] In S1, the solution proposed by the present invention mainly includes three layers: user layer, task scheduling layer and data center layer. In heterogeneous clouds, data can be transmitted between private clouds and public clouds, such as Figure 3 As shown in the architecture diagram, in this system, the scheduler decides whether to accept the user's request based on the usage of private cloud and public cloud. When receiving a user request, the user's request is parsed into a workflow, and a DAG graph is generated based on the task dependency relationship, where the DAG graph is represented by G = (V, E), and the node V i represents the task, edge e ij Represents the dependencies between tasks; then, calls quantum genetic algorithm and ant colony algorithm for task scheduling and resource allocation, and passes the tasks to private cloud or public cloud.

[0045] In S2, an objective function is constructed according to the DAG graph. Specifically, the problem of minimizing the completion time and the virtual machine cost can be defined as:

[0046] Minimize P = (Makespan, Cost) T (1)

[0047] Among them, Makespan is the completion time, and Cost is the virtual machine cost. The completion time Makespan represents the maximum completion time of the task node in the DAG graph, which is expressed as:

[0048]

[0049] Among them, ft i The completion time of the task.

[0050] The virtual machine cost includes the public cloud cost. public and private cloud cost private There are two parts. Private cloud is charged on demand, and public cloud is charged by time. So the virtual machine cost can be expressed as:

[0051] Cost = Cost private +Cost public (3)

[0052] The private cloud cost is calculated by multiplying the execution time of each task on its assigned virtual machine by the unit price, while the public cloud is calculated by multiplying the rental time of each virtual machine by the unit price. The private cloud cost and public cloud cost are expressed as formula (4) and formula (5) respectively, where c p represents the cost per unit time of each virtual machine in the private cloud, c h represents the rental cost per unit time of each virtual machine in the public cloud, st i For task v i Execution start time:

[0053]

[0054] Among them, VM i Represents a virtual machine, VM in a private cloud k Respectively represent the virtual machines in the public cloud, Max_ft k Indicates the maximum task end time, Min_st k Indicates the minimum task start time.

[0055] The objective function is defined as F(x) = α*Makespan + β*Cost, where α+β = 1, then the fitness function is:

[0056]

[0057] Furthermore, chromosome encoding for coordination of charging mechanisms is set based on the objective function.

[0058] In the chromosome encoding part, the chromosome encoding method used in this embodiment is:

[0059] Chrome=(order,vm,type,θ) (7)

[0060] Chromosome encoding decision task V iThe cloud selection probability of each task i is expressed by a single quantum bit as shown in formula (8).

[0061] ∣ψ i >=cosθ i ∣0>+sinθ i ∣1> (8)

[0062] Among them, |0> represents the probability p of selecting a private cloud priv (V i )=cos 2 θ i ,|1> represents the probability p of choosing a public cloud pib (V i )=sin 2 θ i .

[0063] This embodiment sets up chromosome encoding that cooperates with the billing mechanism, mainly to cope with different billing modes of heterogeneous clouds, realize the coordination of task scheduling and billing strategies, and improve resource utilization and efficiency. On the one hand, it can determine the priority of tasks, arrange the execution order according to characteristics, and improve overall efficiency; on the other hand, it can accurately match task execution carriers and virtual machine types according to the billing characteristics of different clouds. For example, tasks that are cost-sensitive and have long execution times can use private cloud resources, and tasks that have high timeliness requirements and short execution times can use public cloud resources to reduce overhead. In addition, the quantum rotation angle can also affect the direction of individual evolution in the quantum genetic algorithm, helping the algorithm to search for the optimal solution and optimize task scheduling.

[0064] In S3, N chromosomes are randomly generated, where the quantum part θ is initialized to be uniformly distributed. For all chromosomes in the population, the fitness (x i ).

[0065] It should be understood that the fitness calculation method can be selected by those skilled in the art as needed.

[0066] In S4, individuals exceeding the fitness threshold are selected, and the Grover search algorithm is used to accelerate the non-dominated sorting, and the non-dominated solution set is obtained and graded, including:

[0067] S401: Construct a quantum oracle, encode the non-dominance relationship judgment, compare the chromosome solutions in parallel, and judge the non-dominance relationship between different chromosome solutions; the non-dominance relationship judgment is that if the completion time and virtual machine cost of one chromosome are better than another, then there is a domination relationship;

[0068] S402: amplifying the probability amplitude of the non-dominated solution in the quantum state through the iteration of the Grover search algorithm until the optimal number of iterations is reached;

[0069] S403: By measuring the quantum state statistical chromosome occurrence probability, the chromosome solution is divided into a multi-level non-dominated solution set according to the chromosome occurrence probability and the sorting and stratification rules of the non-dominated solutions.

[0070] In S401, a non-dominant relationship determination method is first defined. If chromosome x i The completion time and virtual machine cost of chromosome x are better than k , then it is judged that there is a dominance relationship, x i Dominate x k , expressed as:

[0071]

[0072] Among them, Makespan i 、Makespan j Chromosome x i and x j Completion time, Cost i Cost j Chromosome x i and x j The virtual machine cost of , there exists a value k such that chromosome x i The objective function value F k (x i ) is less than or equal to the chromosome x j Objective function value F k (x j ).

[0073] After that, a quantum oracle is constructed to mark the non-dominated solutions and encode the non-dominated relation judgment as U dom , so that U dom ∣x i >|x j >|0>=|x i >|x j >|1>, compare the two chromosome solutions in parallel and determine the non-dominance relationship between different chromosome solutions.

[0074] In S402, the Grover search algorithm is applied to iterate until the optimal number of iterations is reached, i.e. When the amplitude of the non-dominated solution is amplified, the probability amplitude of the non-dominated solution is increased.

[0075] In S403, the quantum state is measured, the probability of chromosome appearance is counted, and the non-dominated solution set Front is obtained according to the sorting and stratification of non-dominated solutions in NSGA-II. 1 , the remaining solutions recursively generate Front 2 , Front 3 wait.

[0076] In this embodiment, individuals are screened by setting fitness thresholds, focusing on better solutions, reducing invalid calculations, and improving algorithm efficiency. Secondly, the Grover search algorithm uses quantum parallelism and amplitude amplification characteristics to quickly determine the non-dominated relationship between chromosome solutions, accelerate the non-dominated sorting process, and greatly shorten the calculation time compared to traditional methods. Furthermore, accurate non-dominated sorting and hierarchical division can obtain non-dominated solution sets at different levels, providing a variety of high-quality solution options for subsequent task scheduling, which helps to comprehensively consider multiple objective factors such as completion time and virtual machine cost in heterogeneous cloud environments, achieve more reasonable resource allocation and better task scheduling, and enhance the overall performance of the system.

[0077] Furthermore, the quantum assisted crowding distance is calculated, and the quantum rotation angles of individuals of different levels are adjusted according to the quantum assisted crowding distance to obtain the individuals to be evolved. The calculation of the quantum assisted crowding distance is used to calculate the distribution density of individuals in the multi-level non-dominated solution set in the target space:

[0078]

[0079] Among them, F k (x i+1 ) represents chromosome x i+1 The value of the k-th target space function, F k (x i-1 ) represents chromosome x i-1 The value under the k-th target space function, represents the maximum value of the entire population under k objective functions, It represents the minimum value of the entire population under k objective functions. For individuals with high levels and large crowding distance, the quantum rotation angle is retained; for individuals with low levels, the quantum rotation angle is reset to a random value to increase randomness.

[0080] In S5, a quantum genetic algorithm is used to perform quantum crossover and mutation operations on evolving individuals. When the average fitness value defined based on the hypervolume meets the convergence condition, the quantum genetic algorithm is stopped.

[0081] Specifically, in the crossover and mutation part, this embodiment uses a quantum genetic algorithm to divide Chrome = (order, vm, type, θ) into a classical algorithm operation area (order, vm, type) and a quantum operation area (θ). First, a cut point is randomly selected for the classical algorithm operation area to perform a single-point crossover, and then a quantum exchange gate is applied to θ. In the mutation part, vm or type is randomly flipped, and then the selected θ is i Apply a rotational perturbation:

[0082] θ i ′ =θ i +Δθ Δθ~N(0,0.1π) (11)

[0083] Since the Pareto solution set is not a single solution, the average fitness value is defined based on the hypervolume. After m iterations, the average fitness value can be recorded as The mth generation Pareto frontier hypervolume formula is (12), and the control parameter C (0≤C≤1) is defined as (13). When C approaches 0 for many consecutive times, the quantum genetic algorithm is judged to have converged, and the ant colony algorithm can be started.

[0084]

[0085] Among them, HV (Pareto_Front m ) is an indicator to measure the quality of the Pareto frontier. The larger the volume, the larger the volume occupied by the Pareto frontier in the target space. m represents the mth generation Pareto frontier, represents all non-dominated solutions found in the mth iteration, HV m represents the hypervolume value of the mth generation Pareto frontier, HV m+1 Represents the hypervolume value of the m+1th generation Pareto frontier.

[0086] In the part of quantum genetics generating Pareto solution sets, this embodiment mainly introduces quantum bits and Grover search acceleration methods to improve the classical genetic algorithm so that it has the ability to handle large-scale calculations.

[0087] In S6, the ant colony algorithm is introduced, specifically:

[0088] (1) Pheromone initialization. Initial pheromone reference value τ C Depends on the task length t i_len and compute node capacity vm j_complex ratio.

[0089]

[0090] The larger the ratio, the longer the task length or the poorer the computing power, the longer the execution time may be, and the lower the priority. The pheromone transformed by the Pareto solution set obtained by the quantum genetic algorithm is defined as τ QGA , the pheromone initialization formula is shown in formula (15).

[0091] τ ij (0) = τ QGA +τ C (15)

[0092] (2) Ant path finding. Place m ants on the First node and search for the next node in turn. The probability calculation of ant k (1≤k≤m) moving from city i to city j is shown in formula (16).

[0093]

[0094] Among them, τ ij(t) is the pheromone concentration on path (i, j) at time t, represents the heuristic information function on the path (i, j) at time t, d ij represents the distance between cities i and j; α and β are pheromone factors and heuristic factors. k (k=1,2,...,n) is used to record the paths that the ants have walked. The paths that have been walked will be added to the taboo table, allowed k The node representing the next hop path that the kth ant is allowed to pass through is a path outside the taboo table. According to the formula, it can be seen that ant k mainly chooses nodes with shorter distances and more pheromones on the path as the next node.

[0095] (3) Pheromone update. In order to prevent the infinite accumulation of pheromones on the path, the pheromone volatility coefficient ρ is used to continuously update the pheromones on each path until the optimal number of iterations is reached, and the optimal task scheduling and resource allocation solution is obtained.

[0096] τ ij (t+1)=ρτ ij (t)+Δτ ij (t) (17)

[0097] In Equation 17, Δτ ij (t) is the pheromone increment left by m ants on the path (i, j) in this search:

[0098]

[0099] In terms of cost control, this specific embodiment uses chromosome encoding in coordination with the billing mechanism to select virtual machine types and assign tasks according to the billing characteristics of different cloud environments, which can effectively reduce virtual machine costs and improve resource utilization. In terms of scheduling efficiency, the Grover search algorithm accelerates non-dominated sorting, and the combination of quantum genetic algorithm and ant colony algorithm makes the algorithm converge faster and can find a better scheduling solution in a shorter time. At the same time, quantum parameter perturbation and quantum-assisted crowding distance calculation increase population diversity, avoid the algorithm from falling into local optimality, and increase the probability of searching for the global optimal solution.

[0100] In view of the shortcomings of traditional algorithms in processing large-scale computing tasks, coping with dynamic task environments and heterogeneous cloud scenarios, and the defect that existing hybrid algorithms are difficult to take into account both global and local optimization, the present invention proposes a heterogeneous cloud multi-objective task scheduling method based on quantum collaborative optimization, generates a DAG graph based on user requests and constructs an objective function, sets chromosome encoding for billing mechanism collaboration, and takes into account the differences in billing in heterogeneous cloud environments; secondly, the Grover search algorithm is used to accelerate non-dominated sorting, and the quantum genetic algorithm is combined for crossover and mutation operations, which solves the inefficiency problem of traditional genetic algorithms in high-dimensional solution space; finally, the Pareto solution set obtained by the quantum genetic algorithm is converted into pheromones, which are initialized and updated using the ant colony algorithm, which overcomes the problems of lack of pheromones and poor global retrieval ability in the early stage of the ant colony algorithm, can better handle task heterogeneity and resource dynamics, take into account both global optimization and local development, and effectively respond to the challenges of heterogeneous cloud environments.

[0101] Embodiment 2

[0102] This embodiment provides a heterogeneous cloud multi-objective task scheduling system based on quantum collaborative optimization, including:

[0103] A task parsing module is configured to generate a DAG graph based on user requests;

[0104] A chromosome encoding module is configured to construct an objective function according to the DAG graph, and to set a chromosome encoding for coordination of a billing mechanism based on the objective function; the chromosome encoding includes task priority, virtual machine allocation, virtual machine type, and quantum rotation angle;

[0105] The fitness evaluation module is configured to randomly generate an initial population and perform quantum fitness evaluation on the chromosomes in the population;

[0106] The non-dominated solution processing module is configured to select individuals that exceed the fitness threshold, use the Grover search algorithm to accelerate the non-dominated sorting, obtain the non-dominated solution set and perform level division; calculate the quantum-assisted crowding distance, adjust the quantum rotation angles of individuals of different levels according to the quantum-assisted crowding distance, and obtain the individuals to be evolved;

[0107] The quantum genetic evolution module is configured to use a quantum genetic algorithm to perform quantum crossover and mutation operations on the evolving individuals, and stop the quantum genetic algorithm when the average fitness value defined based on the hypervolume meets the convergence condition;

[0108] The ant colony scheduling optimization module is configured to convert the Pareto solution set obtained by the quantum genetic algorithm into pheromones, and initialize the pheromones using the ant colony algorithm. After that, the ants find paths according to the preset probability and update the pheromones to complete task scheduling and resource allocation.

[0109] Embodiment 3

[0110] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the steps in the heterogeneous cloud multi-objective task scheduling method based on quantum collaborative optimization as described in the first embodiment above are implemented.

[0111] Embodiment 4

[0112] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the heterogeneous cloud multi-objective task scheduling method based on quantum collaborative optimization as described in the first embodiment are implemented.

[0113] The steps or modules involved in the above embodiments 2 to 4 correspond to those in embodiment 1. For the specific implementation, please refer to the relevant description of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0114] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A heterogeneous cloud multi-objective task scheduling method based on quantum collaborative optimization, characterized in that: include: Generate a DAG graph based on user requests; Constructing an objective function according to the DAG graph, and setting a chromosome encoding for coordination of a billing mechanism based on the objective function; The chromosome encoding includes task priority, virtual machine allocation, virtual machine type and quantum rotation angle; Randomly generate an initial population and evaluate the quantum fitness of the chromosomes in the population; Individuals exceeding the fitness threshold are selected, and the Grover search algorithm is used to accelerate the non-dominated sorting to obtain the non-dominated solution set and perform level division; the quantum-assisted crowding distance is calculated, and the quantum rotation angles of individuals of different levels are adjusted according to the quantum-assisted crowding distance to obtain the individuals to be evolved; The quantum genetic algorithm is used to perform quantum crossover and mutation operations on the evolving individuals. When the average fitness value defined based on the hypervolume meets the convergence condition, the quantum genetic algorithm is stopped. The Pareto solution set obtained by the quantum genetic algorithm is converted into pheromones, and the ant colony algorithm is used to initialize the pheromones. Then the ants find the path according to the preset probability and update the pheromones to complete task scheduling and resource allocation.

2. A heterogeneous cloud multi-objective task scheduling method based on quantum collaborative optimization as claimed in claim 1, characterized in that: The generating of the DAG graph based on the user request specifically includes: parsing the user request into a workflow, and generating the DAG graph according to the task dependency relationship; wherein the node represents the task, and the edge represents the dependency relationship between the tasks.

3. A heterogeneous cloud multi-objective task scheduling method based on quantum collaborative optimization as claimed in claim 1, characterized in that: The constructing of the objective function according to the DAG graph is specifically: constructing the objective function based on minimizing the completion time and the virtual machine cost, wherein the completion time represents the maximum completion time of the task node in the DAG graph; The virtual machine cost includes private cloud cost and public cloud cost; the private cloud cost is charged by multiplying the execution time of each task on its assigned virtual machine by the unit price, and the public cloud cost is charged by multiplying the rental time of each virtual machine by the unit price.

4. A heterogeneous cloud multi-objective task scheduling method based on quantum collaborative optimization as claimed in claim 1, characterized in that: The chromosome encoding coordinated by the billing mechanism is used to determine the cloud selection probability of the task, which is specifically expressed as: Chrome=(order,vm,type,θ) Among them, order represents the task priority order; vm represents the virtual machine corresponding to the task, and the virtual machine is the execution carrier of the associated task in different cloud environments; type represents the virtual machine type, which is selected based on the billing characteristics of different cloud environments; θ represents the quantum rotation angle, which affects the evolutionary direction of the individual in the quantum genetic algorithm.

5. The method for scheduling multi-objective tasks in heterogeneous clouds based on quantum collaborative optimization according to claim 1, characterized in that: The selecting of individuals exceeding the fitness threshold, using the Grover search algorithm to accelerate the non-dominated sorting, obtaining the non-dominated solution set and performing level classification, specifically includes: Construct a quantum oracle to encode the non-dominance relationship judgment, compare the chromosome solutions in parallel, and judge the non-dominance relationship between different chromosome solutions; the non-dominance relationship judgment is that if the completion time and virtual machine cost of one chromosome are better than another, then there is a domination relationship; Through the iteration of the Grover search algorithm, the probability amplitude of the non-dominated solution in the quantum state is amplified until the optimal number of iterations is reached; By measuring the quantum state statistical chromosome occurrence probability, the chromosome solution is divided into a multi-level non-dominated solution set according to the chromosome occurrence probability and the sorting and stratification rules of the non-dominated solutions.

6. A heterogeneous cloud multi-objective task scheduling method based on quantum collaborative optimization as claimed in claim 1, characterized in that: The calculation of quantum-assisted crowding distance is used to calculate the distribution density of individuals in the multi-level non-dominated solution set in the target space, specifically including: Among them, F k (x i+1 ) represents chromosome x i+1 The value of the k-th target space function, F k (x i-1 ) represents chromosome x i-1 The value under the k-th target space function, represents the maximum value of the entire population under k objective functions, Represents the minimum value of the entire population under k objective functions.

7. The method for scheduling multi-objective tasks in heterogeneous clouds based on quantum collaborative optimization according to claim 1, characterized in that: The Pareto solution set obtained by the quantum genetic algorithm is converted into pheromone, and the pheromone is initialized by using the ant colony algorithm. Then, the ants find the path according to the preset probability and update the pheromone to complete the task scheduling and resource allocation; specifically, it includes: Determine a pheromone benchmark value according to the ratio of each task length to the corresponding computing node capacity in the Pareto solution set; initialize the pheromone in the ant colony algorithm based on the benchmark value; Place multiple ants on the first node, and each ant finds a path according to a preset probability; The pheromone volatility coefficient is used to iteratively update the pheromone on each path until the optimal number of iterations is reached, and the optimal task scheduling and resource allocation solution is obtained.

8. A heterogeneous cloud multi-objective task scheduling system based on quantum collaborative optimization, characterized in that: include: A task parsing module is configured to generate a DAG graph based on user requests; A chromosome encoding module is configured to construct an objective function according to the DAG graph, and to set a chromosome encoding for coordination of a billing mechanism based on the objective function; the chromosome encoding includes task priority, virtual machine allocation, virtual machine type, and quantum rotation angle; The fitness evaluation module is configured to randomly generate an initial population and perform quantum fitness evaluation on the chromosomes in the population; The non-dominated solution processing module is configured to select individuals that exceed the fitness threshold, use the Grover search algorithm to accelerate the non-dominated sorting, obtain the non-dominated solution set and perform level division; calculate the quantum-assisted crowding distance, adjust the quantum rotation angles of individuals of different levels according to the quantum-assisted crowding distance, and obtain the individuals to be evolved; The quantum genetic evolution module is configured to use a quantum genetic algorithm to perform quantum crossover and mutation operations on the evolving individuals, and stop the quantum genetic algorithm when the average fitness value defined based on the hypervolume meets the convergence condition; The ant colony scheduling optimization module is configured to convert the Pareto solution set obtained by the quantum genetic algorithm into pheromones, and initialize the pheromones using the ant colony algorithm. After that, the ants find paths according to the preset probability and update the pheromones to complete task scheduling and resource allocation.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by the processor, the steps in the heterogeneous cloud multi-objective task scheduling method based on quantum collaborative optimization are implemented as described in any one of claims 1 to 7.

10. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the heterogeneous cloud multi-objective task scheduling method based on quantum collaborative optimization as described in any one of claims 1-7 are implemented.

Images