Multi-element computing power heterogeneous computing power service management system and method

Abstract

The invention discloses a multivariate computing power heterogeneous computing power service management system, which comprises a user terminal, a task scheduling module, a resource management module and a mixed computing power calculation module, the invention also discloses a multi-element computing power heterogeneous computing power service management method which is realized based on the multi-element computing power heterogeneous computing power service management system. The invention discloses a multivariate computing power heterogeneous computing power service management system and a multivariate computing power heterogeneous computing power service management method, which are characterized in that different computing resources are effectively integrated, and an optimal allocation scheme of the computing resources is obtained by utilizing machine learning according to a resource allocation request, so that computing tasks and the computing resources can be reasonably allocated; the problem that the utilization rate of computing resources is low due to single-type computing resource management and multi-core computing is effectively solved, optimal utilization of the computing resources is achieved, the utilization rate of the computing resources is increased, and energy consumption and cost are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of quantum computing and quantum communication, and in particular to a computing power service management system and method for multi-component heterogeneous computing power. Background Art

[0002] Quantum physics and computer science are constantly developing. Quantum computing (QC) is gradually emerging as a new computing model, which follows the laws of quantum mechanics to control quantum information units for computing. Unlike classical computing, quantum computing follows the laws of quantum mechanics. It is a new computing model that can break through the bottleneck of classical computing power. It has more powerful parallel processing capabilities and higher computing efficiency, and can solve many problems that are difficult to solve with classical computing. However, quantum computing technology is still in its infancy, and the development of hardware facilities and algorithms still faces many challenges.

[0003] Computing services cover cloud computing, edge computing, the Internet of Things and other fields. However, traditional computing services are mainly based on classical computing. Although they can meet the needs to a certain extent, they have problems such as slow computing speed, high energy consumption, and large storage requirements when dealing with complex problems. Using only quantum computing also has challenges such as complex algorithm design and poor stability of quantum bits, making it difficult to be widely used in actual scenarios.

[0004] Most existing computing power service management systems manage a single type of computing resources. They are unable to effectively integrate quantum computing resources and classical computing resources, nor can they reasonably schedule tasks, resulting in low utilization of computing resources and reduced system performance. Summary of the invention

[0005] In order to solve the problem that most existing computing power service management systems manage a single type of computing resource and schedule multi-core computing, resulting in low utilization of computing resources, the present invention provides a computing power service management system and method with multiple computing powers and heterogeneity.

[0006] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0007] A multi-computing power heterogeneous computing power service management system, including a user terminal, a task scheduling module, a resource management module and a hybrid computing power calculation module;

[0008] The user terminal is used to submit computing tasks and processing requests to the task scheduling module; the processing request includes task type, priority and computing resource requirements;

[0009] The task scheduling module is used to classify computing tasks according to task types, adjust the execution order of computing tasks according to priorities, and initiate resource allocation requests to the resource management module according to computing resource requirements; the resource allocation request includes task classification results, execution order and computing resource requirements;

[0010] The resource management module is used to obtain an optimal allocation plan for computing resources using machine learning according to a resource allocation request;

[0011] The hybrid computing power calculation module is used to perform computing tasks according to the optimal allocation plan of computing resources and return the obtained computing results to the user terminal.

[0012] In the above scheme, by effectively integrating different computing resources and using machine learning to obtain the optimal allocation plan for computing resources based on resource allocation requests, computing tasks and computing resources can be reasonably allocated, and the optimal use of computing resources can be achieved, thereby improving the utilization rate of computing resources and reducing energy consumption and costs.

[0013] Preferably, the task types include quantum computing tasks and classical computing tasks.

[0014] Preferably, the computing resource requirements include quantum computing resource requirements and classical computing resource requirements.

[0015] Preferably, the task scheduling module includes a task distinguisher and a task queuer;

[0016] The task distinguisher is used to classify computing tasks according to task types, adjust the execution order of computing tasks according to priorities, and initiate resource allocation requests to the resource management module according to computing resource requirements;

[0017] The task queuer is used to adjust the execution order of computing tasks according to the waiting time and priority of the computing tasks.

[0018] Preferably, the resource management module includes a resource dynamic scheduler, a load balancer and a virtual resource pool;

[0019] The resource dynamic scheduler is used to obtain the optimal allocation plan of computing resources by machine learning according to the resource allocation request, and send the optimal allocation plan to the virtual resource pool and send the computing task to the hybrid computing power computing module;

[0020] The load balancer is used to monitor the load of the virtual resource pool (such as resource utilization, resource remaining capacity and other data), and feed back the load to the resource dynamic scheduler;

[0021] The virtual resource pool is used to mark the location of the computing resources to be used according to the optimal allocation plan, and to feedback the load situation to the load balancer, and to forward the optimal allocation plan to the hybrid computing power calculation module.

[0022] Preferably, the virtual resource pool includes a quantum computing resource pool and a classical computing resource pool.

[0023] In the above solution, virtualization technology is used to perform virtualization management on physical computing resources, so that different application programs can share the same set of physical resources, thereby improving resource utilization.

[0024] Preferably, the optimal allocation scheme of computing resources is obtained by using machine learning through the following steps:

[0025] A1: Initialize the strategy population: set the maximum evolutionary number max_iter and the number of offspring p, set t = 1, and randomly generate M strategy individuals as the initial population P(0);

[0026] A2: Encoding: Encode each strategy individual in the initial population P(0);

[0027] A3: Selection of fitness function: The shortest computing time required for each resource scheduling solution is used as the fitness function;

[0028] A4: Individual evaluation: According to the fitness function, calculate the fitness of each strategy individual in the strategy population P(t) of the tth iteration, 1≤t≤max_iter;

[0029] A5: Selection operation: Evaluate the entire strategy group, sort the fitness of each strategy individual from small to large, retain the strategy individuals with the highest ranking, and discard the strategy individuals with the longest shortest calculation time.

[0030] A6: Crossover: Cross the remaining groups obtained after the selection operation to generate new strategy combinations;

[0031] A7: Mutation: Randomly mutate the codes of some strategy individuals in the strategy population;

[0032] A8: Calculate the fitness value: Decode each strategy individual in the strategy group obtained in step A7 according to the coding rule in step A2 to obtain a resource allocation plan, substitute the resource allocation plan into the fitness function in step A3, and calculate the fitness value;

[0033] A9: Find the optimal allocation plan: If the fitness values ​​of consecutive generations do not change significantly or reach the maximum number of iterations, the resource allocation plan obtained at this time is the optimal allocation plan; otherwise, set t=t+1 and return to step A4.

[0034] In the above solution, machine learning is used to achieve intelligent task scheduling and optimal resource allocation, thereby improving the system's adaptability and intelligence.

[0035] Preferably, the hybrid computing power calculation module includes a quantum computing unit and a classical computing unit;

[0036] The quantum computing unit is used to receive the quantum computing task and the optimal allocation plan for matching the quantum computing task, execute the quantum computing task, and feed back the computing result to the user terminal;

[0037] The classical computing unit is used to receive the classical computing task and the optimal allocation scheme matching the classical computing task, execute the classical computing task, and feed back the computing result to the user terminal.

[0038] In the above solution, by making full use of the parallel computing power of quantum computing and the stability of classical computing, the algorithm advantages are complementary and the efficiency and performance of the overall computing service are improved.

[0039] Preferably, a monitoring module is also included, which is used to monitor the operating status of the system and the execution of the computing tasks, and to feed back the progress of the computing tasks to the user terminal in a timely manner.

[0040] In the above solution, by strengthening the real-time monitoring and feedback mechanism of the system operation status, problems can be discovered and solved in time to ensure the stable operation of the system.

[0041] A method for managing a computing service with multiple computing forces and heterogeneity is implemented based on the aforementioned computing service management system with multiple computing forces and heterogeneity, and includes the following steps:

[0042] S1: Submit computing tasks and processing requests to the task scheduling module through the user terminal;

[0043] S2: The task scheduling module classifies the computing tasks according to the processing requests, adjusts the execution order of the computing tasks, and initiates resource allocation requests to the resource management module;

[0044] S3: The resource management module uses machine learning to obtain the optimal allocation plan for computing resources based on the resource allocation request;

[0045] S4: The hybrid computing power calculation module executes the computing task according to the optimal allocation plan of computing resources, and returns the obtained computing results to the user terminal.

[0046] Beneficial technical effects of the present invention:

[0047] The present invention provides a computing service management system and method for multi-computing power heterogeneity, which effectively integrates different computing resources and uses machine learning to obtain the optimal allocation plan of computing resources according to resource allocation requests, thereby being able to reasonably allocate computing tasks and computing resources, achieve optimal utilization of computing resources, improve the utilization rate of computing resources, and reduce energy consumption and costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of module connection of the present invention;

[0049] Figure 2 A flowchart of the steps for obtaining the optimal allocation scheme in the present invention;

[0050] Figure 3 It is a flowchart of the implementation steps of the technical solution of the present invention;

[0051] Among them: 1. User terminal; 2. Task scheduling module; 21. Task distinguisher; 22. Task queuer; 3. Resource management module; 31. Resource dynamic scheduler; 32. Load balancer; 33. Virtual resource pool; 331. Quantum computing resource pool; 332. Classical computing resource pool; 4. Hybrid computing power calculation module; 41. Quantum computing unit; 42. Classical computing unit; 5. Monitoring module. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments, but the scope of protection claimed in the present invention is not limited to the following specific embodiments.

[0053] Example 1

[0054] like Figure 1 As shown, a multi-computing power heterogeneous computing power service management system includes a user terminal 1, a task scheduling module 2, a resource management module 3 and a hybrid computing power calculation module 4;

[0055] The user terminal 1 is used to submit computing tasks and processing requests to the task scheduling module 2; the processing request includes task type, priority and computing resource requirements;

[0056] The task scheduling module 2 is used to classify computing tasks according to task types, adjust the execution order of computing tasks according to priorities, and initiate resource allocation requests to the resource management module 3 according to computing resource requirements; the resource allocation request includes task classification results, execution order and computing resource requirements;

[0057] The resource management module 3 is used to obtain an optimal allocation plan for computing resources using machine learning according to a resource allocation request;

[0058] The hybrid computing power calculation module 4 is used to execute computing tasks according to the optimal allocation plan of computing resources, and return the obtained computing results to the user terminal 1.

[0059] During the specific implementation process, by effectively integrating different computing resources and using machine learning to obtain the optimal allocation plan for computing resources based on resource allocation requests, it is possible to reasonably allocate computing tasks and computing resources, achieve optimal utilization of computing resources, improve the utilization rate of computing resources, and reduce energy consumption and costs.

[0060] Example 2

[0061] A computing service management system with multiple computing powers and heterogeneity, comprising a user terminal 1, a task scheduling module 2, a resource management module 3 and a hybrid computing power calculation module 4;

[0062] The user terminal 1 is used to submit computing tasks and processing requests to the task scheduling module 2; the processing request includes task type, priority and computing resource requirements;

[0063] More specifically, the task types include quantum computing tasks and classical computing tasks.

[0064] More specifically, the computing resource requirements include quantum computing resource requirements and classical computing resource requirements.

[0065] The task scheduling module 2 is used to classify computing tasks according to task types, adjust the execution order of computing tasks according to priorities, and initiate resource allocation requests to the resource management module 3 according to computing resource requirements; the resource allocation request includes task classification results, execution order and computing resource requirements;

[0066] More specifically, the task scheduling module 2 includes a task distinguisher 21 and a task queuer 22;

[0067] The task distinguisher 21 is used to classify computing tasks according to task types, adjust the execution order of computing tasks according to priorities, and initiate resource allocation requests to the resource management module 3 according to computing resource requirements;

[0068] The task queuer 22 is used to adjust the execution order of the computing tasks according to the waiting time and priority of the computing tasks, that is, the computing task with the longest waiting time is executed first, and the computing task with the highest priority is executed second.

[0069] The resource management module 3 is used to obtain an optimal allocation plan for computing resources using machine learning according to a resource allocation request;

[0070] More specifically, the resource management module 3 includes a resource dynamic scheduler 31, a load balancer 32 and a virtual resource pool 33;

[0071] The resource dynamic scheduler 31 is used to obtain the optimal allocation plan of computing resources by machine learning according to the resource allocation request, and send the optimal allocation plan to the virtual resource pool 33, and send the computing task to the hybrid computing power calculation module 4;

[0072] The load balancer 32 is used to monitor the load status of the virtual resource pool 33 (such as resource utilization, resource remaining capacity, remaining resource location and other data), and feed back the load status to the resource dynamic scheduler 31;

[0073] The virtual resource pool 33 is used to mark the location of the computing resources to be used according to the optimal allocation plan, and to feedback the load situation to the load balancer 32, and to forward the optimal allocation plan to the hybrid computing power calculation module 4.

[0074] More specifically, the virtual resource pool 33 includes a quantum computing resource pool 331 and a classical computing resource pool 332 .

[0075] In the specific implementation process, virtualization technology is used to virtualize the physical computing resources and manage them. Quantum computing resources and classical computing resources are virtualized into multiple virtual resource units, so that different applications can share the same set of physical resources and improve resource utilization.

[0076] The quantum computing resource pool 331 is used to manage and schedule the quantum computing unit 41, including quantum bits, quantum gates, etc., and provide an execution environment for quantum computing tasks;

[0077] The classic computing resource pool 332 is used to manage and schedule the classic computing unit 42, including CPU, memory, etc., and provide an execution environment for classic computing tasks;

[0078] Considering the heterogeneity of the quantum computing unit 41 and the classical computing unit 42, in this embodiment, the virtual resource pool 33 adopts a hierarchical structure design to manage the quantum computing unit 41 and the classical computing unit 42 separately. The top layer of the virtual resource pool 33 is the quantum computing resource pool 331, and the bottom layer is the classical computing resource pool 332. Through this hierarchical structure, different types of resources can be better distinguished and managed.

[0079] More specifically, the optimal allocation of computing resources is obtained by using machine learning through the following steps:

[0080] A1: Initialize the strategy population: set the maximum evolutionary number max_iter and the number of offspring p, set t = 1, and randomly generate M strategy individuals as the initial population P(0);

[0081] A2: Encoding: Encode each strategy individual in the initial population P(0), that is, encode the computing resource position used in each strategy individual as 1, and encode the unused computing resource position as 0;

[0082] A3: Selection of fitness function: The shortest computing time required for each resource scheduling scheme is used as the fitness function, and the total amount of computing resources consumed by the computing task does not exceed the quantum computing resource capacity and the classical computing resource capacity in the virtual resource pool 33 as a constraint;

[0083] A4: Individual evaluation: According to the fitness function, calculate the fitness of each strategy individual in the strategy group P(t) of the tth iteration, that is, the shortest calculation time corresponding to each strategy individual, 1≤t≤max_iter;

[0084] A5: Selection operation: Evaluate the entire strategy group, sort the fitness of each strategy individual from small to large, retain the strategy individuals with the highest ranking, and discard the individuals with the shortest calculation time corresponding to the strategy; the purpose of the selection operation is to directly pass on the optimized individuals to the next generation, or to generate new individuals through pairing and crossover and then pass them on to the next generation;

[0085] A6: Crossover: Cross the remaining groups obtained after the selection operation to generate new strategy combinations;

[0086] A7: Mutation: Randomly mutate the codes of some strategy individuals in the strategy population, for example, 0 becomes 1 or 1 becomes 0;

[0087] A8: Calculate the fitness value: Decode each strategy individual in the strategy group obtained in step A7 according to the coding rule in step A2 to obtain a resource allocation plan, substitute the resource allocation plan into the fitness function in step A3, and calculate the fitness value;

[0088] A9: Find the optimal allocation plan: If the fitness values ​​of consecutive generations do not change significantly or reach the maximum number of iterations, the resource allocation plan obtained at this time is the optimal allocation plan; otherwise, set t=t+1 and return to step A4.

[0089] During the specific implementation process, machine learning is used to achieve intelligent task scheduling and optimal resource allocation, thereby improving the system's adaptability and intelligence.

[0090] The hybrid computing power calculation module 4 is used to execute computing tasks according to the optimal allocation plan of computing resources, and return the obtained computing results to the user terminal 1.

[0091] More specifically, the hybrid computing power calculation module 4 includes a quantum computing unit 41 and a classical computing unit 42;

[0092] The quantum computing unit 41 is used to receive quantum computing tasks and the optimal allocation scheme for matching quantum computing tasks, execute quantum computing tasks, and feed back computing results to the user terminal 1;

[0093] The classical computing unit 42 is used to receive classical computing tasks and optimal allocation solutions matching the classical computing tasks, execute the classical computing tasks, and feed back the computing results to the user terminal 1.

[0094] During the specific implementation process, by making full use of the parallel computing capabilities of quantum computing and the stability of classical computing, we can achieve complementary algorithm advantages and improve the efficiency and performance of the overall computing power service.

[0095] More specifically, it also includes a monitoring module 5, which is used to monitor the operating status of the system and the execution of computing tasks, regularly evaluate the computing service performance to ensure stable operation of the system, and promptly feedback the progress of computing tasks to the user terminal 1 to improve user experience.

[0096] During the specific implementation process, by strengthening the real-time monitoring and feedback mechanism of the system operation status, problems can be discovered and solved in a timely manner to ensure the stable operation of the system.

[0097] Example 3

[0098] like Figure 3 As shown, a method for managing a computing service with multiple computing forces and heterogeneity is implemented based on the aforementioned computing service management system with multiple computing forces and heterogeneity, and includes the following steps:

[0099] S1: Submit computing tasks and processing requests to task scheduling module 2 through user terminal 1;

[0100] If the computing task is a quantum computing task, the task type for processing the request is a quantum computing task, and the computing resource requirement is a quantum computing resource requirement;

[0101] If the computing task is a classic computing task, the task type for processing the request is a classic computing task, and the computing resource requirement is a classic computing resource requirement;

[0102] S2: The task scheduling module 2 classifies the computing tasks according to the processing request, adjusts the execution order of the computing tasks, and initiates a resource allocation request to the resource management module 3; specifically:

[0103] If the computing task is distinguished as a quantum computing task, the task distinguisher 21 analyzes the priority of the quantum computing task and distinguishes the execution level of the quantum computing task; when multiple quantum computing tasks are distinguished, all quantum computing tasks are sorted according to the priority to obtain the execution order of the quantum computing tasks;

[0104] If the computing task is distinguished as a classical computing task, the task distinguisher 21 analyzes the priority of the classical computing task and distinguishes the execution level of the classical computing task; when multiple classical computing tasks are distinguished, all the classical computing tasks are sorted according to the priority to obtain the execution order of the classical computing tasks;

[0105] When there are multiple computing tasks and computing resources are tight, the task queuer 22 is started to calculate the waiting time of the computing tasks, and the task execution order is reasonably arranged according to the waiting time and priority of the computing tasks. In this embodiment, the tasks are first sorted in descending order according to the waiting time, and then sorted in descending order according to the priority of the tasks, with the task with the longest waiting time being given priority;

[0106] For example: there are quantum computing tasks k1, k2, and k3, and the waiting time of k1, k2, and k3 are 1 minute, 5 minutes, and 3 minutes respectively. They are sorted in descending order according to the waiting time as k2, k3, and k1; the priorities of tasks k1, k2, and k3 are the first level, the third level, and the second level. The final execution order of the quantum computing tasks is k2, k1, and k3;

[0107] The task distinguisher 21 continues to analyze the computing resource requirements corresponding to the computing tasks, and generates corresponding resource allocation requests according to the execution order of the computing tasks, and sends them to the resource dynamic scheduler 31 in sequence;

[0108] S3: The resource management module 3 uses machine learning to obtain the optimal allocation plan for computing resources according to the resource allocation request; the specific steps are:

[0109] A1: Initialize the strategy population: set the maximum evolutionary number max_iter and the number of offspring p, set t = 1, and randomly generate M strategy individuals as the initial population P(0);

[0110] A2: Encoding: Encode each strategy individual in the initial population P(0), that is, encode the computing resource position used in each strategy individual as 1, and encode the unused computing resource position as 0;

[0111] A3: Selection of fitness function: The shortest computing time required for each resource scheduling scheme is used as the fitness function, and the total amount of computing resources consumed by the computing task does not exceed the quantum computing resource capacity and the classical computing resource capacity in the virtual resource pool 33 as a constraint;

[0112] A4: Individual evaluation: According to the fitness function, calculate the fitness of each strategy individual in the strategy group P(t) of the tth iteration, that is, the shortest calculation time corresponding to each strategy individual, 1≤t≤max_iter;

[0113] A5: Selection operation: Evaluate the entire strategy group, sort the fitness of each strategy individual from small to large, retain the strategy individuals with the highest ranking, and discard the individuals with the shortest calculation time corresponding to the strategy; the purpose of the selection operation is to directly pass on the optimized individuals to the next generation, or to generate new individuals through pairing and crossover and then pass them on to the next generation;

[0114] A6: Crossover: Cross the remaining groups obtained after the selection operation to generate new strategy combinations;

[0115] A7: Mutation: Randomly mutate the codes of some strategy individuals in the strategy population, for example, 0 becomes 1 or 1 becomes 0;

[0116] A8: Calculate the fitness value: Decode each strategy individual in the strategy group obtained in step A7 according to the coding rule in step A2 to obtain a resource allocation plan, substitute the resource allocation plan into the fitness function in step A3, and calculate the fitness value;

[0117] A9: Find the optimal allocation plan: If the fitness values ​​of consecutive generations do not change significantly or reach the maximum number of iterations, the resource allocation plan obtained at this time is the optimal allocation plan; otherwise, set t=t+1 and return to step A4.

[0118] S4: The hybrid computing power calculation module 4 executes the computing task according to the optimal allocation plan of computing resources, and returns the obtained computing result to the user terminal 1.

[0119] According to the disclosure and teaching of the above description, those skilled in the art to which the present invention belongs may also make changes and modifications to the above embodiments. Therefore, the present invention is not limited to the specific embodiments disclosed and described above, and some modifications and changes to the invention should also fall within the scope of protection of the claims of the present invention. In addition, although some specific terms are used in this specification, these terms are only for the convenience of description and do not constitute any limitation to the present invention.

Claims

1. A computing power service management system with multiple computing power and heterogeneity, characterized in that: It includes user terminal, task scheduling module, resource management module and hybrid computing power calculation module; The user terminal is used to submit computing tasks and processing requests to the task scheduling module; the processing request includes task type, priority and computing resource requirements; The task scheduling module is used to classify computing tasks according to task types, adjust the execution order of computing tasks according to priorities, and initiate resource allocation requests to the resource management module according to computing resource requirements; the resource allocation request includes task classification results, execution order and computing resource requirements; The resource management module is used to obtain an optimal allocation plan for computing resources using machine learning according to a resource allocation request; The hybrid computing power calculation module is used to perform computing tasks according to the optimal allocation plan of computing resources and return the obtained computing results to the user terminal.

2. A computing service management system with multiple computing forces and heterogeneous computing forces according to claim 1, characterized in that: The task types include quantum computing tasks and classical computing tasks.

3. A computing service management system with multiple computing forces and heterogeneous computing forces according to claim 1, characterized in that: The computing resource requirements include quantum computing resource requirements and classical computing resource requirements.

4. A computing service management system with multiple computing forces and heterogeneous computing forces according to claim 1, characterized in that: The task scheduling module includes a task distinguisher and a task queuer; The task distinguisher is used to classify computing tasks according to task types, adjust the execution order of computing tasks according to priorities, and initiate resource allocation requests to the resource management module according to computing resource requirements; The task queuer is used to adjust the execution order of computing tasks according to the waiting time and priority of the computing tasks.

5. The computing service management system of multiple computing power and heterogeneous computing power according to claim 1, characterized in that: The resource management module includes a resource dynamic scheduler, a load balancer and a virtual resource pool; The resource dynamic scheduler is used to obtain the optimal allocation plan of computing resources by machine learning according to the resource allocation request, and send the optimal allocation plan to the virtual resource pool and send the computing task to the hybrid computing power computing module; The load balancer is used to monitor the load of the virtual resource pool and feed back the load to the dynamic resource scheduler; The virtual resource pool is used to mark the location of the computing resources to be used according to the optimal allocation plan, and to feedback the load situation to the load balancer, and to forward the optimal allocation plan to the hybrid computing power calculation module.

6. A computing service management system and method for multi-computing heterogeneous computing power according to claim 5, characterized in that: The virtual resource pool includes a quantum computing resource pool and a classical computing resource pool.

7. A computing service management system with multiple computing forces and heterogeneous computing forces according to claim 1, characterized in that: The optimal allocation of computing resources is obtained by using machine learning through the following steps: A1: Initialize the strategy population: set the maximum evolutionary number max_iter and the number of offspring p, set t = 1, and randomly generate M strategy individuals as the initial population P(0); A2: Encoding: Encode each strategy individual in the initial population P(0); A3: Selection of fitness function: The shortest computing time required for each resource scheduling solution is used as the fitness function; A4: Individual evaluation: According to the fitness function, calculate the fitness of each strategy individual in the strategy population P(t) of the tth iteration, 1≤t≤max_iter; A5: Selection operation: Evaluate the entire strategy group, sort the fitness of each strategy individual from small to large, retain the strategy individuals with the highest ranking, and discard the strategy individuals with the longest shortest calculation time. A6： Crossover: Cross the remaining groups obtained after the selection operation to generate new strategy combinations; A7: Mutation: Randomly mutate the codes of some strategy individuals in the strategy population; A8: Calculate the fitness value: Decode each strategy individual in the strategy group obtained in step A7 according to the coding rule in step A2 to obtain a resource allocation plan, substitute the resource allocation plan into the fitness function in step A3, and calculate the fitness value; A9: Find the optimal allocation plan: If the fitness values ​​of consecutive generations do not change significantly or reach the maximum number of iterations, the resource allocation plan obtained at this time is the optimal allocation plan; otherwise, set t=t+1 and return to step A4.

8. The computing service management system of multiple computing power and heterogeneous computing power according to claim 1, characterized in that: The hybrid computing power calculation module includes a quantum computing unit and a classical computing unit; The quantum computing unit is used to receive quantum computing tasks and the optimal allocation plan for matching quantum computing tasks, execute quantum computing tasks, and feed back computing results to the user terminal; The classical computing unit is used to receive the classical computing task and the optimal allocation scheme matching the classical computing task, execute the classical computing task, and feed back the computing result to the user terminal.

9. The computing service management system of multiple computing power and heterogeneous computing power according to claim 1, characterized in that: It also includes a monitoring module for monitoring the operating status of the system and the execution of computing tasks, and timely feeding back the progress of computing tasks to the user terminal.

10. A method for managing computing services with multiple computing forces and heterogeneity, characterized in that: The following steps are involved: S1: Submit computing tasks and processing requests to the task scheduling module through the user terminal; S2: The task scheduling module classifies the computing tasks according to the processing requests, adjusts the execution order of the computing tasks, and initiates resource allocation requests to the resource management module; S3: The resource management module uses machine learning to obtain the optimal allocation plan for computing resources based on the resource allocation request; S4: The hybrid computing power calculation module executes the computing task according to the optimal allocation plan of computing resources, and returns the obtained computing results to the user terminal.

Images