Method and device for determining computing power scheduling scheme, equipment and medium

Through the combination of quantum approximation optimization algorithm and mixed integer linear programming algorithm, the problem of low computing power scheduling efficiency and resource utilization in the existing technology is solved, and the rapid and effective determination of optimal computing power scheduling scheme is achieved, which improves the scheduling efficiency and resource utilization.

CN118963954AActive Publication Date: 2024-11-15CHINA MOBILE GROUP DESIGN INST +1

Patent Information

Application Number
CN202411065749.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-05
Publication Date
2024-11-15
Estimated Expiration
2044-08-05

AI Technical Summary

Technical Problem

The existing computing power scheduling methods rely on classical algorithms, which leads to long calculation time and easy to fall into local optimal solutions, making it difficult to quickly and effectively determine the optimal computing power scheduling scheme, especially in large-scale computing tasks, resulting in scheduling. Low efficiency and resource utilization.

Method used

The hybrid framework of quantum approximation optimization algorithm and mixed integer linear programming algorithm is used to predict the computing power requirements of the calculation task based on the state parameters of the computing power system, and the target solution space is determined, and the combination of quantum approximation optimization algorithm and mixed integer linear programming algorithm is used. , optimize scheduling schemes to improve efficiency and resource utilization.

Benefits of technology

It significantly shortens the decision-making time of computing power scheduling, improves resource utilization, reduces computing power costs, adapts to the complex and changing computing power demand environment, improves scheduling efficiency by about 30%, improves resource utilization by about 15%, and has a task completion rate Increased by about 10%.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118963954A_ABST
    Figure CN118963954A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of computing power network intelligent scheduling, and provides a method and device for determining a computing power scheduling scheme, equipment and a medium. The method comprises the following steps: determining predicted computing power requirements corresponding to different computing tasks according to state parameters of a computing power system; determining a solution space corresponding to the target computing power scheduling problem by adopting a quantum approximate optimization algorithm according to the state parameters and the predicted computing power demand, and performing solution extraction on the state parameters and the predicted computing power demand to obtain a first computing power scheduling scheme; a mixed integer linear programming algorithm is adopted to refine and optimize the solution space, and a second computing power scheduling scheme is determined; and determining an optimal computing power scheduling scheme corresponding to the computing power system according to the first computing power scheduling scheme and the second computing power scheduling scheme. The method aims at fusing the search acceleration capability of a quantum approximate optimization algorithm and the stability and reliability characteristics of a mixed integer linear programming algorithm, and improving the scheduling efficiency and the resource utilization rate of a computing power system so as to quickly and effectively determine an optimal computing power scheduling scheme.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of intelligent scheduling of computing power networks, and in particular to a method, device, equipment and medium for determining a computing power scheduling solution. Background Art

[0002] At present, with the advancement of the construction of the computing power hub of the "East Data West Computing" national integrated big data center (Data Center, DC), the concept of "Computing Force Network (CFN)" has emerged. The core idea of ​​the computing power network is to connect the geographically distributed computing power center nodes through new network technologies, dynamically and in real time perceive the status of computing power resources, and then coordinate the allocation and scheduling of computing tasks, realize data transmission, and form a network that perceives, allocates, and schedules computing power on a global scale, and on this basis, gather and share computing power, data and application resources. Based on the integrated computing power transaction scheduling platform of "computing power available, controllable, and measurable", the commonly used computing power scheduling methods can include: centralized scheduling, distributed scheduling, hierarchical scheduling, and hybrid scheduling, etc., mainly through dynamic and precise scheduling of the computing power supply side (such as the computing power system) to adjust resource allocation.

[0003] With the rapid development of cloud computing, big data, and artificial intelligence (AI) technologies, the complexity of computing tasks continues to increase, which poses a more severe challenge to computing power scheduling. The computing power scheduling problem is essentially a multi-constraint, multi-objective optimization problem involving multiple aspects such as resource allocation, task scheduling, and cost-benefit analysis. In a cloud computing environment, computing tasks usually need to be executed on multiple physical or virtual servers, and each computing task has specific computing requirements and deadlines. The goal of resource allocation is to minimize resource usage costs or maximize resource utilization while meeting the requirements of all computing tasks. Due to the diversity of computing tasks and computing power resources, and the dynamic changes in demand, computing power scheduling problems are usually large-scale and highly complex.

[0004] Existing computing power scheduling methods mainly rely on classical algorithms for dynamic scheduling of computing power supply side. In the field of traditional computing, computing power scheduling problems are classified as combinatorial optimization problems. However, when using classical algorithms to solve computing power scheduling problems, they often face disadvantages such as long calculation time and easy to fall into local optimal solutions. Especially for large-scale problems, the computational complexity of classical algorithms increases exponentially with the scale of the problem, which limits the application of classical algorithms in real-time or near real-time environments. The whole process has certain limitations, which makes the scheduling efficiency and resource utilization of the computing power system low, and ultimately makes it impossible to quickly and effectively determine the optimal computing power scheduling plan. Summary of the invention

[0005] The embodiments of the present application provide a method, device, equipment and medium for determining a computing power scheduling plan, which are used to solve the technical problem in the prior art that the use of classical algorithms results in low scheduling efficiency and resource utilization of the computing power system, ultimately leading to the inability to quickly and effectively determine the optimal computing power scheduling plan.

[0006] In a first aspect, an embodiment of the present application provides a method for determining a computing power scheduling plan, comprising the following steps.

[0007] According to the state parameters of the computing power system, the predicted computing power requirements corresponding to different computing tasks are determined.

[0008] A quantum approximate optimization algorithm is used to determine a solution space corresponding to the target computing power scheduling problem according to the state parameters and the predicted computing power demand, and solutions are extracted for the state parameters and the predicted computing power demand to obtain a first computing power scheduling solution.

[0009] A mixed integer linear programming algorithm is used to refine and optimize the solution space to determine a second computing power scheduling solution.

[0010] According to the first computing power scheduling scheme and the second computing power scheduling scheme, an optimal computing power scheduling scheme corresponding to the computing power system is determined.

[0011] In one embodiment, the use of a quantum approximate optimization algorithm to determine a solution space corresponding to a target computing power scheduling problem according to the state parameters and the predicted computing power demand includes: performing a two-dimensional unconstrained optimization encoding on the computing power scheduling problem to obtain the target computing power scheduling problem; and using the quantum approximate optimization algorithm to determine a solution space corresponding to the target computing power scheduling problem according to the state parameters and the predicted computing power demand.

[0012] In one embodiment, the predicted computing power requirements corresponding to different computing tasks are determined based on the state parameters of the computing power system, including: determining the state parameters based on the performance indicators, task load, remaining computing power resources and historical scheduling records of the computing power system; inputting the state parameters into a long short-term memory network model to obtain the predicted computing power requirements output by the long short-term memory network model; wherein the long short-term memory network model is trained based on the historical state data of the computing power system and the historical actual computing power requirements.

[0013] In one embodiment, determining the optimal computing power scheduling scheme corresponding to the computing power system according to the first computing power scheduling scheme and the second computing power scheduling scheme includes: determining a first weight corresponding to the first computing power scheduling scheme and a second weight corresponding to the second computing power scheduling scheme; determining the optimal computing power scheduling scheme according to the first computing power scheduling scheme, the second computing power scheduling scheme, the first weight and the second weight; or, using a preset resource rule fusion method to fuse the first computing power scheduling scheme and the second computing power scheduling scheme to obtain the optimal computing power scheduling scheme.

[0014] In one embodiment, the method further includes: monitoring the current network status and task execution status of the computing power system in real time; adjusting the parameters of the quantum approximate optimization algorithm and the parameters of the mixed integer linear programming algorithm according to the current network status and the task execution status.

[0015] In one embodiment, the long short-term memory network model is trained based on the following steps: obtaining the historical status data and the historical actual computing power requirements; inputting the historical status data into the initial long short-term memory network model to obtain the historical predicted computing power requirements output by the initial long short-term memory network model; updating the model parameters of the initial long short-term memory network model according to the historical predicted computing power requirements and the historical actual computing power requirements to obtain a trained long short-term memory network model.

[0016] In a second aspect, an embodiment of the present application provides a device for determining a computing power scheduling scheme, comprising the following modules.

[0017] The data processing module is used to determine the predicted computing power requirements corresponding to different computing tasks based on the state parameters of the computing power system.

[0018] The resource allocation module is used to use a quantum approximate optimization algorithm to determine the solution space corresponding to the target computing power scheduling problem according to the state parameters and the predicted computing power demand, and extract the solution of the state parameters and the predicted computing power demand to obtain a first computing power scheduling plan; use a mixed integer linear programming algorithm to refine and optimize the solution space to determine a second computing power scheduling plan; and determine the optimal computing power scheduling plan corresponding to the computing power system according to the first computing power scheduling plan and the second computing power scheduling plan.

[0019] In a third aspect, an embodiment of the present application provides an electronic device, including a memory, a transceiver, and a processor.

[0020] A memory is used to store a computer program; a transceiver is used to send and receive data under the control of the processor; and a processor is used to read the computer program in the memory and perform the following operations.

[0021] According to the state parameters of the computing power system, the predicted computing power requirements corresponding to different computing tasks are determined; using a quantum approximate optimization algorithm, according to the state parameters and the predicted computing power requirements, the solution space corresponding to the target computing power scheduling problem is determined, and the solution of the state parameters and the predicted computing power requirements is extracted to obtain a first computing power scheduling plan; using a mixed integer linear programming algorithm, the solution space is refined and optimized to determine a second computing power scheduling plan; according to the first computing power scheduling plan and the second computing power scheduling plan, the optimal computing power scheduling plan corresponding to the computing power system is determined.

[0022] In a fourth aspect, an embodiment of the present application provides an electronic device, comprising a processor and a memory storing a computer program, wherein the processor implements the steps of the method for determining a computing power scheduling scheme described in the first aspect when executing the computer program.

[0023] In a fifth aspect, an embodiment of the present application further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for determining a computing power scheduling scheme described in the first aspect.

[0024] In a sixth aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the method for determining the computing power scheduling scheme described in the first aspect.

[0025] The method, device, equipment and medium for determining the computing power scheduling scheme provided in the embodiments of the present application determine the predicted computing power requirements corresponding to different computing tasks according to the state parameters of the computing power system; use a quantum approximate optimization algorithm to determine the solution space corresponding to the target computing power scheduling problem according to the state parameters and the predicted computing power requirements, and extract the solution of the state parameters and the predicted computing power requirements to obtain a first computing power scheduling scheme; use a mixed integer linear programming algorithm to refine and optimize the solution space to determine the second computing power scheduling scheme; determine the optimal computing power scheduling scheme corresponding to the computing power system according to the first computing power scheduling scheme and the second computing power scheduling scheme. This method uses a hybrid framework of a quantum approximate optimization algorithm and a mixed integer linear programming algorithm, aiming to integrate the search acceleration capability of the quantum approximate optimization algorithm and the stable and reliable characteristics of the mixed integer linear programming algorithm, improve the scheduling efficiency and resource utilization of the computing power system, and quickly and effectively determine the optimal computing power scheduling scheme. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the present application or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0027] Figure 1 It is a schematic diagram of the system architecture of the computing power network provided by the existing technology.

[0028] Figure 2 It is a schematic diagram of the hybrid computing architecture of the quantum-classical framework provided in an embodiment of the present application.

[0029] Figure 3 It is a flowchart of a method for determining a computing power scheduling solution provided in an embodiment of the present application.

[0030] Figure 4 It is a scenario diagram of a method for determining a computing power scheduling solution provided in an embodiment of the present application.

[0031] Figure 5 It is a structural diagram of a device for determining a computing power scheduling solution provided in an embodiment of the present application.

[0032] Figure 6 This is one of the structural schematic diagrams of the electronic device provided in the embodiment of the present application.

[0033] Figure 7 This is the second structural schematic diagram of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0035] In order to better understand the embodiments of the present application, the prior art is first described in detail.

[0036] Figure 1 This is a schematic diagram of the system architecture of the computing network provided by existing technology. Figure 1 The computing power network can include: AI empowerment platform, computing network consumers, computing power trading platform, computing power providers, computing power network control plane, end computing nodes, multiple cloud computing centers, and edge computing nodes.

[0037] Based on the above computing power network, the commonly used computing power scheduling methods currently include: centralized scheduling, distributed scheduling, hierarchical scheduling and hybrid scheduling.

[0038] Among them, centralized scheduling: all computing resources are centralized and uniformly managed and allocated by a centralized scheduler. The advantages are strong control and good global optimization effect, but the high degree of centralization may cause single point failures and affect the stability of the entire computing system.

[0039] Distributed scheduling: computing resources are distributed to different nodes, and each node has a certain degree of decision-making power and autonomy. The advantage is high flexibility and the ability to adapt to dynamically changing environments, but there may be problems such as load imbalance and low coordination efficiency.

[0040] Hierarchical scheduling: Computing resources are divided into levels, and each level is managed by a corresponding scheduler. The high-level scheduler is responsible for global optimization, and the low-level scheduler is responsible for local optimization. The advantage is that it takes into account the advantages of centralized and distributed scheduling, but it is more complicated to implement and also has the problem of single point failure.

[0041] Hybrid scheduling: Comprehensive use of centralized scheduling, distributed scheduling and hierarchical scheduling. The advantage is that the most appropriate scheduling strategy is selected according to the actual situation, but the coordination and compatibility between different methods need to be considered.

[0042] The above commonly used computing power scheduling methods mainly rely on classical algorithms for dynamic scheduling of computing power supply side. In the field of traditional computing, computing power scheduling problems are classified as combinatorial optimization problems. However, when using classical algorithms to solve computing power scheduling problems, they often face disadvantages such as long calculation time and easy to fall into local optimal solutions. Especially for large-scale problems, the computational complexity of classical algorithms increases exponentially with the scale of the problem, which limits the application of classical algorithms in real-time or near real-time environments. The whole process has certain limitations, which makes the scheduling efficiency and resource utilization of the computing power system low, and ultimately makes it impossible to quickly and effectively determine the optimal computing power scheduling plan.

[0043] In summary, existing computing power scheduling methods often have difficulty in achieving optimal resource allocation and scheduling strategies when dealing with large-scale, dynamic, and highly heterogeneous computing power scheduling problems. In other words, they cannot quickly and effectively determine the optimal computing power scheduling solution. At the same time, they have limitations in terms of resource utilization, system flexibility, management and maintenance costs, and system reliability. In addition, for large-scale and dynamically changing computing power requirements, existing computing power scheduling methods are also insufficient in predicting computing power requirements and real-time scheduling. Classic algorithms face challenges in terms of real-time performance, optimization effects, and computing resource consumption.

[0044] In order to solve the above technical problems, the embodiment of the present application provides a method for determining a computing power scheduling scheme, by determining the predicted computing power requirements corresponding to different computing tasks according to the state parameters of the computing power system; using a quantum approximate optimization algorithm, according to the state parameters and the predicted computing power requirements, determining the solution space corresponding to the target computing power scheduling problem, and extracting the solution of the state parameters and the predicted computing power requirements to obtain a first computing power scheduling scheme; using a mixed integer linear programming algorithm, refining and optimizing the solution space to determine the second computing power scheduling scheme; according to the first computing power scheduling scheme and the second computing power scheduling scheme, determining the optimal computing power scheduling scheme corresponding to the computing power system. This method uses a hybrid framework of a quantum approximate optimization algorithm and a mixed integer linear programming algorithm, aiming to integrate the search acceleration capability of the quantum approximate optimization algorithm and the stable and reliable characteristics of the mixed integer linear programming algorithm, improve the scheduling efficiency and resource utilization of the computing power system, and quickly and effectively determine the optimal computing power scheduling scheme.

[0045] In order to better understand the embodiments of the present application, a method for determining the computing power scheduling solution provided in the embodiments of the present application is preliminarily explained below.

[0046] Quantum computing is significantly different from traditional classical computing. It is a computing paradigm based on the principles of quantum mechanics. In this computing paradigm, quantum bits (qubits), as the basic units of information storage and processing, exhibit unique physical properties, such as quantum superposition and quantum entanglement. These physical properties give quantum computers the potential to surpass traditional computers when processing specific computing tasks.

[0047] Recent studies have pointed out that, thanks to the physical properties of quantum superposition and quantum entanglement, quantum computing has shown significant advantages over classical computing when processing large-scale and complex data and computing tasks. Potential advantages include faster search speed, higher parallelism, and better global optimal solution search capabilities. Quantum algorithms developed using quantum entanglement have significantly improved the ability of parallel computing, allowing quantum computers to perform exponentially faster when processing actual information processing computing tasks.

[0048] With the advancement of quantum computing technology, quantum algorithms such as the Quantum Approximate Optimization Algorithm (QAOA) can effectively solve certain computing task scheduling problems. In theory, they can accelerate the computing power scheduling problem and increase the speed of determining the optimal computing power scheduling solution.

[0049] Among them, QAOA is a quantum variational algorithm used to solve combinatorial optimization problems. The core idea of ​​QAOA is to use quantum superposition states to search the solution space of the problem. A quantum state is prepared by a parameterized quantum circuit, and the measurement result of the quantum state can give an approximate solution to the optimization problem. For the Mixed Integer Linear Programming (MILP) problem, it can be approximated by converting integer variables into soft constraints, and then QAOA is used to find the global optimal solution or the approximate optimal solution.

[0050] It should be noted that the implementation of QAOA involves two main parts, namely the alternating application of unitary rotation gates and Hamiltonian evolution gates. Among them, the unitary rotation gate is used to prepare the initial superposition state, and the Hamiltonian evolution gate is used to simulate the Hamiltonian of the problem. By optimizing the parameters of the rotation gate, the output state of the quantum circuit can be improved, thus getting closer to the optimal solution of the problem.

[0051] Since the mathematical description of QAOA involves multiple parameters and quantum gate operations, only a simplified version of the QAOA formula is provided here: QAOA( )= C, ).

[0052] in, represents the first parameter vector; represents the second parameter vector; QAOA( ) represents the quantum search results; Indicates the depth or number of layers of QAOA; Indicates the first rotation angle corresponding to the k-th layer depth or number of layers; Indicates the second rotation angle corresponding to the k-th layer depth or number of layers; represents the diagonalized Hamiltonian; C represents the cost function; represents a unitary revolving door; represents the Hamiltonian evolution gate. Among them, the first rotation angle and the second rotation angle Determined by an optimization algorithm such as gradient descent.

[0053] The following is a detailed description of other related terms involved in the method for determining the computing power scheduling solution provided in the embodiment of the present application.

[0054] (1) The MILP algorithm is a mathematical optimization algorithm used to solve linear programming problems that contain both integer and continuous variables. In MILP problems, both the objective function and the constraints are linear, but some or all of the variables are restricted to integer values.

[0055] Specifically, the computing power scheduling problem can usually be modeled as a large-scale MILP problem. Assume there is a computing task set T = and a computing resource set R= , where t i represents the i-th computing task among n computing tasks, n ≥ 2, r j represents the jth computing resource among m computing resources, m ≥ 2. Each computing task t i Requires a certain amount of computing resources i,j Come in time i Completed within, each computing resource r j Each corresponds to a capacity c j The goal is to minimize the resource usage cost or maximize resource utilization while meeting all task requirements. In the MILP problem, parameters such as resource cost and task benefit can be adjusted according to the actual scenario. The MILP problem can be mathematically expressed as: Minimize ; Subject to , i ; , j ; , i , j ; in, Represents a decision variable. If the computational task Assigned to resources On, then =1, otherwise, =0.

[0056] (2) Quadratic Unconstrained Binary Optimization (QUBO) is a combinatorial optimization problem that adds two key constraints to the quadratic programming (QP): unconstrained conditions and binary decision variables.

[0057] Specifically, the computing power scheduling problem can be modeled as a QUBO problem. Indicates the allocation status of computing resources; Q represents an n×n symmetric matrix, Qij Represents the penalty or reward for computing task assignments.

[0058] (3) The Long Short-Term Memory (LSTM) model is a machine learning model used to predict future task computing power requirements P(t).

[0059] That is to say, the method for determining the computing power scheduling plan provided in the above-mentioned embodiment of the present application combines the search acceleration capability of quantum algorithms (such as QAOA) when dealing with complex optimization problems and the stable, reliable characteristics and feasibility of classical algorithms (such as MILP algorithms) in practical applications, thereby improving the scheduling efficiency and resource utilization of the computing power system, so as to quickly and effectively determine the optimal computing power scheduling plan.

[0060] For example, Figure 2 Schematic diagram of the hybrid computing architecture of the quantum-classical framework provided in the embodiment of the present application. Figure 2 The quantum-classical framework includes a quantum computing module, a classical computing module, a timing prediction module, and a hybrid scheduling decision module. The quantum computing module uses a quantum computer or a quantum simulator to perform QAOA and efficiently search for large-scale optimization problems. The classical computing module uses a classical computer to execute the MILP algorithm to refine and optimize the quantum search results. The timing prediction module predicts future task computing power requirements based on historical data and machine learning technology to provide a basis for scheduling decisions. The hybrid scheduling decision module combines quantum search results and classical optimization results to generate the optimal computing power scheduling plan.

[0061] The following describes in detail a method for determining a computing power scheduling solution provided in an embodiment of the present application, taking an electronic device as an example.

[0062] Figure 3 is a flow chart of a method for determining a computing power scheduling solution provided in an embodiment of the present application. Figure 3 , an embodiment of the present application provides a method for determining a computing power scheduling plan, which may include the following steps 301-304.

[0063] Step 301: Determine the predicted computing power requirements corresponding to different computing tasks based on the state parameters of the computing power system.

[0064] Among them, the state parameters refer to a series of indicators or data that describe the current operating status, performance and resource utilization of the computing system. Optionally, the state parameters may include: performance indicators, task load, remaining computing resources, task requirements and network conditions, etc.

[0065] Computing tasks refer to specific work units that need to be performed on the computing system. Optionally, computing tasks may include: data analysis, model training, image processing, and scientific computing.

[0066] Predicting computing power demand refers to the prediction of computing power resources required to perform a specific computing task based on the current state parameters of the computing power system and the characteristics of the computing task. The entire prediction process helps electronic devices allocate sufficient computing power resources in advance to ensure that computing tasks can be executed smoothly and maximize the overall performance and efficiency of the computing power system.

[0067] In some embodiments, the electronic device determines the predicted computing power requirements corresponding to different computing tasks based on the state parameters of the computing power system, which may include: the electronic device determines the state parameters based on the performance indicators, task load, remaining computing power resources and historical scheduling records of the computing power system; the electronic device inputs the state parameters into the long short-term memory network model to obtain the predicted computing power requirements output by the long short-term memory network model.

[0068] Among them, performance indicators refer to a series of quantitative standards that measure the performance of a computing system when performing computing tasks.

[0069] Task load refers to the total amount and complexity of computing tasks that the computing system is currently executing or about to execute.

[0070] Remaining computing resources refer to the resources remaining in the computing system that can be used for new or additional computing tasks after removing the resources allocated to the computing tasks being executed.

[0071] The historical scheduling record refers to a detailed record of the allocation and scheduling of computing resources by the electronic device over a period of time in the past. Optionally, the historical scheduling record may include: task completion time, resource utilization rate, task success rate, etc.

[0072] The long short-term memory network model is a time series prediction model that is trained based on the historical state data of the computing system and the historical actual computing power requirements. Optionally, the historical state data may include: the computing requirements, task submission time, task completion time, and other features of each historical computing task in the past period of time.

[0073] In the process of data collection, electronic devices can first obtain the state parameters of the computing system, such as performance indicators, task load, remaining computing resources, and historical scheduling records, and then use the state parameters as input data to input into the long short-term memory network model. The long short-term memory network model captures the long-term dependencies in the state parameters, and predicts the computing power requirements of future tasks. The predicted computing power requirements are obtained with high accuracy, providing data support for the subsequent determination of the first computing power scheduling plan and the second computing power scheduling plan. The whole process combines time series prediction, which can effectively optimize the allocation of computing power resources and task scheduling. In addition, by guiding computing power scheduling through time series prediction, adapting to the dynamic characteristics of computing power requirements, it can also predict changes in computing power requirements in real time, making subsequent scheduling decisions more forward-looking.

[0074] In some embodiments, the long short-term memory network model is trained based on the following steps: the electronic device obtains historical status data and historical actual computing power requirements; the electronic device inputs the historical status data into the initial long short-term memory network model to obtain the historical predicted computing power requirements output by the initial long short-term memory network model; the electronic device updates the model parameters of the initial long short-term memory network model based on the historical predicted computing power requirements and the historical actual computing power requirements to obtain a trained long short-term memory network model.

[0075] In the process of training the initial long short-term memory network model, the historical state data and the historical actual computing power requirements can be obtained first, and the historical state data can be used as training data; then, the training data is input into the initial long short-term memory network model, and the long-term dependencies in the historical state parameters are captured by the initial long short-term memory network model to predict the computing power requirements of future tasks and obtain the historical predicted computing power requirements; since there is a certain difference between the historical predicted computing power requirements and the historical actual computing power requirements, the model parameters of the above-mentioned initial long short-term memory network model can be updated according to the loss value between the historical predicted computing power requirements and the historical actual computing power requirements to obtain a trained long short-term memory network model, so as to be able to accurately predict the computing power requirements of different computing tasks in the future.

[0076] Step 302: Using a quantum approximate optimization algorithm, determine the solution space corresponding to the target computing power scheduling problem according to the state parameters and the predicted computing power demand, and extract the solution of the state parameters and the predicted computing power demand to obtain a first computing power scheduling plan.

[0077] After determining the state parameters and predicting the computing power requirements, the electronic device can execute the quantum approximate optimization algorithm in the quantum simulator, and perform a global optimization search for resource allocation based on the state parameters and predicted computing power requirements to obtain quantum search results, that is, the solution space corresponding to the target computing power scheduling problem, and perform quantum state measurement based on the state parameters and predicted computing power requirements to extract the solution and obtain the first computing power scheduling plan. The entire quantum search process makes full use of the superposition and entanglement state characteristics of quantum bits, and can quickly converge to the approximate optimal solution in a large-scale search space.

[0078] In some embodiments, the electronic device uses a quantum approximate optimization algorithm to determine the solution space corresponding to the target computing power scheduling problem based on state parameters and predicted computing power requirements, which may include: the electronic device performs two-dimensional unconstrained optimization encoding on the computing power scheduling problem to obtain the target computing power scheduling problem; the electronic device uses a quantum approximate optimization algorithm to determine the solution space corresponding to the target computing power scheduling problem based on state parameters and predicted computing power requirements.

[0079] In the process of determining the solution space corresponding to the target computing power scheduling problem, the electronic device can first obtain the computing power scheduling problem, and perform a two-dimensional unconstrained optimization encoding on the computing power scheduling problem to obtain the target computing power scheduling problem. The target computing power scheduling problem is a quadratic unconstrained optimization problem. In this way, a vector or matrix can be found to make the corresponding quadratic function reach the maximum value (i.e., the minimum value or maximum value). This encoding method helps to abstract complex computing power scheduling problems and facilitates the subsequent use of quantum approximate optimization algorithms for solving them. That is, a quantum approximate optimization algorithm is used to perform a global optimization search for resource allocation based on state parameters and predicted computing power requirements to obtain the solution space corresponding to the target computing power scheduling problem. The accuracy of the solution space is also relatively high.

[0080] Specifically, in the process of encoding the computing power scheduling problem, the various elements of the computing power scheduling problem (such as computing power resources, task requirements, constraints, etc.) can be mathematically modeled to form a clear optimization objective function and constraint set; then, the decision variables are defined in the quadratic function, and each decision variable usually represents the computing task t i Assigned to resource r j Then, according to the optimization goal (such as minimizing resource usage cost or maximizing resource utilization), the corresponding quadratic objective function is constructed, namely the target computing power scheduling problem. Among them, the quadratic objective function is used to reflect the performance of different computing power scheduling schemes.

[0081] It should be noted that the two-dimensional unconstrained optimization coding and quantum approximate optimization algorithm can solve the computing power scheduling problem in a fast, efficient and innovative way. The whole process not only utilizes the powerful computing power of quantum computing, but also approaches the optimal solution through continuous iteration and improvement of the optimization algorithm.

[0082] Step 303: Use a mixed integer linear programming algorithm to refine and optimize the solution space and determine the second computing power scheduling plan.

[0083] Since the above solution space is a rough solution, in order to improve the accuracy of the solution space, the solution space can be used as the initial solution of the mixed integer linear programming algorithm, and combined with the mixed integer linear programming algorithm's expertise in processing precise constraints and local optimization, the mixed integer linear programming algorithm is used, and mathematical tools (such as digital optimization technology (CPLEX) or Gurobi and other tools) are used to refine and optimize the solution space to obtain the second computing power scheduling solution. Among them, the second computing power scheduling solution can meet all resource constraints (such as server capacity, task allocation rules, etc.) and task completion time requirements, and has high accuracy and feasibility.

[0084] Specifically, the optimization goal in the mixed integer linear programming algorithm is defined as minimizing the resource usage cost or maximizing the resource utilization rate; the continuous variable is the amount of resources allocated to each computing task; the integer variable is usually used to represent the decision variable, such as the server number to which the computing task is assigned, the execution order of the computing task, etc. Then, according to the target computing power scheduling problem, the objective function is constructed. For example, if the goal is to minimize the resource usage cost, the objective function may be the sum of the resource usage cost.

[0085] In addition, the resource constraints in the mixed integer linear programming algorithm need to be defined. Optionally, the resource constraints include: resource constraints, task dependencies, performance requirements, and integer constraints, among which resource constraints are used to ensure that the total usage of each server or resource pool does not exceed its capacity; task dependencies are used to ensure that computing tasks are executed in the correct order when there are dependencies between computing tasks; performance requirements are used to ensure that the performance indicators such as response time and throughput of each computing task meet the requirements; integer constraints are used to ensure that integer variables take integer values.

[0086] Then, the mixed integer linear programming algorithm mentioned above can be used to quickly find the solution of the objective function that satisfies all resource constraints (i.e., the second computing power scheduling solution).

[0087] Step 304: Determine the optimal computing power scheduling scheme corresponding to the computing power system according to the first computing power scheduling scheme and the second computing power scheduling scheme.

[0088] After determining the first computing power scheduling scheme and the second computing power scheduling scheme, the electronic device can process the two computing power scheduling schemes to obtain the optimal computing power scheduling scheme corresponding to the computing power system, which not only reflects the global vision of quantum search, but also ensures that the resource constraints of the computing power system are met. The whole process can ensure that computing power resources are allocated and utilized in the most reasonable way, while avoiding the waste and idleness of computing power resources and improving the resource utilization efficiency of the overall computing power system.

[0089] In some embodiments, the electronic device determines the optimal computing power scheduling scheme corresponding to the computing power system based on the first computing power scheduling scheme and the second computing power scheduling scheme, which may include the following implementation method 1 or implementation method 2.

[0090] Implementation method 1: The electronic device determines a first weight corresponding to a first computing power scheduling scheme and a second weight corresponding to a second computing power scheduling scheme; the electronic device determines an optimal computing power scheduling scheme based on the first computing power scheduling scheme, the second computing power scheduling scheme, the first weight and the second weight.

[0091] The first weight and the second weight may be the same or different, and are not specifically limited here.

[0092] After determining the first computing power scheduling scheme and the second computing power scheduling scheme, the electronic device may first determine the first weight corresponding to the first computing power scheduling scheme, and determine the second weight corresponding to the second computing power scheduling scheme; then, the electronic device performs weighted allocation on the first computing power scheduling scheme, the first weight, the second computing power scheduling scheme, and the second weight to obtain the optimal computing power scheduling scheme.

[0093] In summary, electronic devices use a weight distribution method to assign different weights to different computing power scheduling schemes or each evaluation index therein. Through weighted average, the advantages and disadvantages of each computing power scheduling scheme are comprehensively evaluated to obtain the optimal solution (i.e., the optimal computing power scheduling scheme).

[0094] Implementation method 2: The electronic device adopts a preset resource rule fusion method to fuse the first computing power scheduling plan and the second computing power scheduling plan to obtain the optimal computing power scheduling plan.

[0095] Among them, the preset resource rule fusion method is intended to guide how to effectively merge the first computing power scheduling plan and the second computing power scheduling plan to obtain an optimal computing power scheduling plan that can meet the current and future needs of the computing power system.

[0096] Optionally, the preset resource rule fusion method may include: priority rules, conflict resolution mechanisms, dynamic adjustment strategies, resource optimization principles, etc.

[0097] Among them, the priority rule is used to set priorities for different computing power scheduling schemes or some of their elements (such as resource utilization, task response time, energy consumption, etc.). In the fusion process, the schemes or elements with high priority are given priority.

[0098] Conflict resolution mechanism: During the fusion process, conflicts between different computing power scheduling schemes may occur (such as resource allocation conflicts, task execution order conflicts, etc.). The preset resource rule fusion method should include an effective conflict resolution mechanism to ensure the smooth progress of the fusion process.

[0099] Dynamic adjustment strategy: Considering that the system status and task requirements may change over time, the preset resource rule fusion method should also include a set of dynamic adjustment strategies, allowing the computing power system to fine-tune the computing power scheduling plan according to actual conditions during operation to ensure that the computing power system always remains in the optimal state.

[0100] Resource optimization principle: In the integration process, we always follow the principle of resource optimization. Through reasonable resource allocation and scheduling, we can maximize resource utilization efficiency and reduce energy consumption and costs.

[0101] In some embodiments, the method may also include: the electronic device monitors the current network status and task execution status of the computing system in real time; the electronic device adjusts the parameters of the quantum approximate optimization algorithm and the parameters of the mixed integer linear programming algorithm according to the current network status and task execution status.

[0102] In the process of dynamically adjusting algorithm parameters, electronic devices can adopt a feedback mechanism. Specifically, in the process of executing the optimal target scheduling plan, the current network status and task execution status of the computing power network can be monitored in real time, and the parameters of the quantum approximate optimization algorithm and the mixed integer linear programming algorithm can be dynamically adjusted according to the actual operation data. Even the strategy weights can be adjusted to form a closed-loop optimization and continuously optimize the scheduling effect, thereby improving the reliability of the computing power system and reducing system failures and downtime.

[0103] For example, Figure 4 is a schematic diagram of a scenario of a method for determining a computing power scheduling solution provided in an embodiment of the present application, combined with Figure 2 The quantum-classical framework shown, from Figure 4 It can be seen that in the data collection and demand forecasting stage, the state parameters of the computing power system can be obtained, and based on the state parameters, the predicted computing power requirements corresponding to different computing tasks can be determined.

[0104] In the quantum search stage, the computing power scheduling problem can be encoded by a two-dimensional unconstrained optimization to obtain the target computing power scheduling problem; the quantum approximate optimization algorithm is used to determine the solution space corresponding to the target computing power scheduling problem according to the state parameters and the predicted computing power demand, and the solutions of the state parameters and the predicted computing power demand are extracted to obtain the first computing power scheduling plan.

[0105] In the classical optimization stage, the solution space can be used as the initial solution; the mixed integer linear programming algorithm is used to refine and optimize the initial solution to obtain the exact solution, which is the second computing power scheduling plan.

[0106] In the scheduling decision stage, weight allocation or preset resource rule fusion can be used to determine the optimal computing power scheduling plan based on the first computing power scheduling plan and the second computing power scheduling plan.

[0107] During the dynamic adjustment phase, while executing the optimal target scheduling plan, the current network status and task execution status of the computing power network are monitored in real time. The parameters of the quantum approximate optimization algorithm and the mixed integer linear programming algorithm are dynamically adjusted according to the actual operation data, and even the strategy weights can be adjusted to form a closed-loop optimization and continuously optimize the scheduling effect.

[0108] It should be noted that, during the simulation process, the method for determining the computing power scheduling scheme provided in the embodiment of the present application is used to significantly shorten the decision-making time for computing power scheduling, improve resource utilization, reduce computing power costs, and adapt to complex and changeable computing power demand environments. The specific indicators are as follows: Scheduling efficiency: Compared with using only the traditional computing power scheduling method, the method for determining the computing power scheduling solution provided in the embodiment of the present application reduces the average scheduling decision time by about 30%, showing higher execution efficiency.

[0109] Resource utilization: The average resource utilization of the server using the method for determining the computing power scheduling solution provided in the embodiment of the present application is increased by about 15%, and the existing computing power resources can be utilized more effectively.

[0110] Task completion rate: Within the same time window, the method for determining the computing power scheduling solution provided in the embodiment of the present application increased the task completion rate by approximately 10%, demonstrating a stronger task processing capability.

[0111] In the embodiment of the present application, the predicted computing power requirements corresponding to different computing tasks are determined according to the state parameters of the computing power system; the quantum approximate optimization algorithm is used to determine the solution space corresponding to the target computing power scheduling problem according to the state parameters and the predicted computing power requirements, and the solution of the state parameters and the predicted computing power requirements is extracted to obtain the first computing power scheduling plan; the mixed integer linear programming algorithm is used to refine and optimize the solution space to determine the second computing power scheduling plan; based on the first computing power scheduling plan and the second computing power scheduling plan, the optimal computing power scheduling plan corresponding to the computing power system is determined. This method uses a hybrid framework of the quantum approximate optimization algorithm and the mixed integer linear programming algorithm, aiming to integrate the search acceleration capability of the quantum approximate optimization algorithm and the stable and reliable characteristics of the mixed integer linear programming algorithm, improve the scheduling efficiency and resource utilization of the computing power system, and quickly and effectively determine the optimal computing power scheduling plan.

[0112] The following is a description of a device for determining a computing power scheduling scheme provided in an embodiment of the present application. The device for determining a computing power scheduling scheme described below and the method for determining a computing power scheduling scheme described above can refer to each other.

[0113] Figure 5 Schematic diagram of the structure of the device for determining the computing power scheduling solution provided in the embodiment of the present application. Figure 5 , an embodiment of the present application provides a device for determining a computing power scheduling scheme, including the following data processing module 501 and resource allocation module 502.

[0114] The data processing module 501 is used to determine the predicted computing power requirements corresponding to different computing tasks according to the state parameters of the computing power system; The resource allocation module 502 is used to use a quantum approximate optimization algorithm to determine the solution space corresponding to the target computing power scheduling problem according to the state parameters and the predicted computing power demand, and extract the solution of the state parameters and the predicted computing power demand to obtain a first computing power scheduling plan; use a mixed integer linear programming algorithm to refine and optimize the solution space to determine a second computing power scheduling plan; and determine the optimal computing power scheduling plan corresponding to the computing power system according to the first computing power scheduling plan and the second computing power scheduling plan.

[0115] Optionally, the resource allocation module 502 is specifically used to perform a two-dimensional unconstrained optimization encoding on the computing power scheduling problem to obtain the target computing power scheduling problem; and use the quantum approximate optimization algorithm to determine the solution space corresponding to the target computing power scheduling problem according to the state parameters and the predicted computing power demand.

[0116] Optionally, the data processing module 501 is specifically used to determine the state parameter based on the performance indicators, task load, remaining computing power resources and historical scheduling records of the computing power system; input the state parameter into the long short-term memory network model to obtain the predicted computing power demand output by the long short-term memory network model; wherein the long short-term memory network model is trained based on the historical state data and historical actual computing power demand of the computing power system.

[0117] Optionally, the resource allocation module 502 is specifically used to determine a first weight corresponding to the first computing power scheduling scheme and a second weight corresponding to the second computing power scheduling scheme; determine the optimal computing power scheduling scheme based on the first computing power scheduling scheme, the second computing power scheduling scheme, the first weight and the second weight; or, adopt a preset resource rule fusion method to fuse the first computing power scheduling scheme and the second computing power scheduling scheme to obtain the optimal computing power scheduling scheme.

[0118] Optionally, the data processing module 501 is also used to monitor the current network status and task execution status of the computing power system in real time; and adjust the parameters of the quantum approximate optimization algorithm and the parameters of the mixed integer linear programming algorithm according to the current network status and the task execution status.

[0119] Optionally, the long short-term memory network model is trained based on the following steps: obtaining the historical state data and the historical actual computing power requirements; inputting the historical state data into the initial long short-term memory network model to obtain the historical predicted computing power requirements output by the initial long short-term memory network model; updating the model parameters of the initial long short-term memory network model according to the historical predicted computing power requirements and the historical actual computing power requirements to obtain a trained long short-term memory network model.

[0120] The electronic device involved in the embodiments of the present application may be a device that provides voice and / or data connectivity to a user, a handheld device with a wireless connection function, or other processing devices connected to a wireless modem, etc. In different systems, the names of electronic devices may also be different. For example, in a 5G system, an electronic device may be called a user equipment (UE).

[0121] Figure 6 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application, referring to Figure 6 An embodiment of the present application further provides an electronic device, which may include: a memory 610, a transceiver 620 and a processor 630.

[0122] The memory 610 is used to store computer programs; the transceiver 620 is used to send and receive data under the control of the processor 630; the processor 630 is used to read the computer program in the memory 610 and perform the following operations.

[0123] According to the state parameters of the computing power system, the predicted computing power requirements corresponding to different computing tasks are determined; using a quantum approximate optimization algorithm, according to the state parameters and the predicted computing power requirements, the solution space corresponding to the target computing power scheduling problem is determined, and the solution of the state parameters and the predicted computing power requirements is extracted to obtain a first computing power scheduling plan; using a mixed integer linear programming algorithm, the solution space is refined and optimized to determine a second computing power scheduling plan; according to the first computing power scheduling plan and the second computing power scheduling plan, the optimal computing power scheduling plan corresponding to the computing power system is determined.

[0124] Among them, Figure 6 In the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically one or more processors represented by processor 630 and various circuits of memory represented by memory 610 are linked together. The bus architecture may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface. The transceiver 620 may be a plurality of components, namely, a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium. For different user devices, the user interface 640 may also be an interface capable of externally connecting to the required internal devices.

[0125] The processor 630 is responsible for managing the bus architecture and general processing, and the memory 610 can store data used by the processor 630 when performing operations.

[0126] The processor 630 is used to execute any of the methods provided in the embodiments of the present application according to the obtained executable instructions by calling the computer program stored in the memory 610. The processor and the memory may also be arranged physically separately.

[0127] It should be noted here that the electronic device provided in the embodiment of the present application can implement all the method steps implemented in the above-mentioned method embodiment and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as those in the method embodiment will not be described in detail here.

[0128] Figure 7 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application, such as Figure 7As shown, the electronic device may include: a processor 710, a communication interface 720, a memory 730 and a communication bus 740, wherein the processor 710, the communication interface 720 and the memory 730 communicate with each other through the communication bus 740. The processor 710 may call a computer program in the memory 730 to execute the steps of the method for determining a computing power scheduling scheme, including the following steps.

[0129] According to the state parameters of the computing power system, the predicted computing power requirements corresponding to different computing tasks are determined; using a quantum approximate optimization algorithm, according to the state parameters and the predicted computing power requirements, the solution space corresponding to the target computing power scheduling problem is determined, and the solution of the state parameters and the predicted computing power requirements is extracted to obtain a first computing power scheduling plan; using a mixed integer linear programming algorithm, the solution space is refined and optimized to determine a second computing power scheduling plan; according to the first computing power scheduling plan and the second computing power scheduling plan, the optimal computing power scheduling plan corresponding to the computing power system is determined.

[0130] In addition, the logic instructions in the above-mentioned memory 730 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0131] On the other hand, an embodiment of the present application also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the steps of the method for determining the computing power scheduling scheme provided in the above embodiments, including the following steps.

[0132] According to the state parameters of the computing power system, the predicted computing power requirements corresponding to different computing tasks are determined; using a quantum approximate optimization algorithm, according to the state parameters and the predicted computing power requirements, the solution space corresponding to the target computing power scheduling problem is determined, and the solution of the state parameters and the predicted computing power requirements is extracted to obtain a first computing power scheduling plan; using a mixed integer linear programming algorithm, the solution space is refined and optimized to determine a second computing power scheduling plan; according to the first computing power scheduling plan and the second computing power scheduling plan, the optimal computing power scheduling plan corresponding to the computing power system is determined.

[0133] On the other hand, an embodiment of the present application also provides a processor-readable storage medium, which stores a computer program. The computer program is used to enable the processor to execute the steps of the method for determining the computing power scheduling plan provided in the above embodiments, including the following steps.

[0134] According to the state parameters of the computing power system, the predicted computing power requirements corresponding to different computing tasks are determined; using a quantum approximate optimization algorithm, according to the state parameters and the predicted computing power requirements, the solution space corresponding to the target computing power scheduling problem is determined, and the solution of the state parameters and the predicted computing power requirements is extracted to obtain a first computing power scheduling plan; using a mixed integer linear programming algorithm, the solution space is refined and optimized to determine a second computing power scheduling plan; according to the first computing power scheduling plan and the second computing power scheduling plan, the optimal computing power scheduling plan corresponding to the computing power system is determined.

[0135] The processor-readable storage medium can be any available medium or data storage device that can be accessed by the processor, including but not limited to magnetic storage (such as floppy disks, hard disks, magnetic tapes, magneto-optical disks (MO), etc.), optical storage (such as CD, DVD, BD, HVD, etc.), and semiconductor storage (such as ROM, EPROM, EEPROM, non-volatile memory (NANDFLASH), solid-state drive (SSD)), etc.

[0136] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0137] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0138] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for determining a computing power scheduling scheme, characterized in that: include: Determine the predicted computing power requirements corresponding to different computing tasks based on the state parameters of the computing power system; Using a quantum approximate optimization algorithm, according to the state parameters and the predicted computing power demand, determine a solution space corresponding to the target computing power scheduling problem, and extract solutions for the state parameters and the predicted computing power demand to obtain a first computing power scheduling solution; Using a mixed integer linear programming algorithm, the solution space is refined and optimized to determine a second computing power scheduling solution; According to the first computing power scheduling scheme and the second computing power scheduling scheme, an optimal computing power scheduling scheme corresponding to the computing power system is determined.

2. The method for determining a computing power scheduling scheme according to claim 1, characterized in that: The quantum approximate optimization algorithm is used to determine the solution space corresponding to the target computing power scheduling problem according to the state parameters and the predicted computing power demand, including: Performing a two-dimensional unconstrained optimization encoding on the computing power scheduling problem to obtain the target computing power scheduling problem; The quantum approximate optimization algorithm is adopted to determine the solution space corresponding to the target computing power scheduling problem according to the state parameters and the predicted computing power demand.

3. The method for determining a computing power scheduling scheme according to claim 1, characterized in that: Determining predicted computing power requirements corresponding to different computing tasks according to the state parameters of the computing power system includes: Determining the state parameters according to the performance indicators, task load, remaining computing resources and historical scheduling records of the computing power system; Inputting the state parameters into a long short-term memory network model to obtain the predicted computing power requirement output by the long short-term memory network model; Among them, the long short-term memory network model is trained based on the historical status data and historical actual computing power requirements of the computing power system.

4. The method for determining a computing power scheduling scheme according to claim 1, characterized in that: The determining, according to the first computing power scheduling scheme and the second computing power scheduling scheme, an optimal computing power scheduling scheme corresponding to the computing power system includes: Determine a first weight corresponding to the first computing power scheduling scheme and a second weight corresponding to the second computing power scheduling scheme; determine the optimal computing power scheduling scheme according to the first computing power scheduling scheme, the second computing power scheduling scheme, the first weight and the second weight; or, The first computing power scheduling scheme and the second computing power scheduling scheme are integrated by using a preset resource rule fusion method to obtain the optimal computing power scheduling scheme.

5. The method for determining a computing power scheduling scheme according to any one of claims 1 to 4, characterized in that: The method further comprises: Real-time monitoring of the current network status and task execution status of the computing power system; According to the current network state and the task execution status, the parameters of the quantum approximate optimization algorithm and the parameters of the mixed integer linear programming algorithm are adjusted.

6. The method for determining a computing power scheduling scheme according to claim 3, characterized in that: The long short-term memory network model is trained based on the following steps: Obtaining the historical status data and the historical actual computing power requirements; Inputting the historical state data into an initial long short-term memory network model to obtain the historical predicted computing power demand output by the initial long short-term memory network model; According to the historical predicted computing power requirements and the historical actual computing power requirements, the model parameters of the initial long short-term memory network model are updated to obtain a trained long short-term memory network model.

7. A device for determining a computing power scheduling scheme, characterized in that: include: The data processing module is used to determine the predicted computing power requirements corresponding to different computing tasks based on the state parameters of the computing power system; The resource allocation module is used to use a quantum approximate optimization algorithm to determine the solution space corresponding to the target computing power scheduling problem according to the state parameters and the predicted computing power demand, and extract the solution of the state parameters and the predicted computing power demand to obtain a first computing power scheduling plan; use a mixed integer linear programming algorithm to refine and optimize the solution space to determine a second computing power scheduling plan; and determine the optimal computing power scheduling plan corresponding to the computing power system according to the first computing power scheduling plan and the second computing power scheduling plan.

8. An electronic device comprising a processor and a memory storing a computer program, characterized in that: When the processor executes the computer program, the steps of the method for determining the computing power scheduling plan described in any one of claims 1 to 6 are implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for determining a computing power scheduling solution as described in any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method for determining a computing power scheduling solution as described in any one of claims 1 to 6 are implemented.

Citation Information

Patent Citations

  • Method for determining primitive element of cryptographic hash function, computer program and data processing system

    CN114547638A

  • Intelligent optimization method and system of computing power scheduling for improving power supply reliability

    CN117453398A

  • Power distribution microgrid edge computing power cooperative scheduling method, system, equipment and medium

    CN118228989A

Cited By

  • AI-based computing power resource intelligent scheduling system and method

    CN119576590A