Distributed heterogeneous task scheduling method and system based on evolutionary algorithm

Abstract

The invention discloses a distributed heterogeneous task scheduling method and system based on an evolutionary algorithm. The method comprises the following steps that influence factors of distributed heterogeneous task scheduling are analyzed and quantified; taking the heterogeneous chip, the server and the quantized influence factors of the distributed heterogeneous task scheduling as an optimized objective function, and establishing an overall power consumption planning model; performing scheduling optimization on the distributed heterogeneous tasks by using an improved evolutionary optimization algorithm and taking the overall power consumption planning model as an evaluation basis to obtain an optimal scheduling scheme; through the improved evolutionary algorithm, the overall power consumption planning model and dynamic floating power consumption calculation are combined, the optimal solution can be quickly converged in a large-scale data scene, and the scheduling efficiency and stability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of distributed heterogeneous task scheduling, and in particular to a distributed heterogeneous task scheduling method and system based on an evolutionary algorithm. Background Art

[0002] The Internet of Things uses sensing technology and network communication technology as the main means to achieve ubiquitous connection between people, machines and objects, and provide infrastructure for information perception, information transmission, information processing and other services. With the extensive, stable and efficient data communication brought by the Internet of Things, various Internet of Things applications have penetrated into many fields such as industrial manufacturing, urban management, smart homes, autonomous driving, and environmental monitoring. At present, the Internet of Things is gradually entering a rapid growth stage covering multiple features such as multi-dimensional perception, digital twins, and intelligent interconnection from the initial stage of isolated applications. However, with the explosive growth of Internet of Things devices, how to effectively collect, process, analyze and utilize these distributed heterogeneous massive data with diverse sources and different formats has become a key bottleneck restricting the further development of Internet of Things technology.

[0003] In a distributed heterogeneous environment, the allocation of computing resources needs to consider factors such as task priority, device performance, and energy consumption. However, there is currently a lack of effective energy consumption estimation models and scheduling algorithms, making it difficult to strike a balance between performance and energy consumption; traditional scheduling algorithm design is based on a static power consumption model, which assumes that the device power consumption characteristics remain constant during the task execution cycle. This assumption does not match the actual situation, especially in mobile edge computing scenarios, where devices will experience real-time power consumption fluctuations as task loads, network status, and ambient temperature change. In addition, heterogeneous device integration brings about a dynamic parameter coupling effect. When a new device type is added, its unique power consumption curve will produce nonlinear superposition, resulting in deviations in system-level energy efficiency predictions.

[0004] In the field of computer science, evolutionary computing technology, as a branch of artificial intelligence, has the characteristics of global search capabilities and adaptability to complex optimization problems. It does not require too much auxiliary information and is not limited by the conditions of the search space. It can provide an effective solution to the distributed heterogeneous task scheduling problem. However, there is currently no result estimation model designed for the performance and energy consumption of heterogeneous computing platforms, and the fitness functions of most current heterogeneous scheduling algorithms generally use static energy consumption coefficients. This design ignores the relationship between the real-time status of the device and environmental factors. Summary of the invention

[0005] In response to the above problems, the present invention provides a distributed heterogeneous task scheduling method and system based on evolutionary algorithm, which aims to improve scheduling efficiency by optimizing the algorithm, balance performance and energy consumption, reduce the risk of privacy leakage, and enhance the security of the system.

[0006] To achieve the above object, the present invention provides the following technical solution: a distributed heterogeneous task scheduling method based on evolutionary algorithm, comprising the following steps: Step S1: Analyze and quantify the influencing factors of distributed heterogeneous task scheduling; Step S2: Taking the influencing factors of heterogeneous chips, servers and quantified distributed heterogeneous task scheduling as the optimization objective function, an overall power consumption planning model is established; Step S3: using the optimization algorithm and taking the overall power consumption planning model as the evaluation basis, the distributed heterogeneous tasks are optimized for scheduling to obtain the optimal scheduling solution; The overall power consumption planning model is expressed as: ; In the formula, Indicates the overall power consumption of the server; Represents floating power consumption, Indicates the basic power consumption of heterogeneous chips. Indicates the standby power consumption of the memory device. Indicates the standby power consumption of the storage device. represents fixed overhead; Indicates the system energy efficiency.

[0007] Furthermore, the optimization algorithm adopts an improved evolutionary algorithm. By improving the evolutionary optimization algorithm and taking the overall power consumption planning model as the evaluation basis, the distributed heterogeneous tasks are scheduled and optimized. The specific process of obtaining the optimal scheduling solution is as follows: Step S3.1: Initialize the population of the improved evolutionary optimization algorithm; Step S3.2: Calculate the fitness value of each individual according to the overall power consumption planning model; Step S3.3: Screening individuals in the current population through a selection strategy, where the selection strategy is a combination of roulette wheel selection and random competition selection; Step S3.4: updating individual positions through a perceived pulse evolution strategy, wherein the perceived pulse evolution strategy includes an energy efficiency driven crossover strategy and a power consumption sensitive quantized mutation; Step S3.5: Optimize the current best individual using the local search optimization mechanism; Step S3.6: Determine whether the current iteration meets the termination condition. If not, continue the iteration. If it does, stop the iteration and output the best individual in the current population as the optimal task scheduling solution.

[0008] Furthermore, floating power consumption It is expressed as: ; In the formula, Indicates The consumption caused by the completion of distributed heterogeneous tasks; Indicates the floating value of server power consumption; Indicates The running time required for distributed heterogeneous tasks; Indicates The number of heterogeneous chips required for distributed heterogeneous tasks, Indicates the number of distributed heterogeneous tasks; Basic power consumption of heterogeneous chips It is expressed as: ; In the formula, Indicates the basic power consumption of heterogeneous chips; Indicates the operating voltage of heterogeneous chips; Indicates the working current in standby state; Indicates the maximum running time required to complete all distributed heterogeneous tasks; Memory device standby power consumption It is expressed as: ; In the formula, Indicates the standby power consumption of the memory device. Indicates the operating voltage of the memory device. Indicates the operating current of a single memory device in standby mode. Indicates the number of memory devices; The standby power consumption of storage devices is expressed as: ; In the formula, Indicates the standby power consumption of the storage device. Indicates the operating voltage of the storage device. Indicates the operating current of a single storage device in standby state. Indicates the number of storage devices.

[0009] Further, the population of the evolutionary optimization algorithm is improved It is expressed as: ; In the formula, Indicates the population The location of an individual, , Indicates The servers to which the distributed heterogeneous tasks are assigned. , Indicates the number of distributed heterogeneous tasks; Represents the total number of individuals in the population, each of which represents a possible distributed heterogeneous task scheduling scheme; According to the overall power consumption planning model, the fitness value of each individual is calculated, which is expressed as: ; In the formula, Indicates The fitness value of each individual; express The floating power consumption generated; Indicates The running time of a distributed heterogeneous task.

[0010] Furthermore, the specific process of step S3.3 is: calculating the probability of each individual in the current population being selected and the cumulative probability of each individual being selected, expressed as: ; ; In the formula, Indicates The probability of an individual being selected; Indicates The cumulative probability of an individual being selected is used to determine the individual's position in the roulette wheel selection; Indicates the population individual, that is, A distributed heterogeneous task scheduling solution; Indicates The probability of an individual being selected is the ratio of the individual's fitness value to the total fitness value of the population; Represents the sum of all individual fitness values ​​in the population, used to normalize the probability; Generate probabilistic random numbers , choose to meet Individual , Indicates The cumulative probability of an individual being selected; From satisfying Individual Randomly select two individuals and compare their fitness values, and select the individual with a lower fitness value.

[0011] Furthermore, the specific process of step S3.4 is as follows: By calculating the standard deviation of the population To measure the diversity of the population and dynamically adjust the crossover probability , expressed as: ; ; In the formula, in the formula, represents the average fitness value of the population, The threshold of population diversity is used to determine whether the population is too concentrated or dispersed. represents the temperature parameter, which is used to control the changing speed of the crossover probability; represents the base of natural logarithms; After calculating the crossover probability, the individuals selected by the selection strategy are compared pairwise in the order of the individuals in the population; for each pair of individuals, a first random number between 0 and 1 is generated , the second random number and the third random number ,when Less than , then the pair of individuals is cross-operated, and at the same time, when and satisfy , is the individual coding length, then the pair of individuals arrive The codes between them are cross-interchanged to complete individual crossover; For the mutation operation, power-sensitive quantized mutation is used, and the mutation intensity calculation is expressed as: ; In the formula, Indicates The variable step length of each individual; Indicates The total power consumption of each unit, including operating power consumption and standby power consumption; , They are the upper and lower bounds of the population power consumption, i.e., the maximum and minimum power consumption of all individuals in the current population; Indicates the maximum mutation step length, that is, the maximum adjustment range of the mutation operation; At the same time, when the best individual in the current population does not change in several consecutive iterations, tunnel mutation is triggered, and the tunnel mutation probability is The calculation method is expressed as: ; ; In the formula, It represents potential energy, which is the ratio of the current individual power consumption to energy efficiency; represents kinetic energy, indicating the dynamic rate of change of the population; represents the scaling factor, which is used to adjust the magnitude of kinetic energy; Indicates the current iteration number; Indicated in The standard deviation of the population fitness at the iteration; Generate the fourth random number ,when Less than , then the tunneling variation is triggered, and then according to the variable step length , execute mutation, change gene value, and complete individual mutation.

[0012] Furthermore, the specific process of optimizing the current optimal individual using the local search optimization mechanism is as follows: In each iteration, the individual with the lowest fitness in the current population is selected Conduct local search; local search optimizes the fitness value of the individual by adjusting task allocation; For the individual with the lowest fitness in the current population , randomly select a distributed heterogeneous task and a server , distributed heterogeneous tasks Redistribute to server , and calculate the redistribution The fitness value of , expressed as: ; ; In the formula, Represents the reallocated ; Indicates Distributed heterogeneous tasks in The server assigned in ; when ,take over Redistribution plan; otherwise, Restore to the original state; repeat the above operation Second-rate; express The fitness value of .

[0013] A distributed heterogeneous task scheduling system based on evolutionary algorithm, comprising: Analysis module, used to analyze and quantify the influencing factors of distributed heterogeneous task scheduling; The model building module is used to take the influencing factors of heterogeneous chips, servers and quantified distributed heterogeneous task scheduling as the optimization objective function to establish an overall power consumption planning model; The solution module is used to utilize the improved evolutionary optimization algorithm and the overall power consumption planning model as the evaluation basis to optimize the scheduling of distributed heterogeneous tasks and obtain the optimal scheduling solution.

[0014] Compared with the existing technology, the present invention has the following beneficial effects:

[0015] (1) The present invention uses an improved evolutionary algorithm, combined with an overall power consumption planning model and dynamic floating power consumption calculation, to quickly converge to the optimal solution in large-scale data scenarios, thereby improving scheduling efficiency and stability. The present invention not only considers the task completion time, but also takes into account the overall power consumption of the server, including standby power consumption and operating power consumption. By dynamically adjusting the floating power consumption interval and optimizing the task allocation strategy, the balance between performance and energy consumption is achieved. At the same time, in view of the complexity of heterogeneous computing platforms, the present invention constructs a flexible and reliable energy consumption estimation model that can adapt to the addition of different heterogeneous devices and provide accurate energy consumption evaluation for task scheduling.

[0016] (2) The present invention can comprehensively evaluate the impact of task scheduling on energy consumption by incorporating the overall power consumption of the server, including standby power consumption and operating power consumption, into the optimization target. It provides a comprehensive energy consumption evaluation framework for task scheduling, which can dynamically adjust power consumption allocation to adapt to the characteristics of different tasks and devices; by introducing dynamic floating power consumption intervals based on the power consumption changes of heterogeneous chips during operation, it can more accurately reflect the energy consumption fluctuations during task execution and avoid energy consumption estimation deviations caused by data read and write conflicts or queuing. It improves the flexibility and accuracy of energy consumption estimation, can adapt to the characteristics of different heterogeneous devices, and provide a more reliable basis for optimized scheduling.

[0017] (3) The present invention significantly improves the performance and efficiency of the distributed heterogeneous task scheduling algorithm by introducing adaptive crossover and mutation probability and local search optimization. The adaptive crossover and mutation probability mechanism dynamically adjusts the operation intensity according to the population diversity, avoiding the algorithm from converging prematurely or falling into the local optimum due to fixed probability, thereby enhancing the global search capability and convergence speed. The local search optimization further optimizes the task allocation of the current optimal individual, and further reduces energy consumption and improves the quality of the solution by fine-tuning the exploration of the neighborhood solution space. This enables the algorithm to efficiently find the global optimal solution in large-scale distributed heterogeneous task scheduling, while taking into account energy consumption optimization and task completion time, thereby improving the overall performance and adaptability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0019] like Figure 1 As shown, the present invention provides a technical solution: a distributed heterogeneous task scheduling method based on an evolutionary algorithm, comprising the following steps:

[0020] Step S1: Analyze and quantify the influencing factors of distributed heterogeneous task scheduling.

[0021] Among them, the influencing factors of distributed heterogeneous task scheduling include the overall standby power consumption of the server and the increased power consumption during operation; the overall standby power consumption of the server is divided into the power consumption brought by the hardware equipment and the standby power consumption of the heterogeneous chips. The increased power consumption brought by the server operation includes the increased power consumption brought by the heterogeneous chip calculations and the increased power consumption brought by adding containers.

[0022] The quantification process includes:

[0023] Data collection: collect relevant data by consulting official statistics, research reports, energy consumption records, etc.

[0024] Data standardization: standardize the collected relevant data to facilitate comparison of data from different times and places.

[0025] Data transformation, converting the standardized data into numerical values ​​that can be used for subsequent model calculations, such as scaling the data to a specific range through normalization.

[0026] Data validation, through cross-validation or other methods, ensures the accuracy and reliability of the converted data.

[0027] Step S2: Taking the influencing factors of heterogeneous chips, servers and quantified distributed heterogeneous task scheduling as the optimization objective function, an overall power consumption planning model is established.

[0028] The floating power consumption is calculated for each distributed heterogeneous task. The floating power consumption includes the power consumption caused by heterogeneous chip operation, data conflict, queuing, etc.

[0029] When heterogeneous chips start running, the frequency increase brought by the calculation of heterogeneous chips will cause their own power consumption to increase. The consumption caused by this part of the increase is also very different depending on the heterogeneous chip. The performance consumption generated is not linearly increasing. Due to data read and write conflicts, computer system queuing mechanism and other factors, this part of the data needs a reasonable floating range to represent it. This floating range can be dynamically adjusted according to the number of heterogeneous chips used to ensure that the floating value generated is within a reasonable range; floating power consumption It is expressed as: ; In the formula, Indicates The consumption caused by the completion of distributed heterogeneous tasks; Indicates the floating value of server power consumption; Indicates The running time required for distributed heterogeneous tasks; Indicates The number of heterogeneous chips required for distributed heterogeneous tasks, Indicates the number of distributed heterogeneous tasks.

[0030] At the same time, the power consumption generated by the server in standby mode when it is turned on is also considered. Under certain conditions, as long as there is no large-scale change to the server, this part of the power consumption is basically in a stable range with very small fluctuations. Therefore, in the calculation process, this part of the fluctuation is unified in the power consumption model when the heterogeneous chip is running and integrated into it; the power consumption generated by the server in standby mode when it is turned on includes the basic power consumption of the heterogeneous chip and the energy consumption of peripheral components.

[0031] The basic power consumption of heterogeneous chips refers to the power consumption of each heterogeneous chip in the server in standby mode, which is expressed as: ; In the formula, Indicates the basic power consumption of heterogeneous chips; Indicates the operating voltage of heterogeneous chips; Indicates the working current in standby state; Indicates the maximum running time required to complete all distributed heterogeneous tasks.

[0032] Peripheral component energy consumption includes memory device standby power consumption, storage device standby power consumption, and fixed overhead; memory device standby power consumption is expressed as: ; In the formula, Indicates the standby power consumption of the memory device. Indicates the operating voltage of the memory device. Indicates the operating current of a single memory device in standby mode. Indicates the number of memory devices.

[0033] The standby power consumption of storage devices is expressed as: ; In the formula, Indicates the standby power consumption of the storage device. Indicates the operating voltage of the storage device. Indicates the operating current of a single storage device in standby state. Indicates the number of storage devices.

[0034] Considering the overall energy efficiency of the server, the overall power consumption of the server is obtained by integrating the floating power consumption, the basic power consumption of heterogeneous chips and the energy consumption of peripheral components. : ; In the formula, Represents floating power consumption, Indicates the basic power consumption of heterogeneous chips. Indicates the standby power consumption of the memory device. Indicates the standby power consumption of the storage device. represents fixed overhead, which is the fixed energy consumption of the motherboard, low-speed fan, etc.; Indicates the system energy efficiency.

[0035] Step S3: Using the improved evolutionary optimization algorithm and taking the overall power consumption planning model as the evaluation basis, the distributed heterogeneous tasks are scheduled and optimized to obtain the optimal scheduling solution.

[0036] Step S3.1: Initialize the population of the improved evolutionary optimization algorithm.

[0037] Random Generation individuals, each of which represents a possible distributed heterogeneous task scheduling solution.

[0038] Improving the population of evolutionary optimization algorithms It is expressed as: ; In the formula, Indicates the population The location of an individual, , Indicates The servers to which the distributed heterogeneous tasks are assigned. , Indicates the number of distributed heterogeneous tasks.

[0039] Step S3.2: The fitness value is used to evaluate the quality of each individual. The lower the fitness, the lower the energy consumption of the individual, and the higher the fitness, the greater the energy consumption of the individual. According to the overall power consumption planning model, the fitness value of each individual is calculated, expressed as: ; In the formula, Indicates The fitness value of each individual; express The floating power consumption generated; Indicates The running time of a distributed heterogeneous task.

[0040] Step S3.3: Screen individuals in the current population through a selection strategy, where the selection strategy is a combination of roulette wheel selection and random competition selection.

[0041] The fitness ratio is proportional to the probability of an individual being selected; the probability of each individual being selected in the current population and the cumulative probability of each individual being selected are calculated, expressed as: ; ; In the formula, Indicates The probability of an individual being selected; Indicates The cumulative probability of an individual being selected is used to determine the individual's position in the roulette wheel selection; Indicates the population individual, that is, A distributed heterogeneous task scheduling solution; Indicates The probability of an individual being selected is the ratio of the individual's fitness value to the total fitness value of the population; Represents the sum of all individual fitness values ​​in the population and is used to normalize the probability.

[0042] Generate probabilistic random numbers , choose to meet Individual , Indicates The cumulative probability of an individual being selected; From satisfying Individual Randomly select two individuals and compare their fitness values, and select the individual with lower fitness value to enter the next iteration.

[0043] Step S3.4: Update individual positions through the perceived pulse evolution strategy, which includes an energy-efficiency-driven crossover strategy and a power-sensitive quantized mutation.

[0044] Crossover and mutation are important operators in evolutionary algorithms. Their main function is to ensure that the algorithm results do not converge too quickly and fail to reach the global optimal solution. In traditional evolutionary algorithms, the crossover and mutation probabilities are usually fixed, which limits the optimization ability of the algorithm at different stages. In order to overcome this problem, the present invention introduces a perceptual pulse evolution strategy, which enables the evolutionary algorithm to dynamically adjust the operation intensity according to the diversity of the population.

[0045] By calculating the standard deviation of the population To measure the diversity of the population and dynamically adjust the crossover probability The larger the standard deviation, the higher the population diversity. In this case, the crossover probability is increased to facilitate global search. On the contrary, the crossover probability is reduced to avoid unnecessary computational overhead. ; ; In the formula, represents the average fitness value of the population, The threshold of population diversity is used to determine whether the population is too concentrated or dispersed. represents the temperature parameter, which is used to control the changing speed of the crossover probability; Represents the base of natural logarithms.

[0046] After calculating the crossover probability, the individuals selected by the selection strategy are compared pairwise in the order of the individuals in the population; for each pair of individuals, a first random number between 0 and 1 is generated , the second random number and the third random number ,like Less than , then the pair of individuals is cross-operated. At the same time, if and satisfy ( is the individual coding length), then the pair of individuals in arrive The codes between them are cross-exchanged to complete individual crossover.

[0047] For mutation operations, power-sensitive quantized mutation is used. The higher the power consumption of the individual, the greater the mutation intensity. It is more inclined to perform larger mutation operations on high-power individuals to guide the population to evolve in the direction of low power consumption. The mutation intensity is calculated as follows: ; In the formula, Indicates The variable step length of each individual; Indicates The total power consumption of each unit, including operating power consumption and standby power consumption; , They are the upper and lower bounds of the population power consumption, i.e., the maximum and minimum power consumption of all individuals in the current population; Indicates the maximum mutation step length, that is, the maximum adjustment range of the mutation operation.

[0048] If the best individual in the current population does not change in several consecutive iterations, tunneling mutation is triggered. The tunneling mutation probability is The calculation method is expressed as: ; ; In the formula, It represents potential energy, which is the ratio of the current individual power consumption to energy efficiency; represents kinetic energy, indicating the dynamic rate of change of the population; represents the scaling factor, which is used to adjust the magnitude of kinetic energy; Indicates the current iteration number; Indicated in The standard deviation of the population fitness at iteration .

[0049] Generate the fourth random number ,like Less than , then the tunneling variation is triggered, and then according to the variable step length , execute mutation, change gene value, and complete individual mutation.

[0050] Step S3.5: In order to further improve the optimization performance of the algorithm and avoid falling into the local optimum, the local search optimization mechanism is used to optimize the current optimal individual. The purpose of local search optimization is to fine-tune the current optimal individual to explore its neighborhood solution space and find a better solution. This enhances the algorithm's search ability in the local solution space and can effectively improve the quality of the solution.

[0051] In each iteration, the individual with the lowest fitness in the current population is selected Perform local search. Local search optimizes the fitness value of the individual by adjusting the distributed heterogeneous task allocation.

[0052] For the individual with the lowest fitness in the current population , randomly select a distributed heterogeneous task and a server , distributed heterogeneous tasks Redistribute to server , and calculate the redistribution The fitness value of ;when ,take over Redistribution plan; otherwise, Restore to the original state; repeat the above operation Second-rate; express The fitness value of . ; ; In the formula, Represents the reallocated ; Indicates Distributed heterogeneous tasks in The server assigned in .

[0053] Step S3.6: Determine whether the current iteration meets the termination condition (preset maximum number of iterations). If not, continue the iteration; if it meets, stop the iteration and output the best individual in the current population (the individual with the lowest fitness). ) as the optimal distributed heterogeneous task scheduling solution.

[0054] A distributed heterogeneous task scheduling system based on evolutionary algorithm, comprising:

[0055] The analysis module is used to analyze and quantify the influencing factors of distributed heterogeneous task scheduling.

[0056] The model building module is used to take the influencing factors of heterogeneous chips, servers and quantified distributed heterogeneous task scheduling as the optimization objective function to establish an overall power consumption planning model.

[0057] The solution module is used to utilize the improved evolutionary optimization algorithm and the overall power consumption planning model as the evaluation basis to optimize the scheduling of distributed heterogeneous tasks and obtain the optimal scheduling solution.

[0058] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A distributed heterogeneous task scheduling method based on evolutionary algorithm, characterized in that: The steps include: Step S1: Analyze and quantify the influencing factors of distributed heterogeneous task scheduling; Step S2: Taking the influencing factors of heterogeneous chips, servers and quantified distributed heterogeneous task scheduling as the optimization objective function, an overall power consumption planning model is established; Step S3: using the optimization algorithm and taking the overall power consumption planning model as the evaluation basis, the distributed heterogeneous tasks are optimized for scheduling to obtain the optimal scheduling solution; The overall power consumption planning model is expressed as: ； In the formula, Indicates the overall power consumption of the server; Represents floating power consumption, Indicates the basic power consumption of heterogeneous chips. Indicates the standby power consumption of the memory device. Indicates the standby power consumption of the storage device. represents fixed overhead; Indicates the system energy efficiency.

2. The distributed heterogeneous task scheduling method based on evolutionary algorithm according to claim 1 is characterized by: The optimization algorithm adopts an improved evolutionary algorithm. By improving the evolutionary optimization algorithm and taking the overall power consumption planning model as the evaluation basis, the distributed heterogeneous tasks are scheduled and optimized. The specific process of obtaining the optimal scheduling solution is as follows: Step S3.1: Initialize the population of the improved evolutionary optimization algorithm; Step S3.2: Calculate the fitness value of each individual according to the overall power consumption planning model; Step S3.3: Screening individuals in the current population through a selection strategy, where the selection strategy is a combination of roulette wheel selection and random competition selection; Step S3.4: updating individual positions through a perceived pulse evolution strategy, wherein the perceived pulse evolution strategy includes an energy efficiency driven crossover strategy and a power consumption sensitive quantized mutation; Step S3.5: Optimize the current best individual using the local search optimization mechanism; Step S3.6: Determine whether the current iteration meets the termination condition. If not, continue the iteration. If it meets the condition, stop the iteration and output the best individual in the current population as the optimal task scheduling solution.

3. The distributed heterogeneous task scheduling method based on evolutionary algorithm according to claim 2 is characterized by: Floating power consumption It is expressed as: ； In the formula, Indicates The consumption caused by the completion of distributed heterogeneous tasks; Indicates the floating value of server power consumption; Indicates The running time required for distributed heterogeneous tasks; Indicates The number of heterogeneous chips required for distributed heterogeneous tasks, Indicates the number of distributed heterogeneous tasks; Basic power consumption of heterogeneous chips It is expressed as: ； In the formula, Indicates the basic power consumption of heterogeneous chips; Indicates the operating voltage of heterogeneous chips; Indicates the working current in standby state; Indicates the maximum running time required to complete all distributed heterogeneous tasks; Memory device standby power consumption It is expressed as: ； In the formula, Indicates the standby power consumption of the memory device. Indicates the operating voltage of the memory device. Indicates the operating current of a single memory device in standby mode. Indicates the number of memory devices; The standby power consumption of storage devices is expressed as: ； In the formula, Indicates the standby power consumption of the storage device. Indicates the operating voltage of the storage device. Indicates the operating current of a single storage device in standby state. Indicates the number of storage devices.

4. The distributed heterogeneous task scheduling method based on evolutionary algorithm according to claim 3 is characterized by: Improving the population of evolutionary optimization algorithms It is expressed as: ； In the formula, Indicates the population The location of an individual, , Indicates The servers to which the distributed heterogeneous tasks are assigned. , Indicates the number of distributed heterogeneous tasks; Represents the total number of individuals in the population, each of which represents a possible distributed heterogeneous task scheduling scheme; According to the overall power consumption planning model, the fitness value of each individual is calculated, which is expressed as: ； In the formula, Indicates The fitness value of each individual; express The floating power consumption generated; Indicates The running time of a distributed heterogeneous task.

5. The distributed heterogeneous task scheduling method based on evolutionary algorithm according to claim 4 is characterized by: The specific process of step S3.3 is: calculate the probability of each individual in the current population being selected and the cumulative probability of each individual being selected, expressed as: ； ； In the formula, Indicates The probability of an individual being selected; Indicates The cumulative probability of an individual being selected is used to determine the individual's position in the roulette wheel selection; Indicates the population individual, that is, A distributed heterogeneous task scheduling solution; Indicates The probability of an individual being selected is the ratio of the individual's fitness value to the total fitness value of the population; Represents the sum of all individual fitness values ​​in the population, used to normalize the probability; Generate probabilistic random numbers , choose to meet Individual , Indicates The cumulative probability of an individual being selected; From satisfying Individual Randomly select two individuals and compare their fitness values, and select the individual with a lower fitness value.

6. The distributed heterogeneous task scheduling method based on evolutionary algorithm according to claim 5 is characterized by: The specific process of step S3.4 is: By calculating the standard deviation of the population To measure the diversity of the population and dynamically adjust the crossover probability , expressed as: ； ； In the formula, in the formula, represents the average fitness value of the population, The threshold of population diversity is used to determine whether the population is too concentrated or dispersed. represents the temperature parameter, which is used to control the changing speed of the crossover probability; represents the base of natural logarithms; After calculating the crossover probability, the individuals selected by the selection strategy are compared pairwise in the order of the individuals in the population; for each pair of individuals, a first random number between 0 and 1 is generated , the second random number and the third random number ,when Less than , then the pair of individuals is cross-operated, and at the same time, when and satisfy , is the individual coding length, then the pair of individuals arrive The codes between them are cross-interchanged to complete individual crossover; For the mutation operation, power-sensitive quantized mutation is used, and the mutation intensity calculation is expressed as: ； In the formula, Indicates The variable step length of each individual; Indicates The total power consumption of each unit, including operating power consumption and standby power consumption; , They are the upper and lower bounds of the population power consumption, i.e., the maximum and minimum power consumption of all individuals in the current population; Indicates the maximum mutation step length, that is, the maximum adjustment range of the mutation operation; At the same time, when the best individual in the current population does not change in several consecutive iterations, tunnel mutation is triggered, and the tunnel mutation probability is The calculation method is expressed as: ； ； In the formula, It represents potential energy, which is the ratio of the current individual power consumption to energy efficiency; represents kinetic energy, indicating the dynamic rate of change of the population; represents the scaling factor, which is used to adjust the magnitude of kinetic energy; Indicates the current iteration number; Indicated in The standard deviation of the population fitness at the iteration; Generate the fourth random number ,when Less than , then the tunneling variation is triggered, and then according to the variable step length , execute mutation, change gene value, and complete individual mutation.

7. The distributed heterogeneous task scheduling method based on evolutionary algorithm according to claim 6 is characterized by: The specific process of optimizing the current optimal individual using the local search optimization mechanism is: In each iteration, the individual with the lowest fitness in the current population is selected Conduct local search; local search optimizes the fitness value of the individual by adjusting task allocation; For the individual with the lowest fitness in the current population , randomly select a distributed heterogeneous task and a server , distributed heterogeneous tasks Redistribute to server , and calculate the redistribution The fitness value of , expressed as: ； ； In the formula, Represents the reallocated ; Indicates Distributed heterogeneous tasks in The server assigned in ; when ,take over Redistribution plan; otherwise, Restore to the original state; repeat the above operation Second-rate; express The fitness value of .

8. A distributed heterogeneous task scheduling system based on evolutionary algorithm, applied to a distributed heterogeneous task scheduling method based on evolutionary algorithm as claimed in any one of claims 1 to 7, characterized in that: include: Analysis module, used to analyze and quantify the influencing factors of distributed heterogeneous task scheduling; The model building module is used to take the influencing factors of heterogeneous chips, servers and quantified distributed heterogeneous task scheduling as the optimization objective function to establish an overall power consumption planning model; The solution module is used to utilize the improved evolutionary optimization algorithm and the overall power consumption planning model as the evaluation basis to optimize the scheduling of distributed heterogeneous tasks and obtain the optimal scheduling solution.

Images