Intelligent scheduling method and system for computer cloud resources

Abstract

The invention discloses an intelligent scheduling method and system for computer cloud resources, and belongs to the technical field of computer cloud resource scheduling, and the method comprises the steps: collecting task demand data and state data of cloud nodes, carrying out the preprocessing, calculating priority values, carrying out the clustering of the priority values through employing a K-Means clustering method, generating a clustering result, and carrying out the calculation of the clustering result. Fuzzy reasoning is carried out by applying fuzzy logic in combination with the clustering result, a task allocation scheme set is obtained for analogue simulation, and simulation data are output; based on the simulation data, constructing a decision matrix in an AHP method, calculating an initial weight to perform constraint optimization, obtaining a final weight vector, combining the simulation data and a task allocation scheme set, defining a fitness function to perform task allocation scheme optimization, outputting an optimal task scheduling scheme, and then performing feedback; according to the method, dynamic optimization and real-time adjustment of the computer cloud resource task scheduling scheme are realized, and the computing efficiency is greatly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer cloud resource scheduling, in particular to an intelligent scheduling method and system for computer cloud resources. BACKGROUND

[0002] With the rapid development of computer and cloud computing technology, the scheduling and management of cloud resources have become a key problem for optimizing the utilization of computing resources and improving task processing efficiency. Traditional cloud resource scheduling methods usually rely on static resource allocation strategies, which fail to effectively cope with complex and changing task demands and resource conditions of cloud nodes, especially when facing a large number of task requests and diversified resource constraints, which can easily lead to resource waste and increased delays. In order to address these challenges, in recent years, more and more intelligent scheduling methods have been proposed, among which intelligent scheduling methods based on optimization algorithms are considered an effective way to solve this problem.

[0003] However, existing intelligent scheduling methods for computer cloud resources cannot adjust the weights in time under dynamic changes, thereby affecting the optimization effect of the scheduling results, and when facing large-scale tasks, the computational burden is heavy and the execution efficiency is low. In addition, the scheduling process lacks flexibility and real-time performance. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides an intelligent scheduling method and system for computer cloud resources, which solves the problems of existing intelligent scheduling methods for computer cloud resources, such as inability to adjust weights in time, heavy computational burden, low execution efficiency, and lack of flexibility and real-time performance in the scheduling process.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides an intelligent scheduling method for computer cloud resources, which includes,

[0008] Collecting task demand data and state data of cloud nodes for preprocessing, calculating priority values, and using K-Means clustering method to cluster the priority values to generate clustering results, applying fuzzy logic to combine the clustering results for fuzzy reasoning to obtain a task allocation scheme set for simulation, and outputting simulation data;

[0009] The task demand data includes computing resource demand and timestamp of task generation, and the state data of cloud nodes includes computing capacity, memory capacity, load, and power consumption;

[0010] The simulation data includes simulation values of delay of tasks, resource utilization rate of cloud nodes, and power consumption;

[0011] Based on simulation data, a decision matrix in the AHP method is constructed, initial weights are calculated for constraint optimization, the final weight vector is obtained, the simulation data and the task allocation scheme set are combined, the fitness function is defined for task allocation scheme optimization, the optimal task scheduling scheme is output, and feedback is performed;

[0012] The data is displayed through a visual interface and stored in a database.

[0013] As a preferred scheme of the intelligent scheduling method of computer cloud resources, wherein: based on simulation data, a decision matrix in the AHP method is constructed, initial weights are calculated for constraint optimization, the final weight vector is obtained, the simulation data and the task allocation scheme set are combined, the fitness function is defined for task allocation scheme optimization, each data in the simulation data is used as an evaluation standard, the relative importance value between the task allocation scheme and each evaluation standard is set using an empirical rule, a decision matrix in the AHP method is constructed, the sum of each column in the decision matrix is calculated, and the weighted sum of all task allocation schemes under the evaluation standard is obtained;

[0014] Based on the weighted sum, each relative importance value in the decision matrix is normalized to obtain a normalized matrix, the average value of each row is solved to splice, and the weight vector of the current simulation data is generated;

[0015] The weight vector includes the weight of the current task allocation scheme under each evaluation standard;

[0016] Based on the normalized matrix, the maximum eigenvalue is obtained, and based on the weight vector and the simulation value of the evaluation standard, the final weight vector required by each task allocation scheme is calculated by combining the data envelopment analysis method, and a constraint condition is set;

[0017] The task allocation scheme is used as an individual, a population is generated based on the task allocation scheme set, and the fitness function is defined based on the corresponding simulation data and the final weight vector, and the fitness function is minimized;

[0018] The fitness value is calculated, and the individual is self-adjusted according to the fitness value;

[0019] Based on the self-adjusted individual, the fitness value is recalculated, the gradient of the self-adjusted individual fitness value is calculated using the partial derivative operation, the self-adjusted individual is locally optimized using the update formula of the gradient descent method, and a new individual is generated;

[0020] A random number generator is used to generate random numbers of standard normal distribution to splice a random vector, and a mutation operation is performed on the new individual to obtain a mutated individual;

[0021] The calculation of the fitness value, self-adjustment of the individual, partial derivative operation, local optimization and mutation operation are repeatedly performed, and after the number of iterations reaches the maximum number, the optimal task scheduling scheme is output.

[0022] The optimal task scheduling scheme comprises an execution sequence and an allocation of each task on the cloud node.

[0023] As a preferred scheme of the intelligent scheduling method of computer cloud resources, wherein: the re-feedback is based on the scheduling of the computer cloud resources according to the optimal task scheduling scheme, and after the scheduling, the delay of the task, the resource utilization and the power consumption of the cloud node are recorded in real time and fed back to the fitness function for optimization to obtain a new task scheduling scheme.

[0024] When the fitness value of the new task scheduling scheme is less than the fitness value of the optimal task scheduling scheme, the subsequent scheduling of the computer cloud resources is performed according to the new task scheduling scheme, otherwise the scheduling of the computer cloud resources is continued to be performed according to the optimal task scheduling scheme.

[0025] As a preferred scheme of the intelligent scheduling method of computer cloud resources, wherein: the display through the visual interface refers to generating a comparison chart using a visual tool, and displaying the comparison result of the fitness value of the new task scheduling scheme and the fitness value of the optimal task scheduling scheme through the comparison chart.

[0026] Red indicates that the fitness value of the new task scheduling scheme is greater than or equal to the fitness value of the optimal task scheduling scheme.

[0027] Green indicates that the fitness value of the new task scheduling scheme is less than the fitness value of the optimal task scheduling scheme.

[0028] As a preferred scheme of the intelligent scheduling method of computer cloud resources, wherein: the storage of the data in the database refers to compressing the optimal task scheduling scheme and the comparison chart using a compression algorithm, generating a folder format, storing the folder format in the database, and after storing, adding a time stamp to the folder format and sorting according to the time stamp.

[0029] As a preferred scheme of the intelligent scheduling method of computer cloud resources, wherein: the collection of the task demand data and the state data of the cloud node for preprocessing and calculation of the priority value refers to obtaining the task demand data from the Internet of Things device through an API interface, and recording the state data of each cloud node.

[0030] The task demand data comprises a computing resource demand and a time stamp of task generation.

[0031] The state data of the cloud node comprises a computing capability, a memory capacity, a load and a power consumption.

[0032] All data are denoised, normalized and filled in;

[0033] After using the empirical rule to set the weight of each data in the task demand data, the overall value of the task demand data is calculated by the weighted sum formula, which is defined as the priority value.

[0034] As a preferred scheme of the intelligent scheduling method of computer cloud resources, wherein: after using the K-Means clustering method to cluster the priority values, a clustering result is generated, fuzzy reasoning is applied based on the clustering result, a task allocation scheme set is obtained for simulation, and simulation data is outputted based on the priority values, and the elbow rule is used to set the number of clusters, the priority values are clustered using the K-Means clustering method, and the clustering result is generated, including high, medium and low priority tasks.

[0035] Based on the clustering result, each task in the task demand data and the resource state data are used as input variables of fuzzy logic, the Delphi method is used to set fuzzy rules, and fuzzy reasoning is performed using a fuzzy rule engine to obtain a task allocation scheme set.

[0036] A simulation model is constructed using the simulation tool MATLAB, and according to each task allocation scheme in the task allocation scheme set, the corresponding cloud node state data, tasks and network topology are set, the simulation transmission of the tasks is performed, and the simulation data corresponding to each task allocation scheme is obtained, including the simulation values of the delay of the tasks, the resource utilization rate of the cloud nodes and the power consumption.

[0037] In a second aspect, the present application provides an intelligent scheduling system of computer cloud resources, comprising,

[0038] The acquisition simulation module is used to acquire and preprocess the task demand data and the state data of the cloud nodes, calculate the priority values, and use the K-Means clustering method to cluster the priority values, then apply fuzzy logic for fuzzy reasoning and simulation, and output simulation data.

[0039] The constraint optimization module is used to construct a decision matrix in the AHP method based on the simulation data, calculate the initial weight for constraint optimization, obtain the final weight vector, combine the simulation data with the task allocation scheme set, define the fitness function for task allocation scheme optimization, output the optimal task scheduling scheme, and then perform feedback.

[0040] The storage display module is used to display through a visual interface and store data using a database.

[0041] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the computer program, when executed by the processor, implements any step of the intelligent scheduling method of computer cloud resources according to the first aspect of the present application.

[0042] In a fourth aspect, the present application provides a computer readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements any step of the intelligent scheduling method of computer cloud resources according to the first aspect of the present application.

[0043] The present application has the advantages that: the present application realizes dynamic optimization and real-time adjustment of the computer cloud resource task scheduling scheme by combining simulation data, AHP method, data envelopment analysis method and fitness function optimization, and greatly improves the computing efficiency, and secondly, through the real-time feedback mechanism, the scheduling scheme is continuously optimized to ensure that the task scheduling always maintains the optimal state in the changing environment. BRIEF DESCRIPTION OF DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0045] Fig. 1 The flowchart of the intelligent scheduling method of computer cloud resources in embodiment 1.

[0046] Fig. 2 The structure diagram of the intelligent scheduling system of computer cloud resources in embodiment 1. DETAILED DESCRIPTION

[0047] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail in conjunction with the drawings of the specification.

[0048] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the connotation of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0049] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an independent or alternative embodiment that excludes other embodiments.

[0050] Embodiment 1, reference Figs. 1-2 As a first embodiment of the present application, the embodiment provides an intelligent scheduling method of computer cloud resources, comprising the following steps:

[0051] S1, collecting task demand data and cloud node state data for preprocessing, calculating priority value, and using K-Means clustering method for clustering of priority value, generating clustering result, applying fuzzy logic to combine clustering result for fuzzy reasoning, obtaining task allocation scheme set for simulation, and outputting simulation data;

[0052] Specifically, collecting task demand data and cloud node state data for preprocessing, calculating priority value, that is, obtaining task demand data from Internet of Things devices through API interface, and recording the state data of each cloud node;

[0053] The task demand data includes computing resource demand and task generation timestamp;

[0054] The state data of the cloud node includes computing power, memory capacity, load and power consumption;

[0055] All data are denoised, normalized and filled;

[0056] After using empirical rule to set weight for each data in the task demand data, the overall value of the task demand data is calculated by weighted summation formula, which is defined as priority value.

[0057] The task demand data and cloud node state data are collected from Internet of Things devices through API interface, which can ensure that the scheduling decision is more accurate and dynamic through real-time data acquisition. The preprocessing stage includes denoising, normalization and filling operation. The preprocessing can improve the data quality and make different sources and different types of data comparable under the same standard. The priority value of the task is calculated by weighted summation, which can more scientifically reflect the demand of each task for cloud resources and its urgency. Secondly, the priority value of the task is calculated, so that the present application can dynamically adjust the resource scheduling strategy according to the different demands of the task (such as computing resource, time sensitivity, etc.).

[0058] Further, after using K-Means clustering method for clustering of priority value, clustering result is generated, fuzzy reasoning is carried out by applying fuzzy logic combined with clustering result, task allocation scheme set is obtained for simulation, and simulation data is outputted, that is, based on priority value, elbow rule is used to set the number of clusters, K-Means clustering method is used to cluster priority value, clustering result is generated, including high, medium and low priority tasks;

[0059] Based on the clustering result, each task in the task demand data and the resource state data are taken as input variables of the fuzzy logic, the fuzzy rules are set by using the Delphi method, and the fuzzy rule engine is used for fuzzy reasoning to obtain a task allocation scheme set;

[0060] The allocation scheme refers to the task being allocated to a suitable cloud node;

[0061] The fuzzy rule example content is as follows:

[0062] When the computing resource demand of the task is "high" and the computing ability of the cloud node is "high", the task is allocated to the cloud node;

[0063] When the priority value of the task is "high" in the clustering result and the load of the cloud node is "low", the task is preferentially allocated to the cloud node;

[0064] A simulation model is constructed by using a simulation tool MATLAB, and according to each task allocation scheme in the task allocation scheme set, the corresponding cloud node state data, task and network topology are set, and the simulation transmission of the task is performed to obtain simulation data corresponding to each task allocation scheme, including the simulation values (i.e. predicted values) of the delay of the task, the resource utilization rate and the power consumption of the cloud node.

[0065] The use of the elbow rule ensures the rationality of the number of clusters. Through this method, the selection of the number of clusters can effectively avoid over-clustering or under-clustering, and ensure that the clustering result reflects the actual priority division of the task. By applying the K-Means clustering method, the tasks can be divided into multiple priority groups (high, medium and low priority tasks) according to different characteristics of the task demand (such as computing resource demand and generation timestamp). This operation can flexibly cope with different demands in a multi-task environment through the clustering algorithm, and avoid the imbalance of scheduling caused by manually setting the priority. Through the fuzzy logic engine, the task allocation scheme can be dynamically derived according to the task demand (such as computing resource demand and timeliness) and the resource state of the cloud node (such as computing ability, load and power consumption). The application of fuzzy logic makes the task scheduling no longer rely on accurate data points, but can handle inputs containing uncertainty or fuzziness, thereby improving the flexibility and fault tolerance of the present application. The MATLAB simulation tool is used to simulate the task transmission, which can accurately predict the delay corresponding to each task allocation scheme, the resource utilization rate and the power consumption of the cloud node. Through the prediction of simulation data, potential resource conflicts and bottlenecks can be identified in advance, so as to optimize resource allocation in advance and avoid problems such as resource competition or delay in the task execution process.

[0066] S2, based on simulation data, constructing a decision matrix in AHP method, calculating the initial weight for constraint optimization, obtaining the optimal weight vector, combining simulation data and task allocation scheme set, defining fitness function for task allocation scheme optimization, outputting the optimal task scheduling scheme, and then feeding back;

[0067] Specifically, based on simulation data, a decision matrix in AHP method is constructed, the initial weight is calculated for constraint optimization, the optimal weight vector is obtained, the simulation data and the task allocation scheme set are combined, the fitness function is defined for task allocation scheme optimization, each data in the simulation data is used as an evaluation standard, the relative importance value between the task allocation scheme and each evaluation standard is set by using the empirical rule, the decision matrix in AHP method is constructed, and the sum of each column in the decision matrix is calculated to obtain the weighted sum of all task allocation schemes under the evaluation standard, and the formula is:

[0068]

[0069] In the formula, s e represents the weighted sum under the e-th evaluation standard, n represents the total number of task allocation schemes, u oe represents the relative importance value between the o-th task allocation scheme and the e-th evaluation standard in the simulation data.

[0070] Based on the weighted sum, each relative importance value in the decision matrix is normalized to obtain a normalized matrix, and the average value of each row is solved to splice to generate the weight vector of the current simulation data.

[0071] The formula for normalizing each relative importance value in the decision matrix is:

[0072]

[0073] In the formula, h oe represents the normalized value of the relative importance value between the o-th task allocation scheme and the e-th data in the simulation data.

[0074] The weight vector includes the weight of the current task allocation scheme under each evaluation standard.

[0075] Based on the normalized matrix, the maximum eigenvalue is obtained by eigenvalue decomposition, and then based on the weight vector and the simulation value of the evaluation standard, the optimal weight vector required by each task allocation scheme is calculated by combining the data envelopment analysis method, and the constraint condition is set.

[0076] The formula for calculating the optimal weight vector of each simulation data is:

[0077]

[0078] where max(·) represents the maximum value operation, W o represents the weight vector of the oth task allocation scheme, φ max represents the maximum eigenvalue, m represents the total number of evaluation criteria, w e represents the weight of the e-th evaluation criterion (obtained in the weight vector), d oe represents the simulation value of the oth task allocation scheme in the e-th evaluation criterion in the simulation data.

[0079] The constraint condition is set as follows:

[0080]

[0081] where, represents the upper limit, v e represents the variable of the e-th evaluation criterion, and n represents the total number of task allocation schemes.

[0082] The task allocation scheme is taken as an individual, a population is generated based on the task allocation scheme set, and the corresponding simulation data and the minimum weight vector are combined to define the fitness function, which minimizes the fitness function, as follows:

[0083] f(p) = w1·L(p) + w2·R(p) + w3·E(p)

[0084] where f(p) represents the fitness value of the task allocation scheme in the individual p, w1 represents the weight of the task delay (extracted from the minimum weight vector), L(p) represents the delay in the simulation data corresponding to the task allocation scheme in the individual p, w2 represents the weight of the cloud node resource utilization (extracted from the minimum weight vector), R(p) represents the resource utilization in the simulation data corresponding to the task allocation scheme in the individual p, w3 represents the weight of the cloud node power consumption (extracted from the minimum weight vector), and E(p) represents the power consumption in the simulation data corresponding to the task allocation scheme in the individual p.

[0085] The fitness value is calculated, and the individual is self-adjusted according to the fitness value, as follows:

[0086]

[0087] where p i (t+1) represents the self-adjusted i-th individual p at the t+1th iteration, p i (t) represents the i-th individual p at the tth iteration, a represents the learning rate (which can be set through experiments and related knowledge), N represents the population size, f(p j (t)) represents the fitness value of the j-th individual p at the tth iteration, and f(p i(t) represents the fitness value of the ith individual p at the tth iteration, p j (t) represents the fitness value of the ith individual p at the tth iteration, p

[0088] Based on the self-adjusting individual, the fitness value is recalculated, the gradient of the self-adjusting individual fitness value is calculated using the partial derivative operation, the self-adjusting individual is locally optimized by combining the update formula of the gradient descent method, a new individual is generated, and the formula is:

[0089]

[0090] In the formula, represents the new individual of the ith individual p after local optimization at the t+1th iteration, and represents the step size (which can be set by an empirical rule), represents the gradient of the fitness value of the ith individual p after self-adjustment at the t+1th iteration;

[0091] A random number generator is used to generate random numbers of standard normal distribution to form a random vector, and a mutation operation is performed on the new individual to obtain a mutated individual, and the formula is:

[0092]

[0093] In the formula, represents the mutated individual of the new individual of the ith individual p after local optimization, represents the exploration value (which can be set by experiment and related knowledge), and randn(d) represents the random vector of standard normal distribution;

[0094] The calculation of the fitness value, the self-adjustment of the individual, the partial derivative operation, the local optimization and the mutation operation are repeatedly executed, and when the number of iterations reaches the maximum number, the optimal task scheduling scheme is output;

[0095] The optimal task scheduling scheme includes the execution order and allocation of each task on the cloud node.

[0096] By using simulation data as the evaluation standard, it has high realistic representativeness, unlike traditional methods which rely on theoretical weights or artificial rules, in the present application, the delay, power consumption and resource utilization rate key indicators are obtained by simulation to obtain data expression, which not only truly reflects the resource consumption and performance of the task after execution, but also provides accurate quantitative input for AHP and DEA (Data Envelopment Analysis Method);

[0097] Secondly, by AHP, the decision matrix and normalization processing are constructed, the subjective modeling of the complex relationship between the task scheme and the evaluation standard is realized, the guiding value of the expert experience in the early scheduling weight construction process can be reflected, however, considering that AHP may have consistency deficiency or subjective deviation, the present application further introduces the DEA method for weight optimization, the eigenvalue and linear programming model are used for secondary solving, the influence of each evaluation standard in different task schemes is objectively optimized and adjusted, the cascade combination of AHP and DEA reaches the design of the fusion of theory and data driving, and flexibility and accuracy are considered;

[0098] Moreover, in the design of the fitness function, the weight vector and the simulation index are fused, the weighted minimum multi-objective performance index is formed, the single objective expression that can be optimized is formed, so as to reduce the calculation complexity, the fitness function is not only used for evaluating the scheduling quality of the current individual, but also guides the convergence direction of the algorithm in the whole evolution process, through gradient descent and partial derivative optimization, the efficiency of the task scheme to the optimal solution is effectively improved, and the random vector mutation mechanism breaks the local optimal trap, and strengthens the population diversity and algorithm robustness.

[0099] Further, the scheduling of computer cloud resources is performed based on the optimal task scheduling scheme, and after the scheduling, the delay of the task, the resource utilization rate and the power consumption of the cloud node are recorded in real time and fed back to the fitness function for optimization, and a new task scheduling scheme is obtained;

[0100] When the fitness value of the new task scheduling scheme is less than the fitness value of the optimal task scheduling scheme, the subsequent scheduling of computer cloud resources is performed according to the new task scheduling scheme, otherwise the scheduling of computer cloud resources is performed using the optimal task scheduling scheme.

[0101] The cloud resource scheduling is performed through the optimal task scheduling scheme, this process can guide the scheduling of subsequent tasks through the existing optimal scheme, ensure that the resources are efficiently and reasonably allocated to each task, through the feedback mechanism, the present application can dynamically evaluate the performance of the current scheme in each scheduling period, and compare with the historical optimal scheme, if it is found that the fitness value of the new task scheduling scheme is lower than that of the existing optimal scheme, the current optimal scheme is retained, otherwise it is adjusted;

[0102] Moreover, the optimization process has the following advantages:

[0103] Dynamic adaptability: can quickly respond to the changes of the cloud node resource state, adjust the scheduling scheme in time, thereby improving the resource utilization efficiency and avoiding unnecessary resource waste;

[0104] Self-optimization: Through real-time feedback, the fitness function can self-adjust and continuously improve the quality of task scheduling solutions. This self-optimization mechanism is particularly important when facing changing loads and task demands, effectively avoiding long-term stagnation in local optimal solutions.

[0105] Efficient decision support: Real-time recording and feedback to the fitness function mechanism enables each adjustment to be based on the latest real-time data, avoiding outdated decisions based on historical data.

[0106] S3, display through a visual interface and store data using a database;

[0107] Specifically, display through a visual interface refers to generating a comparison chart using a visualization tool and displaying the comparison result of the fitness value of the new task scheduling solution and the fitness value of the optimal task scheduling solution through the comparison chart;

[0108] Red indicates that the fitness value of the new task scheduling solution is greater than or equal to the fitness value of the optimal task scheduling solution;

[0109] Green indicates that the fitness value of the new task scheduling solution is less than the fitness value of the optimal task scheduling solution.

[0110] By generating a comparison chart using a visualization tool, complex fitness values can be presented to users in a simple and intuitive manner. The comparison chart provides a clear way for users to easily see the differences between the new scheduling solution and the optimal scheduling solution, allowing for better decision-making. Additionally, the use of color differentiation makes the comparison results more eye-catching.

[0111] Further, storing data using a database refers to compressing the optimal task scheduling solution and the comparison chart using a compression algorithm, generating a folder format, storing the folder format using a database, and after storing, adding a timestamp to the folder format and sorting according to the timestamp.

[0112] By compressing the optimal task scheduling solution and comparison chart files before saving, the file size is greatly reduced, optimizing the use of storage resources. The database can more effectively manage large amounts of scheduling solution data. Additionally, organizing multiple related files into a folder format allows the optimal task scheduling solution and its corresponding comparison chart and other data files to remain structured and associated.

[0113] The embodiment also provides an intelligent scheduling system for computer cloud resources, comprising:

[0114] The simulation module is configured to collect task demand data and state data of the cloud nodes for preprocessing, calculate priority values, perform clustering of the priority values using a K-Means clustering method, and perform fuzzy reasoning and simulation simulation using fuzzy logic based on the simulation data, and output simulation data.

[0115] The constraint optimization module is configured to construct a decision matrix in the AHP method based on the simulation data, calculate initial weights for constraint optimization, obtain a final weight vector, combine the simulation data and a task allocation scheme set, define a fitness function for optimization of the task allocation scheme, output an optimal task scheduling scheme, and perform feedback.

[0116] The storage and display module is configured to display through a visual interface and store data using a database.

[0117] The embodiment also provides a computer device suitable for the intelligent scheduling method of computer cloud resources, which includes a memory and a processor.

[0118] The computer device can be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is configured to perform wired or wireless communication with an external terminal. The wireless communication can be achieved through WIFI, a carrier network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device. In addition, the input device can be an external keyboard, touchpad or mouse, etc.

[0119] The embodiment also provides a storage medium, which stores a computer program, and the computer program is executed by a processor to implement the intelligent scheduling method of computer cloud resources proposed in the above embodiment. The storage medium can be implemented by any type of volatile or non-volatile storage devices or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic storage, a flash memory, a magnetic disk, or an optical disk.

[0120] It should be noted that the above embodiment is only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. An intelligent scheduling method of computer cloud resources, characterized in that: include, The task requirement data and cloud node status data are collected and preprocessed to calculate priority values. After clustering the priority values ​​using the K-Means clustering method, the clustering results are generated. Fuzzy logic is applied in combination with the clustering results to perform fuzzy inference, obtain a set of task allocation schemes for simulation, and output simulation data. The task requirement data includes computing resource requirements and task generation timestamps, while the cloud node status data includes computing power, memory capacity, load, and power consumption. The simulation data includes simulated values ​​of task latency, cloud node resource utilization, and power consumption. Based on simulation data, a decision matrix in the AHP method is constructed, initial weights are calculated for constraint optimization, and the final weight vector is obtained. Combining simulation data and a set of task allocation schemes, a fitness function is defined to optimize the task allocation scheme, output the optimal task scheduling scheme, and then feedback is performed. The data is displayed through a visual interface and stored using a database.

2. The intelligent scheduling method of computer cloud resources according to claim 1, characterized in that: The process involves constructing a decision matrix in the AHP method based on simulation data, calculating initial weights for constraint optimization to obtain the final weight vector, and combining simulation data with a set of task allocation schemes to define a fitness function for task allocation scheme optimization. This involves using each data point in the simulation data as an evaluation criterion, using empirical rules to set the relative importance value between the task allocation scheme and each evaluation criterion, constructing a decision matrix in the AHP method, and calculating the sum of each column in the decision matrix to obtain the weighted sum of all task allocation schemes under that evaluation criterion. Based on the weighted summation, each relative importance value in the decision matrix is ​​normalized to obtain a normalized matrix, and the average value of each row is calculated and concatenated to generate the weight vector of the current simulation data. The weight vector contains the weights of the current task allocation scheme under each evaluation criterion; After performing eigenvalue decomposition based on the normalized matrix to obtain the largest eigenvalue, the final weight vector required for each task allocation scheme is calculated based on the weight vector and the simulated value of the evaluation criteria, combined with the data envelopment analysis method, and constraints are set. Treating task allocation schemes as individuals, a population is generated based on the set of task allocation schemes for initialization. Then, combining the corresponding simulation data and the final weight vector, a fitness function is defined and minimized. Calculate fitness values ​​and adjust the individual's self-adjustment based on these fitness values; Based on the self-adjusting individual, the fitness value is recalculated, and the gradient of the fitness value of the self-adjusting individual is obtained by using partial derivative operation. Combined with the update formula of gradient descent method, the self-adjusting individual is locally optimized to generate a new individual. Use a random number generator to generate random numbers from a standard normal distribution, concatenate them to form a random vector, and then combine this vector with the new individual to perform a mutation operation to obtain the mutated individual; Repeatedly perform fitness value calculation, individual self-adjustment, partial derivative calculation, local optimization and mutation operation. When the number of iterations reaches the maximum, output the optimal task scheduling scheme. The optimal task scheduling scheme includes the execution order and allocation of each task on the cloud node.

3. The method for intelligent scheduling of computer cloud resources of claim 2, wherein: The re-feedback indicates that the scheduling of the computer cloud resources is performed based on the optimal task scheduling scheme, and after the scheduling, the delay of the task, the resource utilization rate of the cloud node and the power consumption feedback are recorded in real time to the fitness function for optimization, so as to obtain a new task scheduling scheme; When the fitness value of the new task scheduling scheme is less than the fitness value of the optimal task scheduling scheme, the scheduling of the subsequent computer cloud resources is performed according to the new task scheduling scheme, otherwise the scheduling of the computer cloud resources is continuously performed using the optimal task scheduling scheme.

4. The method of claim 3, wherein: The displaying through the visual interface indicates that a comparison chart is generated using a visual tool, and the comparison result of the fitness value of the new task scheduling scheme and the fitness value of the optimal task scheduling scheme is displayed through the comparison chart; Red indicates that the fitness value of the new task scheduling scheme is greater than or equal to the fitness value of the optimal task scheduling scheme; Green indicates that the fitness value of the new task scheduling scheme is less than the fitness value of the optimal task scheduling scheme.

5. The method for intelligent scheduling of computer cloud resources of claim 4, wherein: The storing of the data using the database indicates that after the optimal task scheduling scheme and the comparison chart are compressed using a compression algorithm, a folder format is generated, the folder format is stored using a database, and after the folder format is stored, a time stamp is added to the folder format, and the folder format is sorted according to the time stamp.

6. The intelligent scheduling of computer cloud resources method of claim 5, wherein: The collecting of the task demand data and the state data of the cloud node for preprocessing and the calculation of the priority value indicate that the task demand data is obtained from the Internet of Things device through an API interface, and the state data of each cloud node is recorded; The task demand data includes computing resource demand and time stamp of task generation; The state data of the cloud node includes computing power, memory capacity, load and power consumption; All data are denoised, normalized and filled; After using an empirical rule to set the weight of each data in the task demand data, the overall value of the task demand data is calculated through a weighted summation formula, which is defined as the priority value.

7. The method of claim 6, wherein: After the priority value is clustered using the K-Means clustering method, a clustering result is generated, fuzzy reasoning is performed by applying fuzzy logic combined with the clustering result, a task allocation scheme set is obtained for simulation simulation, and simulation data is output; Based on the clustering result, each task in the task demand data and the resource state data are used as input variables of fuzzy logic, fuzzy rules are set using the Delphi method, and fuzzy reasoning is performed using a fuzzy rule engine to obtain a task allocation scheme set; A simulation model is constructed using a simulation tool MATLAB, and according to each task allocation scheme in the task allocation scheme set, the corresponding cloud node state data, task and network topology are set, the simulation transmission of the task is performed, and the simulation data corresponding to each task allocation scheme is obtained, including the simulation values of the delay of the task, the resource utilization rate of the cloud node and the power consumption.

8. An intelligent scheduling system of computer cloud resources, based on the intelligent scheduling method of computer cloud resources according to any one of claims 1-7, characterized in that: It comprises, The collection simulation module is used for collecting task demand data and state data of cloud nodes for preprocessing, calculating priority values, and clustering priority values using the K-Means clustering method, and then applying fuzzy logic for fuzzy reasoning and simulation simulation, and outputting simulation data. The constraint optimization module is used for constructing a decision matrix in the AHP method based on the simulation data, calculating initial weights for constraint optimization, obtaining the final weight vector, defining a fitness function for task allocation scheme optimization in combination with the simulation data and the task allocation scheme set, outputting an optimal task scheduling scheme, and then feeding back. The storage display module is used for display through a visual interface and storage of data by using a database. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor executes the computer program to implement the steps of the intelligent scheduling method of computer cloud resources in any one of claims 1-7.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to implement the steps of the intelligent scheduling method of computer cloud resources in any one of claims 1-7.

Images