Platform resource self-adaptive scheduling method and system for multiple service scenes

Abstract

The invention discloses a multi-service scene-oriented platform resource adaptive scheduling method and system, and the method comprises the steps: constructing a simulation service operation time prediction model; based on different resource states of the platform, simulation sample data sets which are generated through simulation and face different service scenes and operation durations of different volume services, a simulation service operation time estimation model is trained; performing parameter fine tuning on the trained simulation service operation time estimation model based on the platform resource operation sample data set to obtain a fine-tuned simulation service operation time estimation model; inputting to-be-estimated traffic volume and platform resource state monitoring data of each service into the fine-tuned simulation service operation time estimation model for prediction to obtain operation time length estimation values of the services of different volumes; on the basis of constraints, on the premise that service processing priorities are adjusted according to operation duration estimated values of services of different volumes and normal development of all the services is guaranteed, frequent regulation and control of resources are avoided by setting a cost target function.

Description

Technical Field

[0001] The present invention relates to the field of power information processing, and in particular to a platform resource adaptive scheduling method and system for multiple business scenarios. Background Art

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.

[0003] The power system is transforming into a complex network dominated by renewable energy and featuring large-scale AC / DC interconnection. The integrated application of massive amounts of multi-source measurement data requires comprehensive monitoring and analysis of energy metering data and marketing archive data from data acquisition systems, as well as multi-service support. The diverse business operations and massive data processing demands place higher demands on the resource scheduling of data application platforms, especially for real-time processing such as distributed power generation group scheduling and control. Therefore, it is necessary to adjust business operation priorities to meet precise timeliness requirements, thereby achieving adaptive scheduling of platform resources.

[0004] Traditional resource scheduling methods often use static rules or fixed threshold strategies, such as a scheduling mechanism based on business priority queues or a response mode that triggers expansion based on resource usage. However, these methods have the following significant problems: (1) Existing methods lack dynamic adaptability and are difficult to cope with scenarios where business volume fluctuates instantaneously and resource demand changes nonlinearly; (2) Traditional deep learning models rely on the completeness of data samples in estimating business runtimes, but in actual environments, there is little available sample data for platform resource operation monitoring, resulting in large prediction errors. (3) Existing scheduling strategies have high resource control costs when adjusting business priorities, and often ignore the problem of high business processing latency caused by frequent resource switching in a short period of time. Summary of the Invention

[0005] In order to overcome the shortcomings of the above-mentioned existing technologies, the present invention provides a platform resource adaptive scheduling method and system for multiple business scenarios, which adjusts the business processing priority according to the estimated running time of businesses of different volumes, and avoids frequent resource regulation by setting a cost objective function under the premise of ensuring the normal development of each business.

[0006] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: In a first aspect, the present invention provides a platform resource adaptive scheduling method for multiple business scenarios, comprising: Obtaining business scenario operation sample data, classifying and preprocessing the business scenario operation sample data according to the business scenario, and forming platform resource operation sample data sets for different business scenarios; Simulating different volume businesses based on the platform resource operation sample data set, setting different resource states of the platform, and simulating the running time of different volume businesses; Construct a simulation business runtime estimation model. This model is trained based on the different resource states of the platform, simulation sample data sets generated for different business scenarios, and the runtime of businesses of different volumes. This yields a trained simulation business runtime estimation model. Fine-tuning parameters of the trained simulation business running time estimation model based on the platform resource running sample data set to obtain a fine-tuned simulation business running time estimation model; Input the traffic volume and platform resource status monitoring data of each business to be estimated into the fine-tuned simulation business running time estimation model to predict the running time of businesses of different volumes. Based on the constraints, the business processing priority is adjusted according to the estimated running time of the businesses of different volumes.

[0007] According to a further technical solution, the business scenario operation sample data includes business data and monitoring data, and the monitoring data includes the proportion of transmission resources, the proportion of storage resources, the proportion of cache resources, and the proportion of computing resources.

[0008] A further technical solution is to use a long-short time neural network algorithm to build a simulation business running time estimation model.

[0009] A further technical solution uses cross entropy to construct a loss function based on the estimated running time of different-volume businesses and the running time of different-volume businesses obtained under an ideal simulation running environment.

[0010] A further technical solution uses a transfer learning algorithm to fine-tune the parameters of the trained simulation business running time estimation model, and uses cross entropy to construct a loss function by comparing the estimated running time of businesses of different sizes with the running time of businesses of different sizes obtained in the actual operating environment.

[0011] According to a further technical solution, the constraints include platform resource load state constraints and load increase and decrease ramping constraints.

[0012] Further technical solutions take into account the platform resource load status constraints and adjust the business processing priority according to the estimated running time of businesses of different sizes; consider the load increase and decrease ramping constraints, and minimize the resource allocation amount and business priority control frequency based on the objective function to maximize resource utilization.

[0013] In a second aspect, the present invention provides a platform resource adaptive scheduling system for multiple business scenarios, comprising: A data acquisition module is configured to: acquire business scenario operation sample data, classify and pre-process the business scenario operation sample data according to the business scenario, and form a platform resource operation sample data set for different business scenarios; A business simulation module is configured to simulate different volume businesses based on the platform resource operation sample data set, set different resource states of the platform, and simulate the running time of different volume businesses; A model building module is configured to: build a simulation business runtime estimation model, train the simulation business runtime estimation model based on different resource states of the set platform, simulation sample data sets for different business scenarios, and the runtime of businesses of different volumes, and obtain a trained simulation business runtime estimation model; A model fine-tuning module is configured to: fine-tune parameters of the trained simulation business runtime estimation model based on the platform resource operation sample data set to obtain a fine-tuned simulation business runtime estimation model; The model prediction module is configured to: input the traffic volume and platform resource status monitoring data of each business to be estimated into the fine-tuned simulation business running time estimation model to perform prediction, and obtain the estimated running time value of different volume businesses; The service adjustment module is configured to adjust the service processing priority according to the estimated running time of the services of different volumes based on the constraints.

[0014] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in a platform resource adaptive scheduling method for multiple business scenarios as described in the first aspect.

[0015] In a fourth aspect, the present invention provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the method for adaptively scheduling platform resources for multiple business scenarios as described in the first aspect are implemented.

[0016] One or more of the above technical solutions have the following beneficial effects: In view of the fact that there is little sample data available for platform resource operation monitoring and it is difficult to complete the construction of a business operation time estimation model using deep learning, the present invention divides the model training process into two stages. First, a sample set is obtained according to an ideal simulation environment to construct a simulation business operation time estimation model, thereby completing the model parameter pre-training; then, based on the actual sample data, transfer learning is constructed to fine-tune the parameters of the simulation business operation time estimation model to obtain a business operation time estimation model suitable for the actual platform operation status.

[0017] The present invention aims at the new generation of electricity consumption information collection system, which has many business types and large operating volume, and some businesses have high requirements for data processing timeliness. Therefore, the platform needs to schedule resources according to business needs. On the basis of adjusting the business processing priority according to the estimated operating time of businesses of different volumes, resource allocation that maximizes resource utilization is achieved by setting a cost objective function. At the same time, under the premise of ensuring the normal development of each business, frequent resource regulation is avoided. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0019] Figure 1 This is a flowchart of a platform resource adaptive scheduling method for multiple business scenarios according to an embodiment of the present invention. DETAILED DESCRIPTION

[0020] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0021] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0022] In the absence of conflict, the embodiments of the present invention and the features thereof may be combined with each other.

[0023] Example 1 like Figure 1 As shown, this embodiment discloses a platform resource adaptive scheduling method for multiple business scenarios, which includes the following steps: S1: Obtain business scenario operation sample data, classify and pre-process the business scenario operation sample data according to the business scenario, and form platform resource operation sample data sets for different business scenarios; In this embodiment, business scenario operation sample data, including business data and monitoring data, is obtained from the next-generation electricity consumption information collection system, the provincial measurement data fusion application platform, and the operation and maintenance monitoring system. The monitoring data includes the proportion of transmission resources, storage resources, cache resources, and computing resources.

[0024] Taking electricity spot trading as an example, business data includes the electricity meter's electricity value, voltage value, current value, user profile, and rates for four time periods: peak and off-peak; monitoring data includes the proportion of transmission resources, storage resources, cache resources, and computing resources.

[0025] The business scenario operation sample data is classified according to the business scenario and preprocessed. For example, in the scenario of electricity spot trading, the business scenario operation sample data includes data collection frequency, transaction duration, data volume, user electricity consumption, etc.

[0026] The data preprocessing process includes operations such as missing data removal, cleaning, and data normalization. Business scenario operation sample data includes the operating volume of different businesses at the same time, platform resource monitoring data, and business operation duration.

[0027] S2: Simulating different volume services based on the platform resource operation sample data set, setting different resource states of the platform, and simulating the running time of different volume services; In this embodiment, based on the Gaussian mixture model, the running time of different volume services is simulated to obtain the different transmission resources, storage resources, cache resources, and computing resources required in an ideal simulation running environment when different volume services are running normally.

[0028] Furthermore, simulation samples are generated based on the Gaussian mixture model. The Gaussian mixture model treats the business scenario running sample data as multiple Gaussian distributions for modeling. The specific steps for simulating different volume businesses based on the Gaussian mixture model to generate simulation sample data are as follows: (1) Initially define the number of Gaussian mixture models and randomly extract multiple subsets from the classified business scenario operation sample data set (platform resource operation sample data set for different business scenarios). The number of subsets should be consistent with the number of Gaussian mixture models, and the sample size ratio of the customized subsets should be 5%.

[0029] (2) Calculate the mean and covariance matrix of each subset, initialize the data distribution parameters of the Gaussian mixture model based on the mean and covariance matrix of the subset, and use the expectation maximization algorithm to optimize the model parameters, that is: Expected steps: Calculate the posterior probability of each data point belonging to each initially defined Gaussian mixture model; Maximization step: Re-divide the sample data according to the posterior probability calculated in the expectation step, that is, the sample set is classified into the subset of the Gaussian mixture model with the maximum posterior probability, and then update the mean and covariance matrix of each Gaussian mixture model to maximize the likelihood function of the subset.

[0030] Iteration: Repeat the expectation step and the maximization step until the parameters of each Gaussian mixture model converge or the preset number of iterations is reached, and the parameters of each Gaussian mixture model are fixed.

[0031] (3) Use the business scenario to run the sample data validation set to evaluate the fit of each Gaussian mixture model, that is, generate the Euclidean distance between the data point and the validation data point. The fit is the inverse of the distance. The larger the distance, the lower the fit. Based on the verification results, feedback is given to step (1) to adjust the model parameters, such as the number of subsets, initialization parameters, etc., until the fit of each Gaussian mixture model is not less than the set fit threshold.

[0032] (4) Based on the Gaussian mixture model with fixed parameters, new data points are sampled from each Gaussian distribution. The generation probability of each data point is determined by the weight after normalization of the model fit. The data points sampled from each Gaussian distribution are combined to form a complete set of simulation sample points.

[0033] Different business volumes are defined based on their operational volume. For example, for spot electricity trading, the operational volume is calculated based on data collection frequency, transaction duration, data volume, and application data items. Data collection frequency is measured in minutes, such as 15 minutes; transaction duration is measured in hours; and data volume is measured in tens of thousands, such as for 30,000 users, 100,000 users, and 150,000 users. The operational volume data item for a business is the time it takes for the provincial-level measurement data fusion application platform to execute the business from start to finish, under a certain platform resource operating state. The platform's processed data output is an hourly indicator curve of user electricity consumption. For example, if an electricity spot trading business is carried out with the participation of 150,000 users, the business operation volume is (60 / 15)*15*1=600,000 users' electricity consumption data, and the resource status data of the transmission resources (such as 30%), storage resources (such as 55%), cache resources (such as 60%), and computing resources (such as 50%) in the platform are monitored. Under this resource status data, the business operation time of the provincial-side measurement data fusion application platform to carry out the electricity spot trading business with the participation of 150,000 users is 20 minutes.

[0034] The platform's resource status data, including transmission resource status data, storage resource status data, cache resource status data, and computing resource status data, are all expressed as percentages. Setting these percentages allows you to configure the ideal simulation environment. For example, the platform's current computing resource status data can be 30%, 40%, 50%, 60%, 70%, and 80%, totaling six scenarios.

[0035] S3: Build a simulation business runtime estimation model. This model is trained based on different resource states of the platform, simulation sample datasets generated for different business scenarios, and the runtimes of businesses of different volumes. This yields a trained simulation business runtime estimation model. In this embodiment, the platform provides different resource status data and simulation sample data sets for different business scenarios generated by simulation as input data, and a long-short time neural network algorithm (LSTM) is used to build a simulation business running time estimation model, output the running time estimation value of different volume businesses, and use the running time of different volume businesses obtained by simulation as the target output value, and feedback is used to optimize the simulation business running time estimation model.

[0036] The specific process is as follows: S301: Based on step S2, the simulation sample data sets for different business scenarios generated by the simulation under different resource status data of the setting platform are obtained as input data, and the running time of different volumes of business under the ideal simulation running environment is obtained as the target output value. A business running time sample data set is constructed and divided into a training set and a test set in a ratio of 7:3; S302: Based on the training set in the business runtime sample data set, the simulation sample data generated by the provincial measurement data fusion application platform in different resource status data and simulation for different business scenarios are input into the simulation business runtime estimation model built using LSTM, and the estimated values ​​of the runtime of different business volumes are output; S303: Based on the estimated runtime values ​​of services of different volumes and the runtime y of services of different volumes obtained under an ideal simulation operating environment, a loss function is constructed using cross entropy. The loss function is fed back to step S302 via a gradient, and the parameters of the simulation service runtime estimation model are adjusted through iterative feedback until the loss function falls below a set threshold. S304: Based on the test set in the business running time sample data set, the simulation business running time estimation model is tested. If the test accuracy is lower than the set accuracy threshold, it is fed back to step S302 to initialize and retrain the simulation business running time estimation model parameters; if the test accuracy is higher than the set accuracy threshold, the simulation business running time estimation model is fixed and the trained simulation business running time estimation model is output.

[0037] S4: Based on the platform resource operation sample data set, a transfer learning algorithm is used to fine-tune the parameters of the trained simulation service runtime estimation model to obtain a fine-tuned simulation service runtime estimation model; In this embodiment, based on the actual platform resource operation sample data set for different business scenarios obtained in step S1, a transfer learning algorithm is used to fine-tune the parameters of the trained simulation business operation time estimation model to obtain a fine-tuned simulation business operation time estimation model, that is, a business operation time estimation model suitable for the actual platform operation status.

[0038] The specific process is as follows: S401: Based on a sample dataset of platform resource operations for different business scenarios, transfer learning is used to perform reward feedback tuning on the parameters of the simulation business runtime estimation model, outputting runtime estimates for businesses of different sizes. S402: The estimated running time of different-volume businesses is compared with the running time of different-volume businesses obtained in the actual operating environment, and a loss function is constructed using cross entropy. The loss function is fed back to step S401 through gradient iteration until the loss function is lower than the set threshold. The parameters are fixed and a business running time estimation model suitable for the actual platform operating status is output.

[0039] S5: Input the traffic volume and platform resource status monitoring data of each business to be estimated into the fine-tuned simulation business running time estimation model for prediction, and obtain the estimated running time value of businesses of different volumes.

[0040] In this embodiment, since some businesses are carried out randomly and with high timeliness, when carrying out adaptive resource scheduling, the platform resource status monitoring data and the operating volume of each business (business operation volume) are used as input data, and based on the business operation time estimation model, the estimated operation time of businesses of different sizes is output.

[0041] S6: Based on the constraints, adjust the business processing priority according to the estimated running time of the different-volume businesses.

[0042] In this embodiment, the platform resource load status and load increase and decrease ramping constraints are taken into consideration, and the business processing priority is adjusted according to the estimated running time of businesses of different sizes. Under the premise of ensuring the normal development of each business, the cost objective function is set to avoid frequent resource regulation.

[0043] The specific process is as follows: S601: Considering the platform resource load constraints, adjust the service processing priority according to the estimated running time of different services as follows:

[0044] in, Indicates that the business processing priority is adjusted according to the estimated running time of different business volumes. The smaller the value, the higher the priority. The business priority is obtained through sorting. Indicates time At point , the volume is Business Estimated runtime of With business Runtime requirements The distance is calculated using Euclidean distance. The smaller the data, the more urgent the business operation. Indicates the moment; Indicates Subsequent time period Businesses run within The volume; 、 、 、 、 Respectively expressed in The total platform business volume, transmission resource ratio, storage resource ratio, cache resource ratio, and computing resource ratio at the moment; Represents a business running time estimation model; The volume of the business running time estimation model output is Business Estimated runtime of .

[0045] The business Its operating volume ,exist Total platform business volume at the moment , Transmission resource ratio , Storage resource ratio , cache resource ratio and computing resource ratio Input business running time estimation model , the output volume is Business Estimated runtime of , the business The runtime requirement time set by the platform (ie business Must be in Processing completed within the time) and Subtraction , The smaller the value, the smaller the redundant time of business processing. (If it is a negative number, it means that the timeliness is lagging, and the business needs to be increased. Priority of processing), all business Sort by, and the smallest value gets the highest business processing priority.

[0046] S602: Considering the subsequent load ramping constraints, while ensuring the normal operation of all services, an objective function is set to avoid the problem of resource regulation caused by sudden services affecting the normal operation of other services. At the same time, the frequency of service priority regulation is minimized. The constructed objective function is specifically as follows:

[0047]

[0048] in, represents the objective function; Indicates the service priority adjustment period; Indicates the business priority adjustment period; Indicates the amount of resources allocated to maximize resource utilization (avoiding frequent scheduling of business priorities, which leads to reduced utilization of platform resources); Indicates the proportion of total platform resource consumption. Minimizing this indicator means that when the platform carries the same business volume, it maximizes the allocation of available resources to the current business and reduces the subsequent business volume. Indicates the current platform Resource utilization; Indicates the proportion of different resource weights; Indicates that under the current platform resource operation status, the business Occupancy The estimated runtime of the resource; Indicates the frequency constraint of service priority adjustment, including transmission resource priority adjustment, storage resource priority adjustment, cache resource priority adjustment, and computing resource priority adjustment operations; 、 The weight coefficients representing resource allocation constraints and service priority adjustment frequency constraints; express Time business right The resource priority adjustment operation, the value is 0 or 1, 0 means no priority adjustment, 1 means priority adjustment.

[0049] maximize Indicates that the platform Maximize resource utilization to ensure that business can be completed in a short time Provide sufficient redundant resources for subsequent business processing; allocate resources to maximize resource utilization , ensuring that the four types of resources have the highest utilization rate under the current total platform business volume; Sure Is 0 or 1, that is, whether the business priority needs to be adjusted, and the frequency constraint of business priority adjustment according to Determine the allocation and scheduling of the four types of resources. That is, adjusting the business priority requires recalculating the proportion of the four types of resources; and There is a resource ratio correlation between Changes will cause changes, which in turn affects the utilization of different resources. It is to maximize the utilization of different resources, so there is a mutual constraint relationship between the two, through the weight coefficient 、 Adjusting the proportion of the two can meet the needs of partially adjusting business priorities while avoiding the reduction in overall resource utilization caused by frequent priority adjustments, thereby improving business processing efficiency and freeing up sufficient redundant resources for subsequent business processing.

[0050] S603: Will be The business data and monitoring data acquired at all times are input into the fine-tuned simulation business runtime estimation model to obtain future predictions. Estimated running time of different services within a time period; S604: Under the premise of ensuring the normal operation of various businesses, The recommended service priority adjustment operation in step S601 is selected for execution or non-execution. Minimize the objective function value within a period, obtain a set of business priority adjustment sequences, and use the obtained output sequence to act on platform resource scheduling.

[0051] Table 1 Accuracy of estimated operation time of user electricity spot trading business

[0052] Table 1 shows the test results of the accuracy of the user electricity spot trading business operation time estimation before and after fine-tuning the parameters. The estimation accuracy is significantly improved after fine-tuning the parameters, indicating the effectiveness of the method described in the present invention.

[0053] Example 2 This embodiment discloses a platform resource adaptive scheduling system for multiple business scenarios, including: A data acquisition module is configured to: acquire business scenario operation sample data, classify and pre-process the business scenario operation sample data according to the business scenario, and form a platform resource operation sample data set for different business scenarios; A business simulation module is configured to simulate different volume businesses based on the platform resource operation sample data set, set different resource states of the platform, and simulate the running time of different volume businesses; A model building module is configured to: build a simulation business runtime estimation model, train the simulation business runtime estimation model based on different resource states of the set platform, simulation sample data sets for different business scenarios, and the runtime of businesses of different volumes, and obtain a trained simulation business runtime estimation model; A model fine-tuning module is configured to: fine-tune parameters of the trained simulation business runtime estimation model based on the platform resource operation sample data set to obtain a fine-tuned simulation business runtime estimation model; The model prediction module is configured to: input the traffic volume and platform resource status monitoring data of each business to be estimated into the fine-tuned simulation business running time estimation model to perform prediction, and obtain the estimated running time value of different volume businesses; The service adjustment module is configured to adjust the service processing priority according to the estimated running time of the services of different volumes based on the constraints.

[0054] Example 3 The purpose of this embodiment is to provide a computing device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method of embodiment 1 when executing the program.

[0055] Example 4 The purpose of this embodiment is to provide a computer-readable storage medium, a computer-readable storage medium having a computer program stored thereon, which performs the steps of the method of embodiment 1 when executed by a processor.

[0056] The steps involved in the apparatuses of Examples 3 and 4 above correspond to those of Method Example 1. For detailed implementation, please refer to the relevant description of Example 1. The term "computer-readable storage medium" should be understood to mean a single medium or multiple media containing one or more instruction sets; it should also be understood to include any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and causing the processor to perform any of the methods of the present invention.

[0057] Those skilled in the art will appreciate that the modules or steps of the present invention described above can be implemented using a general-purpose computer device. Alternatively, they can be implemented using program code executable by a computing device, which can then be stored in a storage device and executed by the computing device. Alternatively, they can be fabricated into separate integrated circuit modules, or multiple modules or steps can be fabricated into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0058] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

[0059] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without any creative work are still within the scope of protection of the present invention.

Claims

1. A platform resource adaptive scheduling method for multiple business scenarios, characterized by: include: Obtaining business scenario operation sample data, classifying and preprocessing the business scenario operation sample data according to the business scenario, and forming platform resource operation sample data sets for different business scenarios; Simulating different volume businesses based on the platform resource operation sample data set, setting different resource states of the platform, and simulating the running time of different volume businesses; Construct a simulation business runtime estimation model. This model is trained based on the different resource states of the platform, simulation sample data sets generated for different business scenarios, and the runtime of businesses of different volumes. This yields a trained simulation business runtime estimation model. Fine-tuning parameters of the trained simulation business running time estimation model based on the platform resource running sample data set to obtain a fine-tuned simulation business running time estimation model; Input the traffic volume and platform resource status monitoring data of each business to be estimated into the fine-tuned simulation business running time estimation model to predict the running time of businesses of different volumes. Based on the constraints, the business processing priority is adjusted according to the estimated running time of the businesses of different volumes.

2. A platform resource adaptive scheduling method for multiple business scenarios according to claim 1, characterized in that: The business scenario operation sample data includes business data and monitoring data, and the monitoring data includes the proportion of transmission resources, storage resources, cache resources, and computing resources.

3. The method for adaptively scheduling platform resources for multiple business scenarios according to claim 1, characterized in that: A long-short time neural network algorithm is used to build a simulation business running time estimation model.

4. The method for adaptively scheduling platform resources for multiple business scenarios according to claim 3, characterized in that: Based on the estimated running time of businesses of different sizes and the running time of businesses of different sizes obtained in an ideal simulation running environment, the cross entropy is used to construct the loss function.

5. The method for adaptively scheduling platform resources for multiple business scenarios according to claim 1, characterized in that: The transfer learning algorithm is used to fine-tune the parameters of the trained simulation business running time estimation model. The estimated running time of businesses of different sizes is compared with the running time of businesses of different sizes obtained in the real operating environment, and the cross entropy is used to construct the loss function.

6. The method for adaptively scheduling platform resources for multiple business scenarios according to claim 1, characterized in that: The constraints include platform resource load state constraints and load increase and decrease ramping constraints.

7. A platform resource adaptive scheduling method for multiple business scenarios according to claim 6, characterized in that: Taking into account the platform resource load status constraints, adjust the business processing priority according to the estimated running time of businesses of different sizes; considering the load increase and decrease ramping constraints, minimize the resource allocation amount and business priority control frequency for maximizing resource utilization based on the objective function.

8. A platform resource adaptive scheduling system for multiple business scenarios, characterized by: include: A data acquisition module is configured to: acquire business scenario operation sample data, classify and pre-process the business scenario operation sample data according to the business scenario, and form a platform resource operation sample data set for different business scenarios; A business simulation module is configured to simulate different volume businesses based on the platform resource operation sample data set, set different resource states of the platform, and simulate the running time of different volume businesses; A model building module is configured to: build a simulation business runtime estimation model, train the simulation business runtime estimation model based on different resource states of the set platform, simulation sample data sets for different business scenarios, and the runtime of businesses of different volumes, and obtain a trained simulation business runtime estimation model; A model fine-tuning module is configured to: fine-tune parameters of the trained simulation business runtime estimation model based on the platform resource operation sample data set to obtain a fine-tuned simulation business runtime estimation model; The model prediction module is configured to: input the traffic volume and platform resource status monitoring data of each business to be estimated into the fine-tuned simulation business running time estimation model to perform prediction, and obtain the estimated running time value of different volume businesses; The service adjustment module is configured to adjust the service processing priority according to the estimated running time of the services of different volumes based on the constraints.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the platform resource adaptive scheduling method for multiple business scenarios according to any one of claims 1 to 7 are implemented.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the platform resource adaptive scheduling method for multiple business scenarios according to any one of claims 1 to 7 are implemented.

Images