Computing power lease management method and device, and storage medium

Abstract

The invention discloses a computing power lease management method and device and a storage medium, and relates to the technical field of digital data processing, and the method comprises the steps: receiving a computing power lease form, and determining a task type, a demand time length and a computing power specification corresponding to the computing power lease form; determining a time sequence-based computing power demand rule of the user according to a historical computing power use record of the user corresponding to the computing power lease form; determining computing power analysis data associated with the task type according to platform historical data; calculating a first computing power demand based on the task type, the demand duration, the computing power specification and the computing power demand rule; calculating a second computing power demand based on the task type, the demand duration, the computing power specification and the computing power analysis data; and determining computing power output responding to the computing power lease form according to the first computing power demand, the second computing power demand and the dynamic weight associated with the date. The technical effect of reducing the computing power use cost can be achieved.

Description

Technical Field

[0001] This application relates to the technical field of digital data processing, and particularly to a computing power leasing management method, device, and storage medium. Background Art

[0002] At present, with the rapid development of artificial intelligence, AI large models have been deeply integrated into many business scenarios such as intelligent customer service, image recognition, and natural language processing. Whether in the model training stage or the actual inference stage, AI large models have extremely high requirements for computing power. However, during the operation of many businesses, the computing power demand shows obvious volatility and does not remain at a high level all the time. For example, during large promotion activities on e-commerce platforms, the computing power demand of intelligent recommendation systems will soar sharply, but during the daily operation stage, the computing power demand will drop significantly.

[0003] To cope with the computing power demand during business peaks, enterprises need to purchase a large number of high-end hardware graphics cards. The high hardware procurement costs and subsequent operation and maintenance and power consumption result in too high computing power costs. Summary of the Invention

[0004] The main purpose of this application is to provide a computing power leasing management method, device, and storage medium, aiming to solve the technical problem that the hardware procurement cost and subsequent operation and maintenance and power consumption result in too high computing power costs.

[0005] To achieve the above object, this application provides a computing power leasing management method, which is applied to a computing power leasing management platform. The computing power leasing management method includes: Receiving a computing power leasing form, and determining the task type, required duration, and computing power specification corresponding to the computing power leasing form; According to the historical computing power usage records of the user corresponding to the computing power leasing form, determining the computing power demand pattern of the user based on time series; Determining the computing power analysis data associated with the task type according to the platform historical data; Calculating a first computing power demand based on the task type, the required duration, the computing power specification, and the computing power demand pattern; and calculating a second computing power demand based on the task type, the required duration, the computing power specification, and the computing power analysis data; Determining the computing power output in response to the computing power leasing form according to the first computing power demand, the second computing power demand, and the dynamic weight associated with the date.

[0006] In an embodiment, before the step of receiving the computing power leasing form and determining the task type, required duration, and computing power specification corresponding to the computing power leasing form, it includes: Responding to an input operation of a form submission control, and determining the task type; Render the requirement details control associated with the task type, where the requirement details controls corresponding to different task types are different; In response to an input operation on the requirement details control, parse out the requirement duration and the computing power specification; Generate the computing power lease form according to the task type, the requirement duration, and the computing power specification.

[0007] In one embodiment, the step of determining the user's computing power demand pattern based on time series according to the user's historical computing power usage records corresponding to the computing power lease form includes: Obtain the user's historical computing power usage records, where the historical computing power usage records include execution time periods, computing power consumption values, and business scenario tags; Perform bucket aggregation on the historical computing power usage records based on a preset time window to generate periodic computing power consumption characteristics; Model the periodic computing power consumption characteristics to identify the trend term, seasonal term, and residual term of the user's computing power demand; Combine the business scenario tags to correct the prediction results of the trend term, seasonal term, and residual term to generate the user's computing power demand pattern based on time series.

[0008] In one embodiment, the step of determining the computing power analysis data associated with the task type according to the platform historical data includes: Obtain all historical computing power request records in the platform historical data that match the task type; Perform three-dimensional classification aggregation according to the task type, execution time period, and business scenario tags; Calculate the computing power consumption characteristic values for each classification, including mean, variance, quantile, and resource utilization efficiency coefficient; Correct the computing power consumption characteristic values in combination with external data; Construct an association model based on the random forest algorithm to analyze the dependence relationship between the task type and the computing power consumption characteristic values, and generate the computing power analysis data associated with the task type.

[0009] In one embodiment, the step of calculating the first computing power demand based on the task type, the requirement duration, the computing power specification, and the computing power demand pattern includes: Decompose the computing power demand pattern into a trend term, a seasonal term, and a random term; Establish a mapping relationship between the task type and the standard computing power unit, and convert the computing power specification into a standardized computing power value; Adopt a linear regression model to perform weighted combination of the standardized computing power value, the requirement duration, and the trend term and seasonal term to determine the regression result; Determine the first computing power requirement based on the time decay factor, the resource utilization efficiency coefficient of the user's historical computing power usage, and the regression result.

[0010] In one embodiment, the step of calculating the second computing power requirement based on the task type, the required duration, the computing power specification, and the computing power analysis data includes: Extract the computing power consumption characteristic values matching the task type from the computing power analysis data, including the historical average computing power consumption, the time period fluctuation coefficient, and the resource utilization efficiency benchmark value; Based on the mapping relationship between the computing power specification and the hardware performance parameters, convert the computing power specification into the standard computing power of a unit; Determine the time period correction performance according to the time period fluctuation coefficient and the standard computing power of the unit; Determine the duration efficiency product according to the required duration and the resource utilization efficiency benchmark value; Based on the gradient boosting tree, process the time period correction performance, the duration efficiency product, and the historical average computing power consumption to determine the second computing power requirement.

[0011] In one embodiment, the step of determining the computing power output in response to the computing power lease form according to the first computing power requirement, the second computing power requirement, and the dynamic weight associated with the date includes: Extract the date type label from the platform historical data and establish the mapping relationship between the date type and the business fluctuation coefficient; Determine the dynamic weight according to the mapping relationship between the date type and the business fluctuation coefficient and the standard deviation of the historical computing power requirement; Determine the computing power output by weighted combination of the first computing power requirement, the second computing power requirement, and the dynamic weight.

[0012] In one embodiment, after the step of determining the computing power output in response to the computing power lease form according to the first computing power requirement, the second computing power requirement, and the dynamic weight associated with the date, it includes: Collect the actual computing power output data; Output an alarm prompt message based on the computing power consumption threshold and the actual computing power output data.

[0013] In addition, to achieve the above object, the present application further provides a computing power lease management device, which includes: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the computer program is configured to implement the steps of the computing power lease management method as described above.

[0014] In addition, to achieve the above object, the present application further provides a storage medium, which is a computer-readable storage medium. A program for implementing the computing power leasing management method is stored on the computer-readable storage medium. The program for implementing the computing power leasing management method is executed by a processor to implement the steps of the computing power leasing management method as described above.

[0015] The present application provides a computing power leasing management method. First, by receiving a computing power leasing form, the present application determines the task type, required duration, and computing power specification corresponding to the computing power leasing form; based on the computing power leasing form and the user's historical computing power usage records, the present application determines the user's computing power demand pattern based on time series; based on the platform historical data, the present application determines the computing power analysis data associated with the task type; calculates the first computing power demand based on the task type, the required duration, the computing power specification, and the computing power demand pattern; and calculates the second computing power demand based on the task type, the required duration, the computing power specification, and the computing power analysis data; determines the computing power output in response to the computing power leasing form according to the first computing power demand, the second computing power demand, and the dynamic weight associated with the date. The technical problem of excessively high computing power costs caused by high hardware procurement costs and subsequent operation and maintenance and power consumption is solved, and the technical effect of reducing the computing power usage cost is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings herein are incorporated into and constitute a part of this specification, showing embodiments consistent with the present application and, together with the specification, are used to explain the principles of the present application.

[0017] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the accompanying drawings required for use in the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0018] Figure 1 It is a schematic flowchart provided for Embodiment 1 of the computing power leasing management method of the present application; Figure 2 It is a schematic flowchart provided for Embodiment 3 of the computing power leasing management method of the present application; Figure 3 It is a schematic flowchart provided for Embodiment 4 of the computing power leasing management method of the present application; Figure 4 It is a schematic hardware structure diagram related to the computing power leasing management device of the present application.

[0019] The implementation, functional features, and advantages of the object of the present application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0021] To better understand the technical solutions of the present application, the following will be described in detail in conjunction with the accompanying drawings of the specification and specific implementation manners.

[0022] Currently, in the current rapid development of artificial intelligence, large AI models have been deeply integrated into many business scenarios such as intelligent customer service, image recognition, and natural language processing. Whether in the model training stage or the actual inference stage, large AI models have extremely high requirements for computing power. However, during the operation of many businesses, the computing power requirements show obvious volatility and do not always remain at a high level. For example, during major promotion activities on e-commerce platforms, the computing power requirements of intelligent recommendation systems will soar sharply, but during the daily operation stage, the computing power requirements will drop significantly. To cope with the computing power requirements during business peaks, enterprises need to purchase a large number of high-end hardware graphics cards, and the high hardware procurement costs, as well as subsequent operation and maintenance and power consumption, result in too high computing power costs.

[0023] Therefore, the present application receives a computing power lease form, determines the task type, required duration, and computing power specifications corresponding to the computing power lease form; determines the user's computing power demand pattern based on time series according to the historical computing power usage records of the user corresponding to the computing power lease form; determines the computing power analysis data associated with the task type according to the platform historical data; calculates the first computing power demand based on the task type, the required duration, the computing power specifications, and the computing power demand pattern; and calculates the second computing power demand based on the task type, the required duration, the computing power specifications, and the computing power analysis data; determines the computing power output in response to the computing power lease form according to the first computing power demand, the second computing power demand, and the dynamic weight associated with the date, achieving the technical effect of reducing the computing power usage cost.

[0024] An energy storage system refers to a system that can store energy through a medium or device and release the energy when energy is needed. An energy storage system usually consists of a medium or device for storing energy, a device for converting energy, a device for managing the stored energy, etc.

[0025] It should be noted that the execution subject of this embodiment can be a computing power lease management platform, or a computing service device with data processing, network communication, and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or a computing power lease management device capable of implementing the above functions. This embodiment does not make specific limitations in this regard. The following takes the computing power lease management platform as the execution subject as an example to illustrate this embodiment and the following embodiments.

[0026] Based on this, the present application proposes a computing power leasing management method for the first embodiment. Please refer to Figure 1 , and the computing power leasing management method includes steps S10 to S50: Step S10: Receive a computing power leasing form, and determine the task type, required duration, and computing power specification corresponding to the computing power leasing form.

[0027] In this embodiment, the computing power leasing form is a document in which a user initiates a computing power leasing application to the computing power leasing management platform, and contains key information such as task type, required duration, and computing power specification. The task type indicates the business scope served by the user's leased computing power, such as image recognition in intelligent security; the required duration refers to the time span during which the user plans to use the computing power; and the computing power specification specifies the specific configuration parameters of the computing power device required by the user, such as the model and quantity of GPUs.

[0028] As an alternative implementation, the platform sets up a leasing application entry both on the Web side and the mobile side. The user fills in the form on the corresponding interface, and the platform receives the form data in real time through the data interface, and uses preset data parsing rules to extract the task type, required duration, and computing power specification.

[0029] Step S20: According to the historical computing power usage records of the user corresponding to the computing power leasing form, determine the computing power demand pattern of the user based on the time series.

[0030] In this embodiment, the historical computing power usage records refer to the detailed data of the user's past leased computing power on the platform, covering information such as usage time and usage volume. The computing power demand pattern based on the time series is the trend and characteristics of the user's computing power demand changing over time obtained by performing time series analysis on these historical data.

[0031] As an alternative implementation, the platform uses the pandas and statsmodels libraries in Python to preprocess and model the user's historical computing power usage records, identify the periodicity, seasonality, and long-term trends in the data, so as to determine the computing power demand pattern of the user.

[0032] Step S30: Determine the computing power analysis data associated with the task type according to the platform historical data.

[0033] In this embodiment, the platform historical data includes the computing power leasing and usage data of all users during the platform operation. The computing power analysis data associated with the task type refers to the computing power usage characteristic data obtained by statistically analyzing the platform historical data for a specific task type, such as the average computing power consumption of the task type and the frequency of the occurrence of the computing power usage peak.

[0034] As an alternative implementation, the platform uses big data analysis tools Hive and Spark to regularly summarize and analyze the historical data of the platform, generate a mapping table of task types and computing power analysis data, and store it in the distributed database HBase for quick query and invocation.

[0035] Step S40, calculate the first computing power demand based on the task type, the required duration, the computing power specification, and the computing power demand pattern; and calculate the second computing power demand based on the task type, the required duration, the computing power specification, and the computing power analysis data.

[0036] In this embodiment, the first computing power demand is a computing power value estimated by comprehensively considering the user's own task type, required duration, computing power specification, and computing power demand pattern; the second computing power demand is a value calculated based on the task type, required duration, computing power specification, and computing power analysis data associated with the task type.

[0037] As an alternative implementation, construct a first computing power demand calculation model based on linear regression, use relevant data as the model input to obtain the first computing power demand; use an algorithm based on statistical average to construct a second computing power demand calculation model, and input the corresponding data to calculate the second computing power demand.

[0038] Step S50, determine the computing power output in response to the computing power lease form according to the first computing power demand, the second computing power demand, and the dynamic weight associated with the date.

[0039] In this embodiment, the dynamic weight associated with the date is a weight coefficient preset according to the influence degree of different dates on the computing power demand. For example, the weights of different dates such as weekdays, holidays, and e-commerce promotion days are different. The computing power output refers to the number of computing power resources finally allocated by the platform to the user.

[0040] As an alternative implementation, calculate the computing power configuration provided for the user through the formula "computing power output = first computing power demand × dynamic weight of the corresponding date + second computing power demand × (1 - dynamic weight of the corresponding date)".

[0041] Exemplarily, user A submits a computing power lease form to the computing power lease management platform. The form shows that the task type is live streaming push acceleration, the required duration is from the start to the end of the summer vacation, a total of two months, and the computing power specification is equipped with 32-core CPUs, 128GB of memory, and 8 high-performance GPUs.

[0042] After the platform receives the form, it extracts the task type, required duration, and computing power specification information. By analyzing the historical computing power usage records of the online education platform, it is found that during holidays, due to the increase in live courses, the computing power demand on the platform has increased significantly, and there is a pattern of relatively low demand on weekends every week. At the same time, the platform obtains the computing power analysis data for the live streaming acceleration task type from the mapping table of task type and computing power analysis data, such as average computing power consumption and peak computing power during holidays. Based on the above data, the platform uses the first computing power demand calculation model and combines the computing power demand pattern of the online education platform itself to calculate the first computing power demand; uses the second computing power demand calculation model and combines the general computing power demand of the live streaming acceleration task type to calculate the second computing power demand. Considering that the summer vacation is a peak business period for the online education platform, the platform sets a higher dynamic weight for the summer vacation. Through formula calculation, it is determined to provide a computing power output of 28-core CPUs, 112GB of memory, and 7 high-performance GPUs for the online education platform, which can not only meet the computing power demand for live streaming acceleration during the summer vacation of the platform but also avoid over-allocation of computing power resources and reduce costs.

[0043] This application first determines the task type, required duration, and computing power specification corresponding to the computing power lease form by receiving the computing power lease form; determines the computing power demand pattern of the user based on time series according to the historical computing power usage record corresponding to the computing power lease form of the user; determines the computing power analysis data associated with the task type according to the historical data of the platform; calculates the first computing power demand based on the task type, the required duration, the computing power specification, and the computing power demand pattern; and calculates the second computing power demand based on the task type, the required duration, the computing power specification, and the computing power analysis data; determines the computing power output in response to the computing power lease form according to the first computing power demand, the second computing power demand, and the dynamic weight associated with the date. It solves the technical problem of excessive computing power costs caused by high hardware procurement costs and subsequent operation and maintenance and power consumption, and achieves the technical effect of reducing the computing power usage cost.

[0044] Based on any of the above embodiments, in the second embodiment of this application, before the step of receiving the computing power lease form and determining the task type, required duration, and computing power specification corresponding to the computing power lease form, it includes: Step S1, in response to the input operation of the form submission control, determine the task type.

[0045] In this embodiment, the form submission control is an interactive element on the user interface for initiating the submission operation of the computing power lease form, such as the "Submit" button in a web page or application; the task type indicates the business category that the user needs to carry out when leasing computing power, such as intelligent customer service and big data analysis.

[0046] As an alternative implementation, the platform deploys a drop-down menu on the interactive page for users to lease computing power. After the user clicks on the menu, they select one option from the pre-set task type options. When the user clicks the "Submit" button, the platform captures this operation and obtains the task type selected by the user.

[0047] Step S2: Render the requirement details control associated with the task type. The requirement details controls corresponding to different task types are different.

[0048] In this embodiment, the requirement details control is a set of interactive components used to collect the specific requirements of users for computing power leasing. Due to the differences in the characteristics of computing power requirements for different task types, the corresponding requirement details controls are also different. For example, for an image rendering task, the user may need to input information such as image resolution and rendering frame rate, while for a natural language processing task, the focus may be on information such as the volume of text processing.

[0049] As an alternative implementation, the platform pre-establishes a mapping relationship between the task type and the requirement details control template. After obtaining the task type selected by the user, according to the mapping relationship, the corresponding requirement details control template is loaded from the server and rendered on the page.

[0050] Step S3: In response to the input operation of the requirement details control, parse out the required duration and the computing power specification.

[0051] In this embodiment, the required duration refers to the time range during which the user plans to use the computing power, and the computing power specification is used to describe the performance parameters of the computing power device required by the user, such as the number of CPU cores and the GPU video memory capacity.

[0052] As an alternative implementation, the platform binds an event catcher to each input item in the requirement details control. When the user enters data in these input boxes and completes the operation, the event catcher is triggered, and the platform obtains the input data and parses the data according to the preset parsing rules. For example, for the required duration, by identifying the time unit and value entered by the user, it is converted into a unified time format; for the computing power specification, according to the hardware parameter description entered by the user, the corresponding hardware configuration information is parsed.

[0053] Step S4: Generate the computing power lease form according to the task type, the required duration, and the computing power specification.

[0054] In this embodiment, the computing power lease form is a file that integrates the key information of the user's computing power lease for subsequent processing by the platform.

[0055] As an alternative implementation, the platform generates a computing power rental form in JSON data format. The information such as the task type, required duration, and computing power specification obtained in the steps is organized according to a preset JSON structure to generate the complete computing power rental form data, which is stored in the database of the server for subsequent computing power rental management operations.

[0056] Exemplarily, a game company plans to rent computing power for game testing. On the computing power rental page of the platform, the staff of the game company selects "Game Testing" from the task type drop-down menu and then clicks the "Submit" button. The platform captures this operation and determines that the task type is "Game Testing". According to the mapping relationship between the task type and the demand details control template, the platform renders on the page the demand details control corresponding to the game testing task, and this control includes input items such as "Testing Duration" and "Required Server Configuration (CPU, GPU, memory, etc.)". The staff of the game company fills in "30 days" in the "Testing Duration" input box and fills in "32-core CPU, 64GB memory, 8 RTX 3080 GPUs" in the "Required Server Configuration" input box. The event capturers bound to these input items are triggered, the input data is obtained, and according to the parsing rules, the required duration is parsed as 30 days, and the computing power specification is 32-core CPU, 64GB memory, 8 RTX 3080 GPUs. Finally, the platform organizes the task type "Game Testing", the required duration "30 days", and the computing power specification "32-core CPU, 64GB memory, 8 RTX 3080 GPUs" in JSON format to generate a computing power rental form and stores it in the database for subsequent use in the computing power rental management process.

[0057] Since the demand details control associated with the task type is rendered, the demand details controls corresponding to different task types are different, and thus the computing power requirements for different task types are quantified in different formats, improving the accuracy of computing power requirement assessment.

[0058] Based on any of the above embodiments, in Embodiment 3 of the present application, with reference to Figure 2 , the steps of determining the user's computing power requirement pattern based on time series according to the computing power rental form corresponding to the user's historical computing power usage records include: Step S21, obtaining the user's historical computing power usage records, where the historical computing power usage records include the execution period, the computing power consumption value, and the business scenario label.

[0059] In this embodiment, the historical computing power usage record is a collection of relevant data on the user's past use of computing power on the computing power leasing management platform. The execution period refers to the specific time range during which the user uses the computing power, which can reflect the distribution of computing power usage in the time dimension; the computing power consumption value represents the actual amount of computing power consumed by the user during this execution period and is a key indicator for measuring computing power demand; the business scenario label is used to identify the specific business scenario corresponding to the user's use of computing power, such as intelligent customer service, image rendering, etc.

[0060] As an alternative implementation, the platform creates a storage table for the user's historical computing power usage records in the database, collects the user's computing power usage information in real time from the log system through a data interface, and stores it in a structured manner according to fields such as the execution period, computing power consumption value, and business scenario label. When it is necessary to obtain the historical computing power usage record of a certain user, relevant data is extracted from the database through an SQL query statement.

[0061] Step S22: Perform bucket aggregation on the historical computing power usage records based on a preset time window to generate periodic computing power consumption characteristics.

[0062] In this embodiment, the preset time window is a fixed time length set in advance, such as one day, one week, or one month, etc., which is used to divide the continuous time axis into several equal time periods. Bucket aggregation is to group the historical computing power usage records according to the preset time window and perform statistical calculations on the computing power consumption values within each time period, such as summation, averaging, etc. The periodic computing power consumption characteristics are obtained through the bucket aggregation operation and reflect the computing power consumption pattern of the user in different time cycles.

[0063] As an alternative implementation, use the pandas library in Python to process the obtained historical computing power usage records. First, convert the execution period to a time index, then use the resample() method to resample according to the preset time window, and perform a summation operation on the computing power consumption values to generate the total computing power consumption within each time window, thereby obtaining the periodic computing power consumption characteristics.

[0064] Step S23: Model the periodic computing power consumption characteristics to identify the trend term, seasonal term, and residual term of the user's computing power demand.

[0065] In this embodiment, modeling is to use a suitable time series analysis model to fit the periodic computing power consumption characteristics to reveal the potential laws in the data. The trend term represents the long-term change trend of computing power demand over time, such as gradually increasing or decreasing; the seasonal term reflects the repeated change pattern of computing power demand within a fixed time cycle, such as weekly or monthly periodic fluctuations; the residual term is the random error part that the model cannot explain.

[0066] As an alternative implementation, the SARIMAX (Seasonal AutoRegressive Integrated Moving Average with Exogenous Regression) model in the statsmodels library of Python is used to model the periodic computing power consumption characteristics. This model can simultaneously consider the trend, seasonality, and randomness of the data. By fitting the model and estimating the parameters, the trend term, seasonal term, and residual term are separated.

[0067] Step S24, combine the business scenario labels to correct the prediction results of the trend term, seasonal term, and residual term, and generate the computing power demand pattern of the user based on the time series.

[0068] In this embodiment, the business scenario labels contain the specific business background information of the user's use of computing power. Different business scenarios may have different impacts on the computing power demand. By combining the business scenario labels, the prediction results of the model can be adjusted and optimized to make it more in line with the actual situation.

[0069] As an alternative implementation, establish a mapping relationship table between the business scenario labels and the computing power demand correction coefficients. For each business scenario label, set the corresponding correction coefficient according to historical data and business experience. After obtaining the prediction results of the trend term, seasonal term, and residual term, look up the corresponding correction coefficient according to the business scenario label and perform weighted correction on the prediction results to finally generate the computing power demand pattern of the user based on the time series.

[0070] Through this embodiment, the computing power demand pattern of the user can be grasped more accurately. Mine the periodic characteristics from a large number of historical computing power usage records, then use the time series model to analyze the trend term, seasonal term, and residual term, and finally combine the business scenario labels for correction, making the prediction results more in line with the actual computing power demand of the user. This helps the computing power leasing management platform allocate computing power resources more reasonably and avoid resource waste or shortage caused by improper allocation of computing power resources. For example, reduce unnecessary computing power supply during the low-demand period of the user's computing power to reduce costs; make preparations for computing power reserves in advance during the peak-demand period to ensure the stability and reliability of the service, thereby improving the operation efficiency and user satisfaction of the platform.

[0071] Based on any of the above embodiments, in the fourth embodiment of the present application, referring to Figure 3 , the steps of determining the computing power analysis data associated with the task type according to the platform historical data include: Step S31, obtain all historical computing power request records in the platform historical data that match the task type.

[0072] In this embodiment, the platform historical data is all the data related to computing power usage accumulated by the computing power leasing management platform during its operation. The historical computing power request record is the detailed information of the computing power usage request initiated by the user to the platform, including task type, execution period, business scenario label, computing power consumption value, etc. The task type is the business category for which the user uses the computing power, such as natural language processing, video encoding, etc.

[0073] As an alternative implementation, the platform can set corresponding indexes in the database and use SQL query statements to filter out all historical computing power request records in the database whose task types are the same as the current task type.

[0074] Step S32, perform three-dimensional classification aggregation according to the task type, execution period, and business scenario label.

[0075] In this embodiment, the three-dimensional classification aggregation is to group the obtained historical computing power request records according to the three dimensions of task type, execution period, and business scenario label, and then perform an aggregation operation on the records within each group. The execution period refers to the specific time range during which the user uses the computing power, and the business scenario label is used to identify the specific business scenario in which the user uses the computing power, such as e-commerce promotion activities, online education course live broadcasts, etc.

[0076] As an alternative implementation, use the pandas library in Python to process the historical computing power request records, group them according to the task type, execution period, and business scenario label through the groupby() method, and then perform aggregation operations such as summing and counting on the computing power consumption values within the groups.

[0077] Step S33, calculate the computing power consumption characteristic values for each classification, including mean, variance, quantile, and resource utilization efficiency coefficient.

[0078] In this embodiment, the computing power consumption characteristic value is a statistical indicator used to describe the computing power consumption situation for each classification. The mean represents the average level of computing power consumption for this classification; the variance reflects the degree of dispersion of computing power consumption; the quantile can provide the numerical situation of computing power consumption at different positions; and the resource utilization efficiency coefficient measures the utilization efficiency of the computing power resources for this classification.

[0079] As an alternative implementation, use the relevant functions of the pandas library to calculate the mean, variance, quantile, etc. of the computing power consumption values for each classification, and at the same time obtain the resource utilization efficiency coefficient by calculating the ratio of the actual computing power consumption to the theoretical maximum computing power consumption.

[0080] Step S34, correct the computing power consumption characteristic values in combination with external data.

[0081] In this embodiment, external data refers to data related to computing power usage from outside the platform, such as market quotation data, industry dynamic data, weather data, etc. These external data may affect the computing power consumption. By combining external data, the computing power consumption characteristic value can more accurately reflect the actual situation.

[0082] As an alternative implementation, establish an association model between external data and the computing power consumption characteristic value, and adjust the computing power consumption characteristic value according to the change of external data. For example, if the market quotation data shows that a certain industry is in the peak season, it may increase the expected computing power consumption of the task types related to this industry, and thus adjust the corresponding computing power consumption characteristic value upward.

[0083] Step S35, construct an association model based on the random forest algorithm, analyze the dependence relationship between the task type and the computing power consumption characteristic value, and generate the computing power analysis data associated with the task type.

[0084] In this embodiment, the random forest algorithm is an ensemble learning algorithm. By constructing multiple decision trees and synthesizing their results, it can effectively handle high-dimensional data and complex non-linear relationships. The association model is used to analyze the internal connection between the task type and the computing power consumption characteristic value. The computing power analysis data associated with the task type is data generated by comprehensively considering the task type and the computing power consumption characteristic value, and can provide support for platform decision-making, such as reasonable computing power configuration suggestions for different task types in different business scenarios, etc.

[0085] As an alternative implementation, use the scikit-learn library in Python to construct a random forest model, take the task type and the computing power consumption characteristic value as input features, obtain the dependence relationship between the task type and the computing power consumption characteristic value through training the model, and then generate the computing power analysis data associated with the task type.

[0086] Through this embodiment, comprehensive and accurate computing power analysis data associated with the task type is generated. Select records matching the task type from a large amount of platform historical data, perform multi-dimensional classification aggregation and characteristic value calculation, then correct with external data, and finally establish an association model using the random forest algorithm, which can deeply explore the internal connection between the task type and the computing power consumption. This helps the computing power leasing management platform to more accurately understand the computing power demand characteristics of different task types in various business scenarios, so as to provide more reasonable computing power leasing plans for users, improve the allocation efficiency of computing power resources, reduce the operating costs of the platform, and at the same time enhance the user experience and the competitiveness of the platform in the market.

[0087] Based on any of the above embodiments, in the fifth embodiment of the present application, the step of calculating the first computing power requirement based on the task type, the required duration, the computing power specification, and the computing power demand pattern includes: Step S41: Decompose the computing power demand pattern into a trend term, a seasonal term, and a random term.

[0088] In this embodiment, the computing power demand pattern is obtained by analyzing the user's historical computing power usage records, reflecting the pattern of the user's computing power demand changing over time. The trend term represents the upward or downward trend of the computing power demand over a long period of time; the seasonal term reflects the repetitive fluctuation pattern of the computing power demand within a fixed period (such as daily, weekly, monthly); the random term includes irregular random changes that cannot be explained by the trend term and the seasonal term.

[0089] As an alternative implementation, use the seasonal decomposition algorithm (such as STL decomposition) in the statsmodels library of Python to decompose the computing power demand pattern data in the form of a time series into a trend term, a seasonal term, and a random term.

[0090] Step S42: Establish a mapping relationship between the task type and the standard computing power unit, and convert the computing power specification into a standardized computing power value.

[0091] In this embodiment, the task type represents the business category for which the user leases computing power, such as video rendering, big data analysis, etc. The standard computing power unit is a benchmark unit set to uniformly measure different computing power specifications. The standardized computing power value is to convert various different computing power specifications into values in units of the standard computing power unit according to the mapping relationship between the task type and the standard computing power unit, facilitating subsequent calculations.

[0092] As an alternative implementation, the platform pre-constructs a mapping table between the task type and the standard computing power unit. For example, for a video rendering task, it is stipulated that a specific model of GPU corresponds to a certain number of standard computing power units. When the computing power specification is obtained, it is converted into a standardized computing power value according to the mapping table.

[0093] Step S43: Use a linear regression model to perform a weighted combination of the standardized computing power value, the required duration, the trend term, and the seasonal term to determine the regression result.

[0094] In this embodiment, the linear regression model is a statistical model used to analyze the linear relationship between variables, and a linear equation is constructed to describe the relationship between the independent variable and the dependent variable. The standardized computing power value, the required duration, the trend term, and the seasonal term are independent variables, and the regression result is the predicted value of the dependent variable. The weighted combination means assigning different weights to each independent variable to reflect its influence degree on the dependent variable.

[0095] As an alternative implementation, a linear regression model is constructed using the scikit-learn library in Python. The normalized computing power value, required duration, trend term, and seasonal term are used as input features. Through model training, the weights of each feature are determined, and then the regression result is calculated.

[0096] Step S44: Determine the first computing power demand based on the time decay factor, the resource utilization efficiency coefficient of the user's historical computing power usage, and the regression result.

[0097] In this embodiment, the time decay factor reflects the change in computing power demand over time and is usually used to adjust the results of long-term predictions to make them more in line with reality. The resource utilization efficiency coefficient of the user's historical computing power usage reflects the efficiency of the user in the past when using computing power and is used to correct the regression result. The first computing power demand is a computing power value that meets the user's business needs calculated by combining the above factors.

[0098] As an alternative implementation, the first computing power demand is calculated using the formula "First computing power demand = regression result × time decay factor × resource utilization efficiency coefficient of the user's historical computing power usage".

[0099] Exemplarily, a short video production company plans to lease computing power for video rendering and submits a form to the computing power leasing management platform. The task type is video rendering, the required duration is 10 days, and the computing power specification is equipped with 4 RTX 3090 GPUs. The platform first obtains the computing power demand pattern of the company and decomposes it into a trend term, a seasonal term, and a random term using STL decomposition. Through a pre-constructed mapping table, 4 RTX 3090 GPUs are converted into a normalized computing power value. Then, the normalized computing power value, the 10-day required duration, and the decomposed trend term and seasonal term are used as inputs to train a linear regression model using the scikit-learn library to obtain the regression result. Considering the timeliness of the short video production business, the time decay factor is set to 0.95. At the same time, based on the company's historical computing power usage data, the resource utilization efficiency coefficient is calculated to be 0.85. Finally, the first computing power demand is calculated through the formula, providing a basis for the platform to reasonably allocate computing power for the company.

[0100] Through this embodiment, considering multiple factors comprehensively, the first computing power requirement is accurately calculated. By decomposing the law of computing power requirements, the changing trend of users' computing power requirements can be carefully grasped; standardizing the computing power specifications unifies the calculation standards; using a linear regression model for weighted combination fully considers the influence of various factors on computing power requirements; combining the time decay factor and the resource utilization efficiency coefficient further optimizes the calculation results. This enables the platform to allocate computing power resources for users more accurately, avoiding waste or insufficiency of computing power resources. On the one hand, it reduces the computing power rental cost for users, and on the other hand, it improves the utilization rate of the platform's computing power resources and enhances the market competitiveness of the platform.

[0101] Based on any of the above embodiments, in the sixth embodiment of the present application, the step of calculating the second computing power requirement based on the task type, the required duration, the computing power specification, and the computing power analysis data includes: Step S45, extracting the computing power consumption characteristic values matching the task type from the computing power analysis data, including the historical average computing power consumption, the time period fluctuation coefficient, and the resource utilization efficiency benchmark value.

[0102] In this embodiment, the computing power analysis data is a data set related to computing power consumption analyzed for different task types based on the platform's historical data. The historical average computing power consumption refers to the average computing power consumed when executing this task type in the past, reflecting the basic requirement level of this task type for computing power. The time period fluctuation coefficient reflects the fluctuation of computing power consumption of this task type in different time periods. For example, the computing power consumption of some tasks will increase significantly in specific time periods (such as peak hours on weekdays). The resource utilization efficiency benchmark value represents the theoretically reasonable computing power resource utilization efficiency when executing this task type. As an alternative implementation, the platform stores the computing power analysis data in the database and creates an index for each task type, and the computing power consumption characteristic values matching the current task type can be quickly extracted by querying the index.

[0103] Step S46, converting the computing power specification into the standard computing power in units based on the mapping relationship between the computing power specification and the hardware performance parameters.

[0104] In this embodiment, the computing power specification describes the computing power hardware configuration information requested by the user, such as the number of cores of the CPU, the model and quantity of the GPU, etc. The hardware performance parameters are quantitative indicators of the computing capabilities of different hardware devices. The standard computing power in units is a unified measurement standard used to convert different computing power specifications into comparable values. For example, it is stipulated that a certain specific model of CPU is used as the standard, and other hardware configurations are converted into standard computing power according to their performance ratios with the standard CPU.

[0105] As an alternative implementation, the platform pre - establishes a mapping table between computing power specifications and hardware performance parameters. When the computing power specifications of the user are obtained, by looking up the mapping table and performing corresponding calculations, it is converted into the standard computing power per unit.

[0106] Step S47, determine the time - period corrected performance according to the time - period fluctuation coefficient and the standard computing power per unit.

[0107] In this embodiment, the time - period corrected performance is a performance value obtained by adjusting the standard computing power per unit after considering the influence of time - period fluctuations on the computing power demand. Since the computing power demand for tasks may vary in different time periods, correcting the standard computing power per unit through the time - period fluctuation coefficient can more accurately reflect the computing power required to complete tasks in a specific time period.

[0108] As an alternative implementation, multiplying the standard computing power per unit by the time - period fluctuation coefficient can obtain the time - period corrected performance.

[0109] Step S48, determine the duration - efficiency product according to the required duration and the resource - utilization efficiency benchmark value.

[0110] In this embodiment, the required duration is the length of time that the user plans to use the computing power. The duration - efficiency product is the result of multiplying the required duration by the resource - utilization efficiency benchmark value, which comprehensively considers the influence of task execution time and resource - utilization efficiency on the computing power demand. For example, if the required duration is long and the resource - utilization efficiency is low, then the required computing power may increase accordingly.

[0111] As an alternative implementation, directly multiplying the required duration by the resource - utilization efficiency benchmark value can obtain the duration - efficiency product.

[0112] Step S49, based on the gradient - boosting tree, process the time - period corrected performance, the duration - efficiency product, and the historical average computing power consumption to determine the second computing power demand.

[0113] In this embodiment, the gradient - boosting tree is an ensemble learning algorithm. It iteratively trains multiple weak learners (decision trees) and combines them into a strong learner, which can well handle non - linear relationships and complex data. By using the time - period corrected performance, the duration - efficiency product, and the historical average computing power consumption as input features, and training and predicting using the gradient - boosting tree algorithm, the final second computing power demand is obtained.

[0114] As an alternative implementation, use the GradientBoostingRegressor class in the scikit - learn library of Python to construct a gradient - boosting tree model, train using the above three features as input, and then predict the second computing power demand through the trained model.

[0115] Exemplarily, an e-commerce platform plans to lease computing power during the "Double 11" promotion event for optimizing the product recommendation system, and the task type is the operation of the product recommendation algorithm. The required duration is 8 days from one week before "Double 11" to the day of "Double 11", and the computing power specification is equipped with 32-core CPUs and 8 NVIDIA Tesla V100 GPUs. The platform extracts from the computing power analysis data that the historical average computing power consumption of the product recommendation algorithm operation task type is 500 standard computing power units, and the period fluctuation coefficient is 1.5 during "Double 11" (indicating that the computing power demand in this period is 50% higher than usual), and the resource utilization efficiency benchmark value is 0.8. Through a pre-established mapping table, converting 32-core CPUs and 8 NVIDIA Tesla V100 GPUs into the standard computing power of units is 1200 standard computing power units. According to the period fluctuation coefficient and the unit standard computing power, the period-corrected performance is calculated as 1200×1.5 = 1800 standard computing power units. According to the required duration and the resource utilization efficiency benchmark value, the duration efficiency product is calculated as 8×0.8 = 6.4. Using the period-corrected performance 1800, the duration efficiency product 6.4, and the historical average computing power consumption 500 as inputs, a gradient boosting tree model is used for training and prediction, and finally the second computing power demand is obtained as 2000 standard computing power units. The platform reasonably allocates computing power resources for the e-commerce platform according to this result.

[0116] By calculating the second computing power demand in this embodiment, various factors such as the historical characteristics of the task type, period fluctuations, resource utilization efficiency, and computing power specifications are fully considered, and the gradient boosting tree algorithm is used to process these complex relationships, making the calculation results more accurate and reliable. This helps the computing power leasing management platform provide a more accurate computing power configuration plan for users, avoid resource waste or insufficiency caused by unreasonable computing power allocation. Reduce the computing power leasing cost of users, improve user satisfaction, and improve the utilization rate of the platform's computing power resources.

[0117] Based on any of the above embodiments, in the seventh embodiment of the present application, the steps of determining the computing power output in response to the computing power leasing form according to the first computing power demand, the second computing power demand, and the dynamic weight associated with the date include: Step S51, extract the date type label from the platform historical data and establish a mapping relationship between the date type and the business fluctuation coefficient.

[0118] In this embodiment, the platform historical data refers to all data related to the computing power leasing business accumulated by the computing power leasing management platform in the past operation process, covering information such as computing power demand and business execution status on different dates. The date type label is an identifier for classifying dates, such as weekdays, weekends, holidays, promotional days, etc. Different date types often correspond to different business busyness and computing power demand characteristics. The business volatility coefficient is a value that measures the degree of fluctuation of business computing power demand relative to the average level under a specific date type. The larger the volatility coefficient, the more drastic the change in business computing power demand under this date type.

[0119] As an optional implementation, the platform can use data mining technology to analyze and classify the date information in the historical data and extract different date type labels. Then, the business computing power demand data under each date type is counted, and its ratio to the average computing power demand is calculated to obtain the corresponding business fluctuation coefficient, and a mapping relationship between the two is established and stored in the database.

[0120] Step S52, determining the dynamic weight according to the mapping relationship between the date type and the business fluctuation coefficient, and the standard deviation of the historical computing power demand.

[0121] In this embodiment, the standard deviation of historical computing power demand is a measure of the degree of dispersion of historical computing power demand data, which reflects the fluctuation range of computing power demand at different time points. The dynamic weight is a weight value dynamically adjusted according to the date type and the fluctuation of historical computing power demand, which is used to reasonably distribute the influence of the first computing power demand and the second computing power demand when calculating the computing power output.

[0122] As an optional implementation, first obtain the business volatility coefficient of the date type from the mapping relationship according to the date corresponding to the current computing power rental form. Then, combined with the standard deviation of historical computing power requirements, the dynamic weight is determined by a preset calculation formula. For example, the formula: dynamic weight = business volatility coefficient / (business volatility coefficient + historical computing power demand standard deviation) can be used. In this way, on dates with large business fluctuations, more attention will be paid to the first computing power demand calculated according to historical rules; and on dates with small business fluctuations, relatively more attention will be paid to the second computing power demand calculated based on the general characteristics of the task type.

[0123] Step S53: determining the computing power output according to the first computing power requirement, the second computing power requirement and a weighted combination of the dynamic weights.

[0124] In this embodiment, the first computing power requirement is a computing power value calculated by comprehensively considering the user's own task type, required duration, computing power specification, and computing power demand pattern; the second computing power requirement is a value calculated based on the task type, required duration, computing power specification, and computing power analysis data associated with the task type. The weighted combination means multiplying the first computing power requirement and the second computing power requirement by their corresponding dynamic weights respectively, and then adding the two results to obtain the final computing power output.

[0125] As an alternative embodiment, the formula: computing power output = first computing power requirement × dynamic weight + second computing power requirement × (1 - dynamic weight) is used. In this way, according to the business characteristics of different dates, the proportion of the two computing power requirements can be flexibly adjusted, so as to more accurately determine the computing power resources provided for the user.

[0126] Exemplarily, user B plans to lease computing power during the May 1st Labor Day to cope with the possible increase in player traffic. The computing power lease form submitted by the platform shows that the first computing power requirement is 8,000 computing power units, and the second computing power requirement is 7,500 computing power units. The computing power lease management platform analyzes from the platform historical data that the May 1st Labor Day belongs to the holiday date type, and its corresponding business fluctuation coefficient is 1.2 (indicating that the business computing power demand on this date type is 20% higher than the average level). At the same time, by calculating the historical computing power demand data, the historical computing power demand standard deviation is obtained as 0.3. According to the above data, the dynamic weight is calculated using the formula: dynamic weight = 1.2 / (1.2 + 0.3) = 0.8. Finally, the computing power output is calculated according to the weighted combination formula: computing power output = 8,000 × 0.8 + 7,500 × (1 - 0.8) = 6,400 + 1,500 = 7,900 computing power units. The platform allocates the corresponding computing power resources to the online game platform according to this result.

[0127] By determining the computing power output in this embodiment, the fluctuation of the business under different date types and the stability of the historical computing power demand are fully considered. Using the dynamic weight to perform weighted combination on the first computing power requirement and the second computing power requirement can flexibly adjust the proportion of the two according to the actual situation, making the computing power output more in line with the actual needs of the business. This helps to avoid the situation of insufficient computing power on dates with large business fluctuations, ensuring the stability of the service and the user experience; at the same time, on dates with small business fluctuations, it can also avoid over-allocation of computing power, thereby reducing the computing power lease cost and improving the utilization efficiency of computing power resources.

[0128] Based on any of the above embodiments, in the eighth embodiment of the present application, after the step of determining the computing power output in response to the computing power lease form according to the first computing power requirement, the second computing power requirement, and the dynamic weight associated with the date, it includes: Step B10, collecting actual computing power output data.

[0129] In this embodiment, the actual computing power output data refers to the computing power-related data actually consumed or used during the process of providing computing power services to users. These data can reflect the real situation of users when using computing power, including real-time computing power consumption values, computing power usage trends in different time periods, etc.

[0130] As an alternative implementation, the platform can deploy sensors or monitoring software at the computing power output nodes to collect the running data of computing power devices in real time, such as CPU usage rate, GPU load, memory occupancy and other information. These data will be collected regularly (such as every second, every minute) and transmitted to the data center of the platform, and stored in the database for subsequent analysis and processing.

[0131] Step B20, output an alarm prompt message based on the computing power consumption threshold and the actual computing power output data.

[0132] In this embodiment, the computing power consumption threshold is one or more computing power consumption limit values preset by the platform, which are used to judge whether the current computing power usage situation is within the normal range. These thresholds can be set according to different task types, user requirements and the resource management strategy of the platform. For example, for some tasks with relatively stable computing power requirements, a fixed upper limit threshold can be set; for some tasks with fluctuating characteristics, multiple different levels of thresholds can be set to cope with different degrees of computing power fluctuations. The alarm prompt message is a prompt message sent by the platform to the administrator or user when the actual computing power output data exceeds or is lower than the set computing power consumption threshold. Its purpose is to remind relevant personnel to pay attention to the computing power usage situation and take measures to make adjustments in a timely manner.

[0133] As an alternative implementation, the platform can write a monitoring program to regularly read the actual computing power output data from the database and compare it with the preset computing power consumption threshold. If it is found that the actual computing power output data exceeds the threshold range, the program will trigger an alarm mechanism to send an alarm prompt message to relevant personnel by means of email, text message, system pop-up window, etc., and at the same time record the detailed information of the alarm event, such as the alarm time, alarm type, specific value exceeding the threshold, etc.

[0134] By collecting actual computing power output data and outputting warning prompt information based on the computing power consumption threshold, the platform can monitor the usage of computing power resources in real time and promptly detect abnormal computing power consumption behaviors. This helps the platform administrator to respond quickly and take corresponding measures to ensure the stability and reliability of the computing power service. For example, when the actual computing power output exceeds the threshold, the administrator can promptly check for issues such as resource abuse and system failures and perform corresponding processing to avoid service interruptions or performance degradation caused by insufficient computing power. At the same time, for users, the warning prompt information can help them understand their computing power usage, reasonably adjust their business requirements, and avoid unnecessary cost waste. In addition, by analyzing the warning data, the platform can further optimize the setting of the computing power consumption threshold to improve the efficiency and accuracy of resource management.

[0135] This application provides a computing power leasing management device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the computing power leasing management method in the first embodiment above.

[0136] The following refers to Figure 4 , which shows a schematic structural diagram of a computing power leasing management device suitable for implementing the embodiments of the present application. The computing power leasing management device in the embodiments of the present application may include, but is not limited to, mobile terminals such as laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistant), PADs (Portable Application Description), PMPs (Portable Media Player), in-vehicle terminals (such as in-vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 4 The shown computing power leasing management device is merely an example and should not impose any limitations on the functions and usage scope of the embodiments of the present application.

[0137] As Figure 4As shown, the computing power leasing management device may include a processing device 1001 (such as a central processing unit, a graphics processing unit, etc.), which may perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 1002 or the program loaded from the storage device 1003 into the random access memory (RAM) 1004. In the random access memory 1004, various programs and data required for the operation of the computing power leasing management device are also stored. The processing device 1001, the read-only memory 1002, and the random access memory 1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems may be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 may allow the computing power leasing management device to communicate with other devices wirelessly or wiredly to exchange data. Although the figure shows a computing power leasing management device with various systems, it should be understood that it is not required to implement or have all the systems shown. Instead, more or fewer systems may be implemented or had.

[0138] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes program codes for performing the methods shown in the flowcharts. In such an embodiment, the computer program may be downloaded and installed from the network through the communication device, or installed from the storage device 1003, or installed from the ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the methods of the embodiments disclosed in the present application are executed.

[0139] The computing power leasing management device provided by the present application adopts the computing power leasing management method in the above-mentioned embodiment, and can solve the technical problem that the computing power cost is too high due to the hardware procurement cost and subsequent operation and maintenance and power consumption. Compared with the prior art, the beneficial effects of the computing power leasing management device provided by the present application are the same as those of the computing power leasing management device provided by the above-mentioned embodiment, and other technical features in the computing power leasing management device are the same as the features disclosed in the method of the previous embodiment, and will not be elaborated here.

[0140] It should be understood that each part disclosed in this application can be implemented by hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in a suitable manner in any one or more embodiments or examples.

[0141] As mentioned above, the above are only specific embodiments of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application, and all should be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

[0142] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., computer programs) stored thereon, and the computer-readable program instructions are used to execute the computing power leasing management method in the above embodiments.

[0143] The computer-readable storage medium provided by this application can be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In this embodiment, the computer-readable storage medium can be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system, device, or device. The program code contained on the computer-readable storage medium can be transmitted by any appropriate medium, including but not limited to: wires, optical cables, radio frequency (RF), etc., or any suitable combination of the above.

[0144] The above computer-readable storage medium can be included in the computing power leasing management device; or it can exist separately without being assembled into the computing power leasing management device.

[0145] The above computer-readable storage medium carries one or more programs, which, when executed by the computing power leasing management device, cause the computing power leasing management device to: receive a computing power leasing form, determine the task type, required duration, and computing power specification corresponding to the computing power leasing form; determine the user's computing power demand pattern based on time series according to the user's historical computing power usage records corresponding to the computing power leasing form; determine the computing power analysis data associated with the task type according to the platform historical data; calculate a first computing power demand based on the task type, the required duration, the computing power specification, and the computing power demand pattern; and calculate a second computing power demand based on the task type, the required duration, the computing power specification, and the computing power analysis data; determine the computing power output in response to the computing power leasing form according to the first computing power demand, the second computing power demand, and the dynamic weight associated with the date.

[0146] Computer program code for performing the operations of the present application may be written in one or more programming languages or combinations thereof. The programming languages include object-oriented programming languages such as Java, Smalltalk, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code may execute entirely on the user's computer, partially on the user's computer, execute as a stand-alone software package, execute partially on the user's computer and partially on a remote computer, or execute entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0147] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagram may represent a module, a segment of a program, or a portion of code that contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than that marked in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, as well as combinations of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system that performs the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0148] The modules described in the embodiments of the present application can be implemented in software or in hardware. In some cases, the name of the module does not constitute a limitation on the unit itself.

[0149] The readable storage medium provided by the present application is a computer-readable storage medium that stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned computing power leasing management method, and can solve the technical problem of high computing power costs caused by hardware procurement costs and subsequent operation and maintenance and power consumption. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by the present application are the same as those of the computing power leasing management method provided by the above embodiments, and will not be elaborated here.

[0150] An embodiment of the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the steps of the computing power leasing management method as described above.

[0151] The computer program product provided by the present application can solve the technical problem of high computing power costs caused by hardware procurement costs and subsequent operation and maintenance and power consumption. Compared with the prior art, the beneficial effects of the computer program product provided by the embodiments of the present application are the same as those of the computing power leasing management method provided by the above embodiments, and will not be elaborated here.

[0152] The above are only the preferred embodiments of the present application, and do not limit the patent scope of the present application. Any equivalent structural or equivalent process transformation made using the specification and drawings of the present application, or directly or indirectly applied in other related technical fields, shall be equally included in the patent scope of the present application.

Claims

1. A computing power leasing management method, characterized in that, Applied to a computing power leasing management platform, the computing power leasing management method includes: Receiving a computing power leasing form and determining the task type, required duration, and computing power specification corresponding to the computing power leasing form; Based on the computing power leasing form, applying the user's historical computing power usage records to determine the user's computing power demand pattern based on time series; Determining the computing power analysis data associated with the task type according to the platform historical data; Calculating a first computing power demand based on the task type, the required duration, the computing power specification, and the computing power demand pattern; and calculating a second computing power demand based on the task type, the required duration, the computing power specification, and the computing power analysis data; Determining the computing power output in response to the computing power leasing form according to the first computing power demand, the second computing power demand, and the dynamic weight associated with the date.

2. The method according to claim 1, wherein Before the step of receiving the computing power leasing form and determining the task type, required duration, and computing power specification corresponding to the computing power leasing form, it includes: In response to an input operation of a form submission control, determining the task type; Rendering the demand details control associated with the task type, and different task types have different corresponding demand details controls; In response to an input operation of the demand details control, parsing out the required duration and the computing power specification; Generating the computing power leasing form according to the task type, the required duration, and the computing power specification.

3. The method according to claim 1, characterized in that, The step of, based on the computing power leasing form, applying the user's historical computing power usage records to determine the user's computing power demand pattern based on time series, includes: Obtaining the user's historical computing power usage records, where the historical computing power usage records include the execution period, the computing power consumption value, and the business scenario label; Performing bucket aggregation on the historical computing power usage records based on a preset time window to generate periodic computing power consumption characteristics; Modeling the periodic computing power consumption characteristics to identify the trend term, seasonal term, and residual term of the user's computing power demand; Combining the business scenario label to correct the prediction results of the trend term, seasonal term, and residual term to generate the user's computing power demand pattern based on time series.

4. The method according to claim 1, characterized in that, The step of determining the computing power analysis data associated with the task type according to the platform historical data, includes: Obtaining all historical computing power request records in the platform historical data that match the task type; Performing three-dimensional classification aggregation according to the task type, execution period, and business scenario label; Calculating the computing power consumption characteristic values for each classification, including the mean, variance, quantile, and resource utilization efficiency coefficient; Correcting the computing power consumption characteristic values in combination with external data; Constructing an association model based on the random forest algorithm to analyze the dependence relationship between the task type and the computing power consumption characteristic values, and generating the computing power analysis data associated with the task type.

5. The method according to claim 1, characterized in that, The step of calculating the first computing power demand based on the task type, the required duration, the computing power specification, and the computing power demand pattern, includes: Decomposing the computing power demand pattern into a trend term, a seasonal term, and a random term; Establishing a mapping relationship between the task type and the standard computing power unit, and converting the computing power specification into a standardized computing power value; Using a linear regression model, the standardized computing power value, the required duration, the trend term, and the seasonal term are weighted and combined to determine the regression result; Based on the time decay factor, the resource utilization efficiency coefficient of the user's historical computing power usage, and the regression result, the first computing power demand is determined.

6. The method according to claim 1, characterized in that, The step of calculating the second computing power demand based on the task type, the required duration, the computing power specification, and the computing power analysis data includes: Extracting computing power consumption characteristic values matching the task type from the computing power analysis data, including historical average computing power consumption, period fluctuation coefficient, and resource utilization efficiency benchmark value; Based on the mapping relationship between the computing power specification and the hardware performance parameters, the computing power specification is converted into the standard computing power per unit; Determining the period correction performance according to the period fluctuation coefficient and the standard computing power per unit; Determining the duration efficiency product according to the required duration and the resource utilization efficiency benchmark value; Based on the gradient boosting tree, processing the period correction performance, the duration efficiency product, and the historical average computing power consumption to determine the second computing power demand.

7. The method according to claim 1, wherein The step of determining the computing power output in response to the computing power lease form according to the first computing power demand, the second computing power demand, and the dynamic weight associated with the date includes: Extracting the date type label from the platform historical data and establishing the mapping relationship between the date type and the business fluctuation coefficient; Determining the dynamic weight according to the mapping relationship between the date type and the business fluctuation coefficient and the historical computing power demand standard deviation; Determining the computing power output according to the weighted combination of the first computing power demand, the second computing power demand, and the dynamic weight.

8. The method according to claim 1, characterized in that After the step of determining the computing power output in response to the computing power lease form according to the first computing power demand, the second computing power demand, and the dynamic weight associated with the date, it includes: Collecting actual computing power output data; Outputting an alarm prompt message based on the computing power consumption threshold and the actual computing power output data.

9. A computing power leasing management device, characterized in that, The computing power lease management device includes: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the computer program is configured to implement the steps of the computing power lease management method according to any one of claims 1 to 8.

10. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the computer-readable storage medium. When the computer program is executed by the processor, it implements the steps of the computing power lease management method according to any one of claims 1 to 8.

Images