Server capacity regulation method and device of performance system and computer readable medium

Abstract

The application discloses a server capacity regulation method, equipment and computer storage medium of a performance system, wherein the server capacity regulation method comprises the following steps: obtaining server historical data to generate a historical data set, the historical data set comprises historical data samples of several time periods, and the historical data samples comprise a time sequence of server task quantity in a unit time; in a first time stage, a task quantity prediction model is trained by using the historical data set, a predicted task quantity time sequence in a to-be-predicted period is generated based on the task quantity prediction model, and the server capacity is regulated by using the predicted task quantity time sequence. The task quantity prediction model is trained based on the historical data samples of the time sequence, the predicted task quantity time sequence in the to-be-predicted period is generated, the server capacity is regulated in advance, and the running stability of the server is ensured.

Description

Technical Field

[0001] This invention relates to the field of fulfillment system technology, and specifically to a server capacity control method, device, and computer-readable medium for a fulfillment system. Background Technology

[0002] In high-concurrency scenarios of promotional activities on e-commerce platforms, it is necessary to ensure that the server resource configuration of the fulfillment system matches the task volume (order volume, access volume, etc.) to prevent problems such as excessive system load or low server utilization during peak task periods.

[0003] In related technologies, server capacity configuration is typically achieved using methods such as static load testing, passive expansion based on CPU utilization threshold detection, or prediction using historical data averages. These methods adjust server capacity based on real-time load pressure, expanding capacity only when the server exceeds its limits. This results in a lag, an inability to respond in advance to temporal fluctuations, and a lack of responsiveness to sudden increases during events. They are poorly adaptable to fulfillment systems with periodic temporal fluctuations and cannot effectively solve problems such as excessively high operating loads or low server utilization in fulfillment systems. Summary of the Invention

[0004] The present invention aims to address, to a certain extent, one of the technical problems in the prior art. To this end, the present invention provides a server capacity control method, device, and computer-readable medium for a fulfillment system, solving the problems of excessive operating load or low server utilization in the fulfillment system.

[0005] In a first aspect, to achieve the above objectives, the present invention provides a server capacity control method for a fulfillment system, comprising:

[0006] Historical data from the server is acquired to generate a historical dataset, which includes historical data samples from several time periods, and the historical data samples include a time series of server task volume per unit time period.

[0007] In the first time phase, a task volume prediction model is trained using the historical dataset, and a predicted task volume time series for the period to be predicted is generated based on the task volume prediction model. The first time phase is a period of time before the period to be predicted.

[0008] Based on the predicted task volume time series, generate peak task volume sub-periods and peak task volume within the predicted period;

[0009] The server capacity is adjusted based on the predicted task volume time series.

[0010] In the second time phase, the CPU utilization and task volume of the server are acquired in real time, and the server capacity is adjusted a second time based on the acquired CPU utilization and task volume. The second time phase is a period between the first time phase and the peak sub-period of the period to be predicted.

[0011] In this technical solution, a prediction model is trained in the first time stage, and the peak task volume sub-period and corresponding peak task volume are generated based on the prediction model within the time period to be predicted. In the second time stage before the peak task volume sub-period arrives within the time period to be predicted, the prediction results generated in the first time stage are optimized and adjusted based on the real-time CPU utilization. The two-stage prediction and adjustment improve the accuracy of the prediction results and the server's ability to handle emergencies.

[0012] Preferably, adjusting the server capacity based on the predicted task volume time series further includes:

[0013] The server baseline capacity is deployed in advance based on the peak task volume.

[0014] Preferably, generating peak task sub-periods and peak task values ​​within the predicted period based on the task volume time series further includes:

[0015] A time-based task volume regression model is constructed using the predicted task volume time series.

[0016] Based on the task volume regression model, the peak task volume sub-periods and peak task volume within the period to be predicted are generated.

[0017] Preferably, deploying the server baseline capacity in advance based on the peak task volume includes:

[0018] Obtain the single-device task processing speed of the server below the safety level, where the safety level is the CPU utilization threshold that the server can withstand.

[0019] The number of server devices is deployed based on the peak task volume and the single-device task processing speed.

[0020] Preferably, the process of obtaining historical data from the server to generate a historical dataset also includes:

[0021] Retrieve the activity status identifier of historical server data;

[0022] The elements in the historical data sample are assigned an activity status identifier.

[0023] Preferably, in the first time phase, the historical dataset is filtered by the activity status identifier, and the task volume prediction model is trained and optimized using the filtered historical dataset.

[0024] Preferably, generating the predicted task volume time series within the predicted period based on the task volume prediction model further includes obtaining the activity status identifier within the predicted period and inputting it into the task volume prediction model to obtain the predicted task volume time series within the predicted period.

[0025] Preferably, the process of obtaining historical data from the server to generate a historical dataset also includes:

[0026] The time features of the period to be predicted are obtained, and the server historical data are filtered using the time features to generate a historical dataset containing the same time features as the period to be predicted; the time features are used to characterize the time period corresponding to the server historical data.

[0027] Preferably, the specific steps for training a task volume prediction model using the historical dataset and generating a predicted task volume time series for the period to be predicted based on the task volume prediction model include:

[0028] S201. Arrange historical data samples in the historical dataset in chronological order to form a time series. Set up a sliding window with k unit time lengths. The historical data obtained by each sliding window is used as one input, and the task volume data of the next unit time corresponding to the sliding window is used as the corresponding label.

[0029] S202. Input the input and corresponding labels into the task volume prediction model (LSTM model) for training;

[0030] S203. Obtain the latest task volume for the next k consecutive time units as input, input it into the trained task volume prediction model, and the task volume prediction model outputs the task volume prediction value for the next 1 time unit.

[0031] S204. Adopt a rolling prediction strategy, append the previously obtained task quantity prediction value to the end of the previously input sequence, remove the task quantity data of the first unit time in the input and use it as the new input, input the task quantity prediction model again, and complete m predictions in a rolling manner.

[0032] S205. Obtain m prediction results and arrange them in chronological order to generate a time series of prediction task volume.

[0033] Preferably, the secondary adjustment of the server capacity based on real-time acquired CPU utilization and task volume per unit time includes:

[0034] The correlation between CPU utilization, task volume per unit time, and number of server devices in historical server data is obtained.

[0035] A safety water level regression model is generated based on the correlation between CPU utilization, task volume per unit time, and number of server devices in historical data.

[0036] The real-time CPU utilization and task volume per unit time are input into the safety level regression model, and the server capacity is adjusted a second time based on the output results; the output results are the number of server devices required to meet the current task volume when the CPU utilization is below the safety level.

[0037] Secondly, this application also provides an electronic device, including a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to implement the server capacity control method as described in any of the above technical solutions.

[0038] Thirdly, this application also provides a computer-readable medium storing computer instructions, which, when executed by a processor, implement the server capacity control method as described in any of the above technical solutions.

[0039] These features and advantages of the present invention will be disclosed in detail in the following specific embodiments and accompanying drawings. The embodiments or means of the present invention will be illustrated in detail with reference to the accompanying drawings, but are not intended to limit the technical solutions of the present invention. Furthermore, each of these features, elements, and components appearing in the following text and drawings is a plurality, and different symbols or numbers are used for convenience of representation, but all represent parts with the same or similar construction or function. Attached Figure Description

[0040] The present invention will be further described below with reference to the accompanying drawings:

[0041] Figure 1 This embodiment presents a flowchart of a server capacity control method for a fulfillment system.

[0042] Figure 2 This is a flowchart of another server capacity control method for a fulfillment system in this embodiment;

[0043] Figure 3 This is a schematic diagram of the structure of an electronic device according to this embodiment. Detailed Implementation

[0044] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described are intended to explain the present invention and should not be construed as limiting the invention.

[0045] The terms "an embodiment," "example," or "example" used in this specification refer to a particular feature, structure, or characteristic described in connection with the embodiment itself that may be included in at least one embodiment disclosed in this patent. The phrase "in some embodiments" appearing in various places in the specification does not necessarily refer to the same embodiment.

[0046] In the description of this application, it should be understood that the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances. Furthermore, in the description of this application, unless otherwise stated, "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship.

[0047] In the embodiments of this application, the fulfillment system is mainly used to manage the entire process of logistics orders on an instant e-commerce platform, including but not limited to order acceptance, scheduling, delivery, and exception handling. In such fulfillment systems, the order volume, access request volume, and other task volumes exhibit obvious periodic fluctuations while also possessing activity characteristics. That is, they show a relatively consistent fluctuation trend within each time period, and the task volume increases significantly during promotional activities, flash sales, and other scenarios. If server capacity cannot be effectively configured, it may cause system crashes during peak task periods and waste server resources during off-peak task periods.

[0048] The application scenario in this embodiment includes a service platform, users, service providers, and riders. The service platform integrates services offered by multiple service providers, displaying these services as service products to users. Users can then select and order service products online, and riders are dispatched to complete the delivery after the user selects a specific service product. In this embodiment, the server is configured on the service platform. The server capacity of the service platform needs to be configured based on task volume data such as user access requests, rider access requests, and order receipts. The server capacity control method of the fulfillment system shown in this embodiment can be adjusted based on the sum of various task volumes, or multiple collaborative fulfillment subsystems can each complete a single task. In this case, each fulfillment subsystem is controlled based on its single task volume, and the fulfillment subsystem completing a single task is considered the fulfillment system of this embodiment.

[0049] like Figure 1 As shown, this embodiment provides a server capacity control method for a fulfillment system, including:

[0050] S100. Obtain historical server data to generate a historical dataset. The historical dataset includes historical data samples for several time periods. The historical data samples include a time series of server task volume within a unit time period. Specifically, the historical data samples can be time series with fixed time intervals or non-fixed time intervals. Each element in the time series can be represented by the average amount of tasks completed within a unit time period or by the total amount of tasks within the time interval.

[0051] S200. In the first time phase, a task volume prediction model is trained using the historical dataset, and a predicted task volume time series for the period to be predicted is generated based on the task volume prediction model. The first time phase is a period of time several hours before the period to be predicted. Specifically, the task volume prediction model can be trained using statistical or machine learning models such as Autoregressive Integrated Moving Average (ARIMA) or Long Short-Term Memory (LSTM). The first time phase is several hours before the period to be predicted. For example, if the period to be predicted is Saturday of a certain week, then the first time phase is several days before Saturday.

[0052] S300. Adjust the server capacity based on the predicted task volume time series; specifically, the predicted task volume time series includes the predicted task volume of several time nodes or time periods, and adjust the server capacity before the time point or time period corresponding to each element to ensure the server capacity in advance.

[0053] S400. In the second time phase, the CPU utilization and task volume per unit time of the server are acquired in real time. Based on the acquired CPU utilization and task volume per unit time, the server capacity is adjusted a second time. The second time phase is a period between the first time phase and the peak sub-period of the period to be predicted. It should be noted that in this embodiment, the second time phase is between the first time phase and the period to be predicted, or the second time phase is before the peak period of the period to be predicted. The task volume and peak period of the period to be predicted are predicted in the first time phase. The period before the peak period within the period to be predicted is set as the second time phase. For example, if the period to be predicted is the day of an event, then the first time phase is the day before the day of the event. Assuming the predicted peak period is from 11:00 to 12:00 noon on the day of the event, then the second time phase can be set as a period before the peak period of the day of the event. In this embodiment, the time interval between the second time phase and the period to be predicted (the peak period for which server capacity adjustment is required) is less than the time interval between the first time phase and the period to be predicted. By acquiring CPU utilization and task volume data in real time before the peak period on the day of the event, the accuracy of the first-stage prediction results can be further verified. If a sudden event occurs at this time, the accuracy of the first-stage prediction results can be verified, and secondary adjustments can be made to ensure the stability of the server operation.

[0054] In this embodiment, a time series of server task volume within a unit of time is used as a historical data sample to collect the task volume change trend of historical data. For example, the order volume of the food delivery e-commerce platform changes over time in a day. The task volume prediction model is trained using historical data containing periodic change patterns, so that the output of the prediction model conforms to the periodic change pattern, ensuring the accuracy of the prediction results. At the same time, the prediction model can output a time series of predicted task volume, so that the system can configure server capacity in advance to cope with the peak task volume period within the cycle and ensure the stability of server operation.

[0055] In some embodiments, the specific steps of training a task volume prediction model using the historical dataset and generating a predicted task volume time series for the period to be predicted based on the task volume prediction model include:

[0056] S201. Arrange historical data samples in the historical dataset in chronological order to form a time series. Set up a sliding window containing k unit time lengths (e.g., k=24, the task volume data for each hour within 24 hours). The historical data obtained by each sliding window is used as one input, and the task volume data of the next unit time (the task volume data of the next hour) corresponding to the sliding window is used as the corresponding label.

[0057] S202. Input the input and corresponding labels into a Long Short-Term Memory (LSTM) network model for training; the model learns the variation pattern of historical task volume under daily, weekly and other cycles by minimizing the mean square error (MSE) between the prediction and the true value through backpropagation and Adam optimizer.

[0058] S203. Obtain the latest task volume for the next k consecutive units of time as input, input it into the trained task volume prediction model, and the task volume prediction model outputs the task volume prediction value for the next unit of time.

[0059] S204. Adopt a rolling prediction strategy, append the previously obtained task quantity prediction value to the end of the previously input sequence, remove the task quantity data of the first unit time in the input and use it as the new input, input the task quantity prediction model again, and complete m predictions in a rolling manner.

[0060] S205. Obtain m prediction results and arrange them in chronological order to generate a time series of prediction task volume.

[0061] In some embodiments, adjusting the server capacity based on the predicted task volume time series further includes:

[0062] S301. Generate peak sub-periods and peak values ​​of task volume within the period to be predicted based on the predicted task volume time series.

[0063] S302. Deploy the server baseline capacity in advance based on the peak task volume.

[0064] In this embodiment, the peak period of task volume and the level of task volume during the prediction period are determined by predicting the task volume time series. Before the peak period arrives, the server's baseline capacity is expanded according to the predicted task volume level. After the peak period ends, the server's baseline capacity is adjusted back to the normal level. This ensures stable server operation during the peak period while improving CPU utilization at other times and avoiding resource waste.

[0065] Specifically, generating peak sub-periods and peak task volumes within the predicted period based on the task volume time series also includes:

[0066] S3011. Construct a time-based task volume regression model using the predicted task volume time series;

[0067] S3012. Based on the task volume regression model, generate the peak task volume sub-period and peak task volume within the period to be predicted.

[0068] For example, the predicted task volume time series is a discrete correspondence between the predicted task volume (average task volume per unit time) and time, while the task volume regression model is a continuous linear correspondence between the predicted task volume and time. That is, the task volume regression model is a linear regression model. The task volume regression model generates the peak sub-periods and peak values ​​of the task volume within the predicted period. Fitting the discrete time series to a linear regression model can improve the accuracy of the prediction. For example, if the predicted period is a future day, then the peak sub-period is one hour or half an hour. It should be noted that a predicted period may contain one or more peak sub-periods of the task volume, which can be determined based on the task volume threshold.

[0069] In some embodiments, pre-deploying the server baseline capacity based on the peak task volume includes:

[0070] S3021. Obtain the single-device task processing speed of the server below the safety level, where the safety level is the CPU utilization threshold that the server can withstand. For example, if the safety level is set to a CPU utilization of 35%, then when the server's CPU utilization is below 35%, the server is operating below the safety level; when the server's CPU utilization is above 35%, the server is operating above the safety level, meaning the service is overloaded, which may result in lag, data loss, response delays, etc. It should be noted that the specific value of the safety level can be set according to requirements and the highest number of access tasks. For example, the CPU utilization of the safety level can be set to 35% on activity days, 50% on holidays, and 60% on ordinary days. In some embodiments, the value of the safety level is set and bound differently based on the activity status identifier, meaning that the CPU utilization value corresponding to the safety level can be obtained by reading the activity status identifier of the current activity.

[0071] S3022. Deploy the number of server devices based on the peak task volume and the single-device task processing speed. In the fulfillment system, the server includes multiple single devices working simultaneously. The task volume of the server is the sum of the task volumes of each single device. The number of server devices required to be deployed at the peak task volume can be calculated by using the peak task volume and the single-device task processing speed, so that the corresponding number of server devices can be started before the peak task volume.

[0072] In some embodiments, obtaining historical server data to generate a historical dataset further includes:

[0073] Retrieve the activity status identifier of historical server data;

[0074] Assign an activity status identifier to each element in the historical data sample;

[0075] In the first time phase, the historical dataset is filtered by the activity status identifier, and the task volume prediction model is trained and optimized using the filtered historical dataset. In this embodiment, the first time phase is a period of time before the period to be predicted, for example, the day before the period to be predicted.

[0076] In the fulfillment system of this embodiment, the activity status has a significant impact on the server workload. For example, the server workload will increase significantly during daily promotions and flash sales. Therefore, by obtaining the activity status identifier of the server's historical data, it is possible to filter out historical data under the same or similar activities. By training the workload prediction model with historical data samples that have the same activities as the time period to be predicted, the prediction accuracy of the prediction model can be further improved.

[0077] In some embodiments, generating a predicted task volume time series for a given period based on the task volume prediction model further includes obtaining activity status identifiers for the given period and inputting them into the task volume prediction model to obtain the predicted task volume time series for the given period. When different activities are scheduled to take place within the given period, the task volume prediction model can output different prediction results. For example, when predicting tasks on an activity day, the task volume prediction model can output prediction results that include the influence of activity factors based on historical data of the same or similar activities.

[0078] In some embodiments, obtaining historical data from the server to generate a historical dataset further includes:

[0079] The temporal characteristics of the period to be predicted are obtained, and historical server data is filtered using these characteristics to generate a historical dataset containing historical server data with the same temporal characteristics as the period to be predicted. The temporal characteristics characterize the time period corresponding to the historical server data. In the fulfillment system of this embodiment, temporal characteristics have a significant impact on the system's workload. For example, the workload changes significantly on a particular day of the week, or during a particular season or month of the year. The temporal characteristics described in this embodiment include at least one of the following: season, month, week, hour, and minute. When predicting the workload for a period with temporal characteristics, if summer is used as the temporal characteristic, historical server data that is also from summer is filtered to improve the accuracy of the workload prediction model.

[0080] Specifically, the method for adjusting the server capacity based on real-time CPU utilization and task volume per unit time includes:

[0081] S401. Obtain the correlation between CPU utilization, task volume per unit time, and number of server devices in historical server data.

[0082] S402. Generate a safety water level regression model based on the correlation between CPU utilization, task volume per unit time and number of server devices in historical data.

[0083] S403. Input the real-time CPU utilization and the workload per unit time into the safety level regression model, and adjust the server capacity a second time based on the output result; the output result is the number of server devices required to meet the current workload when the CPU utilization is below the safety level.

[0084] In this technical solution, a safe water level regression model is constructed based on CPU utilization, task volume per unit time, and number of server devices to predict the short-term trend of server task volume. When the CPU utilization of the server is higher than the safe water level, the peak number of server devices that need to be expanded is calculated based on the short-term trend prediction of the peak task volume. Then, the server capacity is adjusted a second time based on the calculation results.

[0085] like Figure 2 As shown, this embodiment provides a method for server capacity control in a fulfillment system:

[0086] S501. Obtain historical server data to generate a historical dataset. The historical dataset includes historical data samples for several time periods. The historical data samples include time series of server task volume and CPU utilization within a unit of time. For example, obtain the time series of task volume and CPU utilization for the past 90 days.

[0087] S502. Obtain the activity status identifier and time feature of historical data from the server, and filter the historical dataset using the activity status identifier and time feature; for example, the activity status identifier is an activity tag, and the time feature is Saturday;

[0088] S503. In the first time period, a task volume prediction model is trained using the selected historical dataset, and a predicted task volume time series for the period to be predicted is generated based on the task volume prediction model.

[0089] S504. Construct a time-based task volume regression model using the predicted task volume time series;

[0090] S505. Based on the task volume regression model, generate the peak task volume sub-period and peak task volume within the period to be predicted.

[0091] S506. Obtain the single-device task processing speed of the server below the safety level, where the safety level is the CPU utilization threshold that the server can withstand.

[0092] S507. Deploy the number of server devices based on the peak task volume and the single device task processing speed;

[0093] S508. In the second time phase, the CPU utilization and task volume of the server are obtained in real time.

[0094] S509. Obtain the correlation between CPU utilization, task volume per unit time, and number of server devices in historical server data.

[0095] S510. Generate a safety water level regression model based on the correlation between CPU utilization, task volume per unit time and number of server devices in historical data.

[0096] S511. Input the real-time CPU utilization and the workload per unit time into the safety level regression model, and adjust the server capacity a second time based on the output result; the output result is the number of server devices required to meet the current workload when the CPU utilization is below the safety level.

[0097] This embodiment implements a two-stage server capacity control through a first time stage and a second time stage. In the first stage, historical datasets are filtered using activity tag features and time features, and the filtered datasets are used to train the model to improve the model's prediction accuracy. The task volume prediction model and the safety water level regression model can predict the trend of task volume changes in the future, so as to expand capacity in advance before the peak of task volume, avoid the server's CPU utilization from exceeding the safety water level, and ensure the stability of system operation.

[0098] Secondly, such as Figure 3 As shown, this application also provides an electronic device, including a processor 101 and a memory 102. The memory 102 stores a computer program. When the processor 101 executes the computer program, it implements the server capacity control method as described in any of the above technical embodiments. The electronic device may also include one or more I / O interfaces 103 connected between the processor 101 and the memory 102, configured to enable information interaction between the processor 101 and the memory 102.

[0099] Thirdly, this application also provides a computer-readable medium storing computer instructions, which, when executed by a processor, implement the server capacity control method as described in any of the above embodiments.

[0100] It should be noted that the user information (including but not limited to user account information, user identity information, etc.) and personal data (including but not limited to personal email data, personal folder data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0101] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer storage medium or transmitted through the computer storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., digital versatile discs (DVDs)), or semiconductor media (e.g., solid-state drives (SSDs)).

[0102] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. The aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks. Unless otherwise specified, the technical features of this embodiment and its implementation can be combined arbitrarily.

[0103] The embodiments described above are merely preferred embodiments of this application and are not intended to limit the scope of this application. Any modifications and improvements made by those skilled in the art to the technical solutions of this application without departing from the spirit of this application should fall within the protection scope defined by the claims.

[0104] The foregoing has described specific embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A method for server capacity control in a fulfillment system, characterized in that, include: Historical data from the server is acquired to generate a historical dataset, which includes historical data samples from several time periods, and the historical data samples include a time series of server task volume per unit time period. In the first time phase, a task volume prediction model is trained using the historical dataset, and a predicted task volume time series for the period to be predicted is generated based on the task volume prediction model. The first time phase is a period of time before the period to be predicted. Based on the predicted task volume time series, generate peak task volume sub-periods and peak task volume within the predicted period; The server capacity is adjusted based on the predicted task volume time series. In the second time phase, the CPU utilization and task volume of the server are obtained in real time. The second time phase is a period between the first time phase and the peak sub-period of the period to be predicted. The correlation between CPU utilization, task volume per unit time, and number of server devices in historical server data is obtained. A safety water level regression model is generated based on the correlation between CPU utilization, task volume per unit time, and number of server devices in historical data. The real-time CPU utilization and task volume per unit time are input into the safety water level regression model, and the server capacity is adjusted a second time based on the output results. The output results are the number of server devices required to meet the current task volume when the CPU utilization is below the safety water level. The safety water level is the CPU utilization threshold that the server can withstand.

2. The server capacity control method for the fulfillment system according to claim 1, characterized in that, Adjusting the server capacity based on the predicted task volume time series includes: Based on the peak task volume, deploy the server's baseline capacity in advance; Based on the task volume time series, generating the peak sub-periods and peak values ​​of task volume within the period to be predicted also includes: A time-based task volume regression model is constructed using the predicted task volume time series. Based on the task volume regression model, the peak task volume sub-periods and peak task volume within the period to be predicted are generated.

3. The server capacity control method for the fulfillment system according to claim 2, characterized in that, Based on the peak task volume, the server baseline capacity is deployed in advance, including: Obtain the single-device task processing speed of the server below the safe water level; The number of server devices is deployed based on the peak task volume and the single-device task processing speed.

4. The server capacity control method for the fulfillment system according to claim 1, characterized in that, Obtaining historical server data to generate historical datasets also includes: Retrieve the activity status identifier of historical server data; Assign an activity status identifier to each element in the historical data sample; In the first time phase, the historical dataset is filtered by the activity status identifier, and the task volume prediction model is trained and optimized using the filtered historical dataset.

5. The server capacity control method for the fulfillment system according to claim 4, characterized in that, The process of generating a predicted task volume time series for the period to be predicted based on the task volume prediction model also includes obtaining the activity status identifier for the period to be predicted and inputting it into the task volume prediction model to obtain the predicted task volume time series for the period to be predicted.

6. The server capacity control method for the fulfillment system according to claim 1, characterized in that, Obtaining historical server data to generate historical datasets also includes: The time features of the period to be predicted are obtained, and the server historical data are filtered using the time features to generate a historical dataset containing the same time features as the period to be predicted; the time features are used to characterize the time period corresponding to the server historical data.

7. The server capacity control method for a fulfillment system according to any one of claims 1 to 6, characterized in that, The specific steps for training a task volume prediction model using the historical dataset and generating a predicted task volume time series for the period to be predicted based on the task volume prediction model include: S201. Arrange historical data samples in the historical dataset in chronological order to form a time series. Set up a sliding window with k unit time lengths. The historical data obtained by each sliding window is used as one input, and the task volume data of the next unit time corresponding to the sliding window is used as the corresponding label. S202. Train the task volume prediction model by inputting the input and the corresponding label; S203. Obtain the latest task volume for the next k consecutive time units as input, input it into the trained task volume prediction model, and the task volume prediction model outputs the task volume prediction value for the next 1 time unit. S204. Adopt a rolling prediction strategy, append the previously obtained task quantity prediction value to the end of the previously input sequence, remove the task quantity data of the first unit time in the input and use it as the new input, input the task quantity prediction model again, and complete m predictions in a rolling manner. S205. Obtain m prediction results and arrange them in chronological order to generate a time series of prediction task volume.

8. An electronic device, characterized in that, It includes a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the server capacity control method as described in any one of claims 1 to 7.

9. A computer-readable medium, characterized in that, The computer-readable medium stores computer instructions, which, when executed by a processor, implement the server capacity control method according to any one of claims 1 to 7.

Images