Method and device for controlling abnormal use of computing power of intelligent computing center cloud platform

Abstract

The invention provides a computing power abnormal use control method and device for an intelligent computing center cloud platform, and relates to the technical field of intelligent computing centers, intelligent computing centers and computing power infrastructures, and the method comprises the steps: S1, obtaining historical computing power use amount data of a user in a first period; s2, inputting the historical computing power consumption data into a preset time sequence model, and predicting to obtain computing power demand information of the user in the next period of the first period; s3, dividing the computing power budget of the user into a first computing power budget and a second computing power budget according to the computing power demand quantity information and a preset user strategy; the first computing power budget is greater than or equal to the minimum computing power demand of the user in operation within a preset range, and the second computing power budget is used for responding to the elastic computing power demand of the user in operation outside the preset range; and S4, under the condition that the computing power demand of the user is greater than or equal to at least one of the first computing power budget and the second computing power budget, executing at least one of flow limiting processing and alarm processing on the operation of the user.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical fields of intelligent computing centers, smart computing centers and computing power infrastructure, and specifically to a method and device for controlling abnormal use of computing power on an intelligent computing center cloud platform. Background Art

[0002] With the rapid development of artificial intelligence technology, "intelligent computing centers" and "intelligent computing centers" have emerged.

[0003] An "Intelligent Computing Center" is a facility that uses large-scale heterogeneous computing resources, including general-purpose and intelligent computing power, to provide the computing power, data, and algorithms required for AI applications (such as AI deep learning model development, model training, and model inference). The Intelligent Computing Center encompasses facilities, hardware, and software, and provides a full stack of capabilities, from bottom-level computing power to top-level application enablement.

[0004] “Intelligent Computing Center” includes but is not limited to “Smart Computing Center”.

[0005] "Intelligent Computing Center" refers to an artificial intelligence computing center. It is a type of computing power infrastructure that is based on artificial intelligence theory, adopts artificial intelligence computing architecture, and provides computing power services, data services, and algorithm services required for artificial intelligence applications.

[0006] "Computing power" is the core of the "Intelligent Computing Center", "Intelligent Computing Center Cloud Platform" and "Intelligent Computing Center". It is the ability of computer equipment or computing / data center to process information. It is the ability of computer hardware and software to work together to execute certain computing needs. It is the computing power to achieve target result output by processing information data. It is a new type of productivity that integrates information computing power, network carrying capacity, and data storage capacity. It mainly provides services to society through computing power infrastructure.

[0007] Amid the rapid development of cloud computing and big data technologies, efficient computing power management and cost control are crucial for intelligent computing centers, the core infrastructure supporting massive data processing, complex model training, and high-concurrency business operations. Existing technologies commonly employ quota allocation or post-payment cost control models. Quota management implements static budget control by presetting a maximum limit on computing power usage. However, due to a lack of dynamic forecasting of user business trends and fluctuations in computing power demand, critical operations are often forced to be interrupted mid-execution due to budget depletion, severely impacting business continuity. While post-payment models offer flexible billing based on actual usage, they rely on users to independently control computing power consumption and lack proactive intervention mechanisms for sudden computing power spikes. This can easily lead to cost overruns caused by excessive resource utilization, or to idle and wasted resources due to uneven resource allocation. Therefore, since the emergence of intelligent computing centers, addressing the lag in budget control in existing computing power management and achieving optimal computing power allocation and refined cost control has been a pressing issue. Summary of the Invention

[0008] The present invention provides a method and device for controlling abnormal use of computing power of an intelligent computing center cloud platform, which can solve the problem of lagging budget control in existing computing power management and realize the optimized configuration of computing power and refined control of costs.

[0009] In order to solve the above-mentioned technical problems, the present invention is achieved as follows:

[0010] In a first aspect, the present invention provides a method for controlling abnormal use of computing power of an intelligent computing center cloud platform, the method comprising:

[0011] Step S1: Obtain the historical computing power usage data of the user in the first cycle;

[0012] Step S2: Inputting the historical computing power usage data into a preset time series model to calculate the computing power demand information of the user in a second period, where the second period is the period after the first period;

[0013] Step S3: Divide the user's computing power budget into a first computing power budget and a second computing power budget based on the computing power demand information and a preset user policy; wherein the first computing power budget is greater than or equal to the user's minimum computing power demand for operations within a preset range, and the second computing power budget is used to respond to the user's flexible computing power demand for operations outside the preset range;

[0014] Step S4: When the computing power demand of the user is greater than or equal to at least one of the first computing power budget and the second computing power budget, at least one of flow limiting processing and alarm processing is performed on the user's job.

[0015] In one embodiment, the method further comprises:

[0016] Step S5: When the second period ends, obtain the actual computing power demand of the user in the second period;

[0017] Step S6: Based on the actual computing power demand and the computing power demand information, adjust the parameters of the preset time series model.

[0018] In one embodiment, step S4 includes at least one of the following:

[0019] Step S41: When the computing power demand of the user is greater than or equal to the first computing power budget, suspending the user's operations outside the preset range;

[0020] Step S42: When the computing power requirement of the user is greater than or equal to the second computing power budget, perform flow limiting processing on the user's operations outside the preset range;

[0021] Step S43: When the computing power demand of the user is greater than or equal to the second computing power budget, an alarm is issued to the user.

[0022] In one embodiment, the preset time series model is any one of an autoregressive integrated moving average (ARIMA) model, an exponential smoothing model, and a deep learning model based on a long short-term memory network (LSTM).

[0023] In one embodiment, the historical computing power usage data includes at least one of a computing power usage log, a computing power running task job queue length, a historical peak computing power usage, and a service level agreement (SLA) level.

[0024] In one embodiment, the preset user policy includes at least one of the following:

[0025] a first user policy, configured to allocate computing power to the first computing power budget and the second computing power budget according to the user's level, wherein a proportion of the first computing power budget corresponding to a high-priority user is higher than a proportion of the first computing power budget corresponding to a low-priority user;

[0026] a second user policy, configured to assign weights to the first computing power budget and the second computing power budget according to a priority of the job;

[0027] The third user policy is used to obtain a preset allocation ratio of the first computing power budget and the second computing power budget through a preconfigured user interface.

[0028] In a second aspect, the present invention provides a device for controlling abnormal use of computing power of an intelligent computing center cloud platform, the device comprising:

[0029] The data acquisition module is used to obtain the historical computing power usage data of users in the first cycle;

[0030] an information prediction module, configured to input the historical computing power usage data into a preset time series model to calculate the computing power demand information of the user in a second period, where the second period is the period following the first period;

[0031] a budget division module, configured to divide the computing power budget of the user into a first computing power budget and a second computing power budget based on the computing power demand information and a preset user policy; wherein the first computing power budget is greater than or equal to the minimum computing power demand of the user for operations within a preset range, and the second computing power budget is used to respond to the flexible computing power demand of the user for operations outside the preset range;

[0032] An overspending processing module is used to perform at least one of flow limiting and alarm processing on the user's job when the user's computing power demand is greater than or equal to at least one of the first computing power budget and the second computing power budget.

[0033] In a third aspect, the present invention provides a server comprising: a processor, a memory, and a program stored in the memory and runnable on the processor. When the program is executed by the processor, the steps of the method for controlling abnormal computing power usage of the intelligent computing center cloud platform as described in the first aspect above are implemented.

[0034] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for controlling abnormal use of computing power of an intelligent computing center cloud platform as described in the first aspect above.

[0035] In a fifth aspect, the present invention provides a computer program product comprising computer instructions, which, when executed by a processor, implement the steps of the method for controlling abnormal use of computing power of an intelligent computing center cloud platform as described in the first aspect above.

[0036] In the present invention, the abnormal computing power usage control method running on the intelligent computing center cloud platform predicts the computing power demand of the user in the next cycle and then splits the budget into a first computing power budget and a second computing power budget, namely a fixed demand guarantee part and an elastic demand response part. This ensures the stability of the computing power supply for the user's basic operations while retaining the flexibility to deal with sudden demands. When it is detected that the actual demand exceeds the budget threshold, timely intervention is made through the current limiting and alarm mechanism, effectively avoiding excessive consumption of computing power and the risk of system overload. At the same time, the dynamic adjustment strategy balances resource utilization and user service quality, significantly improving the cost control capability of the intelligent computing center cloud platform and providing users with a more reliable and efficient computing power service experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present invention. The same reference symbols are used throughout the drawings to represent the same components. In the drawings:

[0038] Figure 1 This is a flow chart of a method for controlling abnormal computing power usage of an intelligent computing center cloud platform in the present invention;

[0039] Figure 2 This is a structural diagram of a device for controlling abnormal computing power usage of an intelligent computing center cloud platform in the present invention;

[0040] Figure 3 The figure is a schematic structural diagram of an electronic device in the present invention. DETAILED DESCRIPTION

[0041] The following will clearly and completely describe the technical solutions of the present invention in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0042] First, the technical terms involved in the present invention are briefly explained below.

[0043] The "computing power" mentioned in the present invention refers to: the ability of computer equipment or computing / data centers to process information, the ability of computer hardware and software to work together to execute certain computing requirements, and the computing power to achieve target result output by processing information data. It is a new type of productivity that integrates information computing power, network carrying capacity, and data storage capacity, and mainly provides services to society through computing power infrastructure.

[0044] The "computing power" (CP) mentioned in the present invention refers to: the ability of a data center server to process data and output results. It is a comprehensive indicator to measure the computing power of a data center, including general computing power, super computing power and intelligent computing power. The commonly used unit of measurement is the number of floating-point operations performed per second (FLOPS, 1EFLOPS=10^18FLOPS). The larger the value, the stronger the comprehensive computing power. According to calculations, 1EFLOPS is approximately the computing power output of 5 Tianhe-2A or 500,000 mainstream server CPUs or 2 million mainstream notebooks. The calculation formula is: CP=CP 通用 +CP 智能 +CP 超级 .

[0045] The "carrying capacity" (Network Power, NP) mentioned in the present invention refers to: it is the performance of the data transmission capability of the computing power facility, which includes comprehensive capabilities such as network architecture, network bandwidth, transmission latency, intelligent management and scheduling, etc. It involves network transmission within and between data centers, and is a comprehensive indicator for measuring network transmission scheduling capabilities.

[0046] The "Storage Power" (SP) described in this invention refers to the comprehensive capabilities of a data center in terms of data storage capacity, performance, security and reliability, and environmental friendliness. It is a comprehensive indicator for measuring a data center's data storage capacity, encompassing both external storage devices such as storage arrays and internal server storage. Storage capacity is commonly measured in exabytes (EB, 1EB = 2^60 bytes), while performance is commonly measured in IOPS / TB (Input / Output Operations Per Second / TB). Disaster recovery ratio is a key indicator of security and reliability.

[0047] The "computing power infrastructure" mentioned in the present invention refers to a new type of information infrastructure that integrates information computing power, network carrying capacity, and data storage capacity, and can realize the centralized calculation, storage, transmission and application of information.

[0048] The "new information infrastructure" mentioned in the present invention refers to: mainly including network infrastructure such as 5G networks, fiber-optic broadband networks, backbone networks, international communication networks, satellite Internet, computing power infrastructure such as data centers, general computing power centers, intelligent computing centers, supercomputing centers, and new technology facilities such as artificial intelligence, blockchain, and quantum computing.

[0049] The "computing power" mentioned in the present invention includes: general computing power, intelligent computing power and super computing power.

[0050] The "general computing power" mentioned in the present invention refers to the computing power provided by servers based on CPU (Central Processing Unit) chips, which is used to support basic general computing such as cloud computing and edge computing.

[0051] The "intelligent computing power" mentioned in this invention refers to: a computing platform based on specialized chips such as GPU (Graphics Processing Unit), FPGA (Field Programmable Gate Array), and ASIC (Application Specific Integrated Circuit) for various innovative artificial intelligence applications, such as natural language processing and machine vision.

[0052] The "supercomputing power" mentioned in the present invention refers to the computing power provided by high-performance computing clusters such as supercomputers. It utilizes the centralized computing resources of multiple computer systems working in parallel and uses a dedicated operating system to handle extremely complex or data-intensive problems. It is mainly used for calculations in cutting-edge scientific fields, such as planetary simulation, drug molecule design, genetic analysis, etc.

[0053] The "intelligent computing center" described in this article refers to a facility that provides the computing power, data, and algorithms required for artificial intelligence applications (such as AI deep learning model development, model training, and model inference) by utilizing large-scale heterogeneous computing resources, including general-purpose computing power (CPU) and intelligent computing power (GPU, FPGA, ASIC, etc.). The intelligent computing center encompasses facilities, hardware, and software, and provides a full stack of capabilities, from bottom-level computing power to top-level application enablement.

[0054] The "intelligent computing center cloud platform" mentioned in the present invention refers to: a cloud computing platform that provides comprehensive services based on the hardware resources and software resources of the intelligent computing center.

[0055] The "intelligent computing center" mentioned in the present invention includes but is not limited to the "intelligent computing center".

[0056] The "intelligent computing center" mentioned in the present invention is an artificial intelligence computing center, which is a type of computing power infrastructure based on artificial intelligence theory, adopts artificial intelligence computing architecture, and provides computing power services, data services and algorithm services required for artificial intelligence applications.

[0057] The "computing power center" mentioned in the present invention refers to: a facility that is mainly composed of infrastructure such as wind, fire, water, electricity, and IT hardware and software equipment, and has computing power, transportation capacity, and storage capacity, including general data centers, intelligent computing centers, supercomputing centers, etc.

[0058] The "supercomputing center" mentioned in the present invention refers to: a supercomputing data center, which is a data center based on a supercomputer or a large-scale computing cluster, which can provide large-scale computing, storage and network services and other functions, and is widely used in application scenarios such as aerospace, national defense, oil exploration, climate modeling and genome sequencing.

[0059] The "computing resources" mentioned in the present invention refer to: technologies and facilities with information computing, transmission, storage and application capabilities required for the development of a digital society, including but not limited to computing resources such as CPUs and GPUs, network resources such as switches and routers, computing resources such as storage arrays and distributed storage, security resources such as firewalls and intrusion detection systems, and supporting and guarantee resources such as wind, fire, water and electricity.

[0060] The "computing power operation task" mentioned in the present invention refers to: a specific workload or job executed on computing power resources that requires a certain amount of computing power support, usually involving complex data processing, numerical calculations, model training or simulation scenarios.

[0061] For details, see Figure 1 , Figure 1 This is a flow chart of a method for controlling abnormal computing power usage of an intelligent computing center cloud platform provided by an embodiment of the present invention. The method specifically includes the following steps:

[0062] Step S1: Obtain the user's historical computing power usage data in the first cycle.

[0063] In the above steps, a complete portrait of the user's computing power behavior is constructed through multi-source data collection (logs, queues, SLAs, etc.). In addition, the embodiment of the present application adopts period granularity adaptation to support data collection at different period levels such as day, week, and month. For example, the daily period collects daily data, and the monthly period summarizes the data of the current month. At the same time, the combination of real-time and historical data is adopted, which includes both the current period data collected in real time and the long-term trend data of the historical period.

[0064] Therefore, the above steps provide a comprehensive and dynamic data foundation for subsequent prediction models, avoiding prediction bias caused by single-dimensional data. At the same time, user needs can be differentiated by SLA level, providing a differentiated basis for minimum budget allocation.

[0065] Step S2: Input the historical computing power usage data into a preset time series model to calculate the computing power demand information of the user in the second period, where the second period is the next period after the first period.

[0066] In the embodiment of the present application, a lightweight time series model such as ARIMA, exponential smoothing, or a deep learning model such as a long short-term memory network (LSTM) can be selected according to user needs. The historical computing power usage data (such as daily computing power usage and the length of the computing power running task job queue) is standardized and input into the model. As a result, the model can output the computing power demand forecast value (Dx) for the second period (such as the next day, next week, next month), including the peak value, average value, and fluctuation range.

[0067] It should be noted that the embodiments of the present application support flexible switching of models of different complexity (lightweight and deep learning) to meet the needs of different scenarios. At the same time, using independent prediction of periodic granularity, the prediction models of daily, weekly, and monthly periods can be trained independently to avoid period conflicts. For example, daily period prediction does not rely on monthly period data. In addition, model parameters can also be dynamically optimized based on historical error rates (for example, adjusting the hidden layer size of LSTM).

[0068] In this way, the above steps can avoid resource shortage or waste by predicting user computing power requirements in advance, and can also improve prediction accuracy through multi-model adaptation.

[0069] Step S3: Based on the computing power demand information and the preset user policy, the user's computing power budget is divided into a first computing power budget and a second computing power budget; wherein the first computing power budget is greater than or equal to the user's minimum computing power demand for operations within a preset range, and the second computing power budget is used to respond to the user's elastic computing power demand for operations outside the preset range.

[0070] In some embodiments of the present application, the user's computing power budget can be divided into a first computing power budget and a second computing power budget based on computing power demand information and preset user policies. In a specific embodiment, for the calculation of the guaranteed budget (B), if the predicted value (Dx) is greater than or equal to the guaranteed minimum value, then B = min (Dx, the user sets the guaranteed upper limit). If Dx is less than the guaranteed minimum value, then B = Dx, E = 0 (the elastic budget is 0). For the calculation of the elastic budget (E), it can be obtained that E = Dx-B.

[0071] It's worth noting that the aforementioned user-level policies can prioritize high-priority users (such as enterprise users) with a higher guaranteed budget share, or employ job-priority policies, allowing for temporary expansion of the flexible budget for high-priority jobs (such as real-time tasks). Alternatively, a custom ratio policy can be employed, allowing users to set the guaranteed / flexible ratio through the configuration interface, such as a fixed 1:1 ratio or a dynamic ratio adjusted based on job type.

[0072] Furthermore, this embodiment of the application can dynamically adjust the guaranteed and flexible budgets based on predicted values ​​and user policies, rather than fixed values. Furthermore, this embodiment of the application supports multi-dimensional policy configurations such as user level, job priority, and custom ratios, allowing the flexible budget to be dynamically adjusted based on actual needs, such as when a high-priority job triggers a temporary capacity expansion.

[0073] In this way, the embodiments of the present application can ensure the minimum computing power requirement (i.e., the guaranteed budget amount) for key tasks, avoid job interruptions due to budget exhaustion, and respond to sudden demands through flexible budgets to reduce resource waste.

[0074] Step S4: When the computing power demand of the user is greater than or equal to at least one of the first computing power budget and the second computing power budget, at least one of flow limiting processing and alarm processing is performed on the user's job.

[0075] In one specific embodiment, when a user's actual computing power consumption exceeds the guaranteed budget, throttling can be triggered, such as pausing non-critical jobs. In another specific embodiment, when a user's actual computing power consumption exceeds the flexible budget, throttling (such as limiting high-priority jobs) or alarms (such as sending alarm notifications) can be triggered.

[0076] It can be understood that the embodiments of the present application can adopt two response strategies: current limiting processing and alarm processing. Current limiting processing is to dynamically adjust resource allocation according to job priority (for example, suspend low-priority jobs), and alarm processing is to remind users of the risk of overspending through email, text message or system notification.

[0077] In this way, the embodiment of the present application triggers different levels of flow control / alarms through the guaranteed budget and flexible budget respectively, thereby preventing resource waste or service interruption due to budget overspending, and can also balance resource allocation and user needs through a layered response mechanism.

[0078] In one embodiment, the method further comprises:

[0079] Step S5: When the second period ends, obtain the actual computing power demand of the user in the second period;

[0080] Step S6: Based on the actual computing power demand and the computing power demand information, adjust the parameters of the preset time series model.

[0081] In some embodiments, at the end of the second period (e.g., day, week, or month), the data collection module may collect in real time the actual computing power usage data of the user during the period, including:

[0082] Actual computing power consumption, that is, the total amount of CPU, GPU, memory and other resources actually used by the user during the cycle;

[0083] Job completion status, such as the number of jobs submitted by users, completion rate, timeout rate, etc.

[0084] Resource utilization, such as utilization of various resource types, GPU utilization, memory usage, etc.

[0085] Records of abnormal events, such as job interruptions and throttling events caused by insufficient resources.

[0086] Therefore, the actual demand data can be compared with the computing power demand information (Dx) output by the prediction model to generate an error analysis report, such as the deviation rate and fluctuation range between the predicted value and the actual value.

[0087] It's worth noting that the embodiments of this application support independent data collection at different time periods, such as daily, weekly, and monthly, ensuring that the data granularity matches the prediction model. This also comprehensively reflects actual user needs by combining multi-dimensional data such as computing power consumption, job completion rate, and resource utilization. Furthermore, by recording abnormal events such as throttling and overspending, it provides more accurate feedback for subsequent model adjustments.

[0088] In some other embodiments, the actual demand data is compared with the predicted value (Dx) and the error rate can be calculated, for example Specifically, dynamically adjusting the parameters of the time series model according to the error rate may include:

[0089] Model structure optimization, such as increasing the number of hidden layer nodes in the LSTM model to adapt to nonlinear fluctuations;

[0090] Weight adjustment, for example, adjusting the difference order or smoothing coefficient of the ARIMA model based on the historical error rate;

[0091] Adapt the cycle granularity. If the daily cycle prediction error is found to be large, increase the training frequency of the daily cycle model or introduce more features, such as job type and SLA level.

[0092] Therefore, the adjusted model parameters can be stored in the preset time series model for use in the next round of prediction.

[0093] As can be seen, this embodiment of the present application achieves continuous improvement in model prediction accuracy through a "prediction-execution-feedback-adjustment" cycle, and automatically adjusts model parameters based on the error rate, rather than relying on manual intervention. This improves the accuracy of the prediction model and reduces resource waste or service interruptions caused by prediction bias. Furthermore, dynamic parameter adjustment can also be used to adapt to dynamic changes in user needs, such as sudden tasks and seasonal fluctuations.

[0094] In one embodiment, step S4 includes at least one of the following:

[0095] Step S41: When the computing power demand of the user is greater than or equal to the first computing power budget, suspending the user's operations outside the preset range;

[0096] Step S42: When the computing power requirement of the user is greater than or equal to the second computing power budget, perform flow limiting processing on the user's operations outside the preset range;

[0097] Step S43: When the computing power demand of the user is greater than or equal to the second computing power budget, an alarm is issued to the user.

[0098] In some embodiments, the computing power consumption of the user in the current cycle can be tracked in real time to determine whether it exceeds the minimum budget (first computing power budget) or the flexible budget (second computing power budget). If the user's computing power demand is greater than or equal to the minimum budget (B), the minimum budget mechanism is triggered to directly limit non-critical jobs and ensure core tasks. If the user's computing power demand is greater than or equal to the flexible budget (E), the current limiting processing mechanism or the alarm mechanism can be triggered. Among them, the current limiting strategy can be dynamically adjusted based on the job priority, for example, giving priority to high-priority jobs. The alarm strategy can be combined with differentiated configurations such as user level, SLA level, etc., for example, high-priority users trigger more urgent alarms.

[0099] It is understandable that the above-mentioned flow limiting strategy can be dynamically adjusted based on the job priority, and the alarm strategy can be configured in a differentiated manner based on user level, SLA level, etc. For example, a corporate user has a surge in computing power demand due to sudden tasks within a monthly cycle. If the minimum budget is 60 units and the actual demand is 70 units, the system will suspend non-critical jobs. If the flexible budget is 40 units and the actual demand is 50 units, the system will limit the flow of non-critical jobs. If the flexible budget is exhausted, the system will send an alarm via email, prompting the user to adjust the budget or optimize the job. In this way, the critical tasks of corporate users can be guaranteed, while the flow limiting and alarm mechanisms can avoid resource waste and improve overall resource utilization.

[0100] In one embodiment, the preset time series model is any one of an autoregressive integrated moving average (ARIMA) model, an exponential smoothing model, and a deep learning model based on a long short-term memory network (LSTM).

[0101] Optionally, the preset time series model in the embodiment of the present application can be an autoregressive integrated moving average (ARIMA) model. Specifically, the ARIMA model can perform a stationary test on the user's historical computing power usage data. If it is non-stationary, differential processing is performed. The parameters (p, d, q) of the ARIMA model can also be determined through ACF (autocorrelation function) and PACF (partial autocorrelation function) graphs. The ARIMA model can also be trained using historical computing power usage data to output a predicted value (Dx), which can then be used as the computing power demand for the next cycle and input into the budget partitioning module. The ARIMA model has a good prediction effect on linear trends and seasonal fluctuations, and is suitable for scenarios with low computing resource requirements.

[0102] Optionally, the preset time series model can adopt an exponential smoothing model. Specifically, the exponential smoothing model can perform denoising processing (such as sliding average) on the historical computing power usage data, select single exponential smoothing (applicable to non-trend / seasonal data) or double exponential smoothing (applicable to trend data) according to the data characteristics, and optimize the smoothing coefficient (α) by minimizing the prediction error (such as mean square error), thereby inputting the predicted value Dx into the budget partitioning module as the computing power demand for the next cycle. In the embodiment of the present application, the preset time series model adopts an exponential smoothing model that only requires the weighted average of historical data, has low computational complexity, and can quickly respond to data changes by adjusting the smoothing coefficient α, and has a faster response speed to sudden changes in demand.

[0103] Optionally, the preset time series model can adopt an LSTM-based deep learning model. Specifically, the LSTM-based deep learning model can standardize the historical computing power usage data (such as Z-score normalization) and construct a time series window. By designing the LSTM network structure (such as input layer, hidden layer, output layer), and setting hyperparameters (such as the number of hidden layer nodes and learning rate), the LSTM model is trained using the historical computing power usage data, and the predicted value (Dx) is output. Finally, the predicted value Dx is input into the budget allocation module as the computing power demand for the next cycle.

[0104] In this way, the embodiments of the present application can flexibly select time series models to adapt to different user scenarios, balance prediction accuracy and resource consumption, and improve overall computing power management efficiency.

[0105] In one embodiment, the historical computing power usage data includes at least one of a computing power usage log, a computing power running task job queue length, a historical peak computing power usage, and a service level agreement (SLA) level.

[0106] In some specific embodiments, the data collection module collects historical computing power usage data of users in real time, including but not limited to the following:

[0107] Computing power usage logs, which include the number of jobs submitted by users in a historical period, job types (e.g., compute-intensive, I / O-intensive), and resource consumption (e.g., CPU, GPU, memory, etc.);

[0108] The length of the computing power running task job queue can be the number of jobs that users have queued for execution in the historical period;

[0109] Historical peak values, i.e., the peak value of computing power usage by users in a historical period, such as a sudden increase in GPU usage on a certain day;

[0110] SLA level refers to the user's Service Level Agreement (SLA) requirements. For example, high-priority users must guarantee a minimum computing power.

[0111] In one embodiment, the preset user policy includes at least one of the following:

[0112] a first user policy, configured to allocate computing power to the first computing power budget and the second computing power budget according to the user's level, wherein a proportion of the first computing power budget corresponding to a high-priority user is higher than a proportion of the first computing power budget corresponding to a low-priority user;

[0113] a second user policy, configured to assign weights to the first computing power budget and the second computing power budget according to a priority of the job;

[0114] The third user strategy is used to obtain a preset allocation ratio of the first computing power budget and the second computing power budget through a preconfigured user interface.

[0115] In a specific embodiment, user levels can be defined according to user types (such as corporate users, scientific research institution users, ordinary users) or SLA levels (such as high priority, medium priority, low priority). In addition, the budget ratio can be updated in real time according to changes in user levels (such as upgrading to high priority). In this way, the embodiment of the present application can dynamically adjust the minimum budget ratio by user level to ensure resource protection for high-priority users, and also supports administrators to adjust the budget ratio thresholds of different levels through the configuration interface. In addition, the minimum budget of high-priority users can also be allocated first to avoid interruptions caused by resource occupation by low-priority users.

[0116] In a specific embodiment, the job priority can be defined according to the job type (such as real-time tasks, batch tasks, low-priority tasks) or user-defined tags (such as "urgent" and "regular"). In addition, the budget weight can be adjusted in real time according to the change of job priority (such as task status update). In this way, the embodiment of the present application can adjust the minimum budget ratio in real time through the job priority to ensure the allocation of resources for high-priority tasks, while supporting customized job priorities. In addition, the minimum budget for high-priority jobs in the embodiment of the present application can also be temporarily expanded to avoid task failure due to insufficient resources.

[0117] In another specific embodiment, a user interface (such as a web console or an API interface) can be provided to allow the user to input a customized minimum budget and flexible budget ratio. Specifically, the budget ratio input by the user can be stored in a user profile or database. In the budget division module, the minimum budget and flexible budget are dynamically adjusted according to the user-defined ratio. Among them, the embodiment of the present application supports users to flexibly configure the budget ratio according to business needs (such as a fixed ratio, a dynamic ratio, or a differentiated ratio triggered based on the job type). Moreover, the user-defined budget ratio can be saved for a long time to avoid repeated configuration. In addition, it also supports combined use with the first and second user policies (such as high-priority users can further customize the budget ratio).

[0118] In this way, the above-mentioned embodiments of the present application can flexibly adapt to different user needs, balance resource guarantee and cost control, and improve the overall computing power management efficiency by combining the application of the first, second, and third user strategies.

[0119] See Figure 2 , Figure 2 The embodiment of the present invention provides a device 20 for controlling abnormal computing power usage of an intelligent computing center cloud platform, the device comprising:

[0120] The data acquisition module 21 is used to obtain the historical computing power usage data of the user in the first period;

[0121] An information prediction module 22 is configured to input the historical computing power usage data into a preset time series model to calculate the computing power demand information of the user in a second period, where the second period is the period following the first period;

[0122] A budget division module 23 is configured to divide the user's computing power budget into a first computing power budget and a second computing power budget based on the computing power demand information and a preset user policy; wherein the first computing power budget is greater than or equal to the user's minimum computing power demand for operations within a preset range, and the second computing power budget is used to respond to the user's flexible computing power demand for operations outside the preset range;

[0123] The overspending processing module 24 is used to perform at least one of flow limiting and alarm processing on the user's job when the user's computing power demand is greater than or equal to at least one of the first computing power budget and the second computing power budget.

[0124] In one embodiment, the computing power abnormal usage control device 20 of the intelligent computing center cloud platform further includes:

[0125] A first acquisition module is configured to acquire, when the second period ends, the actual computing power demand of the user in the second period;

[0126] A parameter adjustment module is used to adjust the parameters of the preset time series model based on the actual computing power demand and the computing power demand information.

[0127] In one embodiment, the overspending processing module 24 includes at least one of the following:

[0128] a job suspending unit, configured to suspend the user's job outside the preset range when the user's computing power demand is greater than or equal to the first computing power budget;

[0129] a current limiting processing unit, configured to perform current limiting processing on operations of the user outside the preset range when the computing power demand of the user is greater than or equal to the second computing power budget;

[0130] A user alarm unit is configured to issue an alarm to the user when the computing power demand of the user is greater than or equal to the second computing power budget.

[0131] In one embodiment, the preset time series model is any one of an autoregressive integrated moving average (ARIMA) model, an exponential smoothing model, and a deep learning model based on a long short-term memory network (LSTM).

[0132] In one embodiment, the historical computing power usage data includes at least one of a computing power usage log, a computing power running task job queue length, a historical peak computing power usage, and a service level agreement (SLA) level.

[0133] In one embodiment, the preset user policy includes at least one of the following:

[0134] a first user policy, configured to allocate computing power to the first computing power budget and the second computing power budget according to the user's level, wherein a proportion of the first computing power budget corresponding to a high-priority user is higher than a proportion of the first computing power budget corresponding to a low-priority user;

[0135] a second user policy, configured to assign weights to the first computing power budget and the second computing power budget according to a priority of the job;

[0136] The third user policy is used to obtain a preset allocation ratio of the first computing power budget and the second computing power budget through a preconfigured user interface.

[0137] In the embodiment of the present invention, the computing power abnormal use control device 20 of the intelligent computing center cloud platform not only ensures the stability of the computing power supply for the user's basic operations, but also retains the flexibility to deal with sudden demands. Moreover, when it is detected that the actual demand exceeds the budget threshold, it can intervene in time through the current limiting and alarm mechanism, effectively avoiding the excessive consumption of computing power resources and the risk of system overload. At the same time, it balances resource utilization and user service quality through dynamic adjustment strategies, significantly improving the resource scheduling efficiency, cost control capabilities and operational stability of the intelligent computing center cloud platform, and providing users with a more reliable and efficient computing power service experience. In addition, the computing power abnormal use control device 20 of the intelligent computing center cloud platform provided in the embodiment of the present invention is capable of realizing each process of each embodiment of the computing power abnormal use control method of the intelligent computing center cloud platform, and the technical features correspond one to one, and can achieve the same technical effect. In order to avoid repetition, it will not be repeated here.

[0138] It should be noted that the computing power abnormal usage control device 20 of the intelligent computing center cloud platform in the embodiment of the present invention can be a device, or a component, integrated circuit, or chip in an electronic device.

[0139] The embodiment of the present invention further provides an electronic device, see Figure 3 , Figure 3 The electronic device includes a memory 31, a processor 32, and a program or instruction stored in the memory 31. When the program or instruction is executed by the processor 32, the program or instruction can be realized. Figure 1 Any steps in the corresponding embodiment of the method for controlling abnormal use of computing power of the intelligent computing center cloud platform and the same beneficial effects are achieved will not be repeated here.

[0140] The processor 32 may be a CPU, an ASIC, an FPGA or a GPU.

[0141] Those skilled in the art will appreciate that all or part of the steps of the embodiment of the method for controlling abnormal computing power usage of the intelligent computing center cloud platform can be accomplished through hardware related to program instructions, and the program can be stored in a readable medium.

[0142] The embodiment of the present invention further provides a readable storage medium, on which a computer program is stored, which can realize the above-mentioned Figure 1 The corresponding steps in the embodiment of the abnormal use control method of computing power of the intelligent computing center cloud platform can achieve the same technical effect. To avoid repetition, they are not described here. The storage medium is such as read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk.

[0143] The present application also provides a computer program product including computer instructions, which, when executed by a processor, implement the above Figure 1 The various processes of the embodiment of the abnormal computing power usage control method of the intelligent computing center cloud platform shown in the figure can achieve the same technical effect. To avoid repetition, they will not be described here.

[0144] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0145] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better embodiment. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present invention.

[0146] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are protected by the present invention.

Claims

1. A method for controlling abnormal use of computing power in an intelligent computing center cloud platform, characterized in that: include: Step S1: Obtain the historical computing power usage data of the user in the first cycle; Step S2: Inputting the historical computing power usage data into a preset time series model to calculate the computing power demand information of the user in a second period, where the second period is the period after the first period; Step S3: Divide the user's computing power budget into a first computing power budget and a second computing power budget based on the computing power demand information and a preset user policy; wherein the first computing power budget is greater than or equal to the user's minimum computing power demand for operations within a preset range, and the second computing power budget is used to respond to the user's flexible computing power demand for operations outside the preset range; Step S4: When the computing power demand of the user is greater than or equal to at least one of the first computing power budget and the second computing power budget, at least one of current limiting processing and alarm processing is performed on the user's job.

2. The method according to claim 1, characterized in that The method further comprises: Step S5: When the second period ends, obtain the actual computing power demand of the user in the second period; Step S6: Based on the actual computing power demand and the computing power demand information, adjust the parameters of the preset time series model.

3. The method according to claim 1, characterized in that The step S4 includes at least one of the following: Step S41: When the computing power demand of the user is greater than or equal to the first computing power budget, suspending the user's operations outside the preset range; Step S42: When the computing power demand of the user is greater than or equal to the second computing power budget, perform flow limiting processing on the user's operations outside the preset range; Step S43: When the computing power demand of the user is greater than or equal to the second computing power budget, an alarm is issued to the user.

4. The method according to claim 1, wherein The preset time series model is any one of an autoregressive integrated moving average (ARIMA) model, an exponential smoothing model, and a deep learning model based on a long short-term memory network (LSTM).

5. The method according to any one of claims 1 to 4, characterized in that The historical computing power usage data includes at least one of the computing power usage log, the length of the computing power running task job queue, the historical peak value of computing power usage, and the service level agreement SLA level.

6. The method according to any one of claims 1 to 4, characterized in that The preset user policy includes at least one of the following: a first user policy, configured to allocate computing power to the first computing power budget and the second computing power budget according to the user's level, wherein a proportion of the first computing power budget corresponding to a high-priority user is higher than a proportion of the first computing power budget corresponding to a low-priority user; a second user policy, configured to assign weights to the first computing power budget and the second computing power budget according to a priority of the job; The third user strategy is used to obtain a preset allocation ratio of the first computing power budget and the second computing power budget through a preconfigured user interface.

7. A device for controlling abnormal computing power usage of an intelligent computing center cloud platform, characterized in that: include: The data acquisition module is used to obtain the historical computing power usage data of users in the first cycle; an information prediction module, configured to input the historical computing power usage data into a preset time series model to calculate the computing power demand information of the user in a second period, where the second period is the period following the first period; a budget division module, configured to divide the computing power budget of the user into a first computing power budget and a second computing power budget based on the computing power demand information and a preset user policy; wherein the first computing power budget is greater than or equal to the minimum computing power demand of the user for operations within a preset range, and the second computing power budget is used to respond to the flexible computing power demand of the user for operations outside the preset range; An overspending processing module is used to perform at least one of flow limiting and alarm processing on the user's job when the user's computing power demand is greater than or equal to at least one of the first computing power budget and the second computing power budget.

8. An electronic device, characterized in that: include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, the steps of the method for controlling abnormal use of computing power of an intelligent computing center cloud platform as described in any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for controlling abnormal use of computing power of an intelligent computing center cloud platform according to any one of claims 1 to 6.

10. A computer program product, characterized in that It includes computer instructions, which, when executed by a processor, implement the steps of the method for controlling abnormal use of computing power of an intelligent computing center cloud platform as described in any one of claims 1 to 6.

Images