Energy-saving method and device for computing power cloud platform of common computing power intelligent computing center

Abstract

The invention provides an energy-saving method and device for a computing power cloud platform of a common computing power intelligent computing center, 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 index data corresponding to the computing power cloud platform at a first time point; s2, inputting the index data into a target load prediction model to predict load data of the computing power cloud platform at a second time point to obtain second load data; s3, determining second system data corresponding to the computing power cloud platform at a second time point according to the second load data; and S4, adjusting the system data of the computing power cloud platform into second system data at a second time point. On the premise that the performance of the computing power cloud platform is guaranteed, the computing power resource consumption and the economic cost of the computing power cloud platform are greatly reduced, and wide application of the general computing power is facilitated.

Description

Technical Field

[0001] The present invention relates to the technical fields of intelligent computing centers, intelligent computing centers, and computing power infrastructure, and specifically relates to an energy-saving method and device for a computing power cloud platform of an intelligent computing center for inclusive computing power. Background Art

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

[0003] An "intelligent computing center" refers to a facility that uses large-scale heterogeneous computing power resources, including general computing power and intelligent computing power, and mainly provides the required computing power, data, and algorithms for artificial intelligence applications (such as scenarios of artificial intelligence deep learning model development, model training, and model inference, etc.). An intelligent computing center covers facilities, hardware, and software, and can provide full-stack capabilities from underlying computing power to top-level application enabling.

[0004] The "intelligent computing center" includes but is not limited to the "intelligent computing center".

[0005] An "intelligent computing center", that is, an artificial intelligence computing center, is a type of computing power infrastructure based on artificial intelligence theory, adopting an artificial intelligence computing architecture, and providing computing power services, data services, and algorithm services required for artificial intelligence applications.

[0006] "Computing power" is the core of "intelligent computing centers" and "intelligent computing centers". It is the ability of computer devices or computing / data centers to process information, the ability of computer hardware and software to cooperate to jointly execute a certain computing requirement, the computing ability to achieve the output of target results by processing information data, and a new type of productive force integrating information computing power, network carrying capacity, and data storage capacity. It mainly provides services to society through computing power infrastructure.

[0007] Currently, through cloud servers, it is possible to directly provide computing power resources required by users, and users can use the computing power resources to perform tasks such as model development. However, due to different task types and required computing power of different users, the computing power cloud platform usually remains in a high-energy-consuming state to ensure that the computing power resources are always on standby and can respond in a timely manner to the users' demands for computing power resources.

[0008] It can be seen that in the prior art, there are problems of large consumption of computing power resources and high economic costs of the computing power cloud platform, which is not conducive to the wide application of inclusive computing power. Summary of the Invention

[0009] Embodiments of the present invention provide an energy-saving method and device for a computing power cloud platform of an intelligent computing center for inclusive computing power, which solve the problems in the prior art of large consumption of computing power resources and high economic costs of the computing power cloud platform, and are not conducive to the wide application of inclusive computing power.

[0010] To solve the above problems, the present invention is implemented as follows:

[0011] In a first aspect, the present invention provides an energy-saving method for a computing power cloud platform of an inclusive computing power intelligent computing center, including:

[0012] Step S1, obtain the index data corresponding to the computing power cloud platform at a first time point, where the index data includes first load data and first system data, the first load data is used to indicate the load state of the computing power cloud platform at the first time point, and the first system data is used to indicate the hardware parameters of the computing power cloud platform corresponding to the load state at the first time point;

[0013] Step S2, input the index data into a target load prediction model to predict the load data of the computing power cloud platform at a second time point, and obtain second load data, where the second time point is a time point after the first time point;

[0014] Step S3, determine the second system data corresponding to the computing power cloud platform at the second time point according to the second load data;

[0015] Step S4, at the second time point, adjust the system data of the computing power cloud platform to the second system data.

[0016] Optionally, the first load data includes first performance data and first task data, the first performance data is the system performance data of the computing power cloud platform for processing computing power tasks at the first time point, and the first task data is the task data of the computing power cloud platform for processing computing power tasks at the first time point. The step S1, obtain the index data corresponding to the computing power cloud platform at the first time point, includes:

[0017] Step S11, obtain the first performance data corresponding to the computing power cloud platform at the first time point, where the first performance data includes at least one of the following: the utilization rate of the graphics processing unit (GPU) core and the occupancy rate of the video memory bandwidth;

[0018] Step S12, obtain the first task data corresponding to the computing power cloud platform at the first time point, where the first task data includes at least one of the following: task queue information, task size information, and task processing time information. Among them, the task queue information includes multiple target computing power tasks, the task size information is the task volume size corresponding to each target computing power task among the multiple target computing power tasks, and the task processing time information is the time consumed for processing each target computing power task among the multiple target computing power tasks;

[0019] Step S13: Obtain the first system data corresponding to the computing power cloud platform at the first time point.

[0020] Optionally, step S2: Input the metric data into the target load prediction model to predict the load data of the computing power cloud platform at the second time point, and obtain the second load data, including:

[0021] Step S21: Input the first performance data, the first task data, and the first system data into the target load prediction model to predict the load data of the computing power cloud platform at the second time point, and obtain the second load data and the target confidence level. The target confidence level is used to indicate the prediction confidence level of the target load prediction model for the second load data.

[0022] Optionally, step S3: Determine the second system data corresponding to the computing power cloud platform at the second time point according to the second load data, including:

[0023] Step S31: Based on the target confidence level and the second load data, determine the second system data corresponding to the computing power cloud platform at the second time point.

[0024] Optionally, step S4: At the second time point, adjust the system data of the computing power cloud platform to the second system data, including:

[0025] Step S41: At the second time point, determine the third system data of the computing power cloud platform. The third system data is the system data of the computing power cloud platform at the second time point;

[0026] Step S42: Determine the difference between the second system data and the third system data;

[0027] Step S43: Within the target time period, adjust the third system data multiple times according to the difference until the third system data is adjusted to the second system data. In the multiple adjustments, each adjustment adjusts the third system data by a preset percentage. The start time point of the target time period is the second time point.

[0028] Optionally, step S4: At the second time point, adjust the system data of the computing power cloud platform to the second system data, including:

[0029] Step S44: At the second time point, update the GPU voltage parameter based on the second system data to update the third system data of the computing power cloud platform to the second system data. The third system data is the system data of the computing power cloud platform at the second time point;

[0030] and / or,

[0031] Step S45: At the second time point, call the application programming interface based on the second system data to update the third system data of the computing power cloud platform to the second system data.

[0032] Optionally, before the step S2 of inputting the metric data into the target load prediction model to predict the load data of the computing power cloud platform at the second time point to obtain the second load data, the method further includes:

[0033] Step S5: Obtain a sample data set from the historical resource database. The sample data set includes multiple historical metric data corresponding to multiple historical time points. Each historical metric data in the multiple historical metric data includes historical load data and historical system data. The historical load data is used to indicate the load status of the computing power cloud platform at the historical time point, and the historical system data is used to indicate the hardware parameters of the computing power cloud platform corresponding to the load status at the historical time point;

[0034] Step S6: Train the load prediction model based on the sample data set to obtain the target load prediction model.

[0035] Optionally, the first system data includes at least one of the following: GPU temperature, GPU power consumption, GPU voltage, and GPU frequency. The second system data includes at least one of the following: GPU temperature, GPU power consumption, GPU voltage, and GPU frequency.

[0036] In a second aspect, the present invention further provides an energy-saving device for a computing power cloud platform of an inclusive computing power intelligent computing center. The device includes:

[0037] An acquisition module, configured to acquire metric data corresponding to the computing power cloud platform at a first time point. The metric data includes first load data and first system data. The first load data is used to indicate the load status of the computing power cloud platform at the first time point, and the first system data is used to indicate the hardware parameters of the computing power cloud platform corresponding to the load status at the first time point;

[0038] A prediction module, configured to input the metric data into the target load prediction model to predict the load data of the computing power cloud platform at a second time point to obtain second load data. The second time point is a time point after the first time point;

[0039] A determination module, configured to determine the second system data corresponding to the computing power cloud platform at the second time point according to the second load data;

[0040] An adjustment module, configured to adjust the system data of the computing power cloud platform to the second system data at the second time point.

[0041] In a third aspect, the present invention further provides an electronic device, including a processor, a memory, and a computer program stored on the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps in the energy-saving method for the computing power cloud platform of the inclusive computing power intelligent computing center as described in the first aspect above.

[0042] In a fourth aspect, the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps in the energy-saving method for the computing power cloud platform of the inclusive computing power intelligent computing center as described in the first aspect above.

[0043] In a fifth aspect, the present invention further provides a computer program product, including computer instructions. When the computer instructions are executed by a processor, they implement the steps in the energy-saving method for the computing power cloud platform of the inclusive computing power intelligent computing center as described in the first aspect above.

[0044] The present invention provides an energy-saving method and device for a computing power cloud platform of an inclusive computing power intelligent computing center. The method includes: Step S1, obtaining index data corresponding to the computing power cloud platform at a first time point; Step S2, inputting the index data into a target load prediction model to predict the load data of the computing power cloud platform at a second time point, obtaining second load data; Step S3, determining second system data corresponding to the computing power cloud platform at the second time point according to the second load data; Step S4, at the second time point, adjusting the system data of the computing power cloud platform to the second system data. By obtaining the monitored index data corresponding to the first time point and then using the target load prediction model to predict the load data of the computing power cloud platform at the second time point to obtain the second load data, the system data is adjusted to the second system data at the second time point, so that the computing power cloud platform achieves the effect of significantly reducing the computing power resource consumption and economic cost of the computing power cloud platform on the premise of ensuring performance, which is beneficial to the wide application of inclusive computing power. Description of the Drawings

[0045] To more clearly illustrate the technical solutions of the present invention, the following will briefly introduce the drawings required for the description of the present invention. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0046] Figure 1It is a flowchart of an energy-saving method for a computing power cloud platform of an inclusive computing power intelligent computing center provided by the present invention;

[0047] Figure 2 It is a schematic flowchart of the computing power cloud platform provided by the present invention;

[0048] Figure 3 It is a structural diagram of an energy-saving device for a computing power cloud platform of an inclusive computing power intelligent computing center provided by the present invention;

[0049] Figure 4 It is a structural diagram of an electronic device provided by the present invention. Detailed implementation manners

[0050] Next, the technical solutions in the present invention will be clearly and completely described in conjunction with the accompanying drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present invention.

[0051] The "computing power" described in the present invention refers to: the ability of a computer device or a computing / data center to process information, the ability of computer hardware and software to cooperate to jointly execute a certain computing requirement, the computing ability to output a target result by processing information data, a new type of productive force integrating information computing power, network carrying capacity, and data storage capacity, and mainly providing services to society through computing power infrastructure.

[0052] The "computational power" (Computational Power, CP) described in the present invention refers to: the ability of a data center server to process data and output results, a comprehensive index to measure the computing ability of a data center, including general computing ability, supercomputing ability, and intelligent computing ability. The commonly used measurement unit is the number of floating-point operations per second (FLOPS, 1 EFLOPS = 10^18 FLOPS), and the larger the value, the stronger the comprehensive computing ability. It is estimated that 1 EFLOPS is approximately the computing power output of 5 Tianhe 2A or 500,000 mainstream server CPUs or 2 million mainstream laptops. The calculation formula is: CP = CP 通用 + CP 智能 + CP 超级 .

[0053] The "network power" (Network Power, NP) described in the present invention refers to: the performance of the data transmission ability of computing power facilities, including the comprehensive ability including network architecture, network bandwidth, transmission delay, intelligent management and scheduling, etc., involving network transmission inside and between data centers, and is a comprehensive index to measure the network transmission scheduling ability.

[0054] The "Storage Power (SP)" described in the present invention refers to the comprehensive ability of a data center in four aspects: data storage capacity, performance, security and reliability, and green and low-carbon. It is a comprehensive indicator for measuring the data storage capacity of a data center, including external storage devices such as storage arrays and built-in storage devices of servers. The commonly used measurement unit for storage capacity is exabyte (EB, 1EB = 2^60 bytes), the commonly used measurement unit for performance is the number of read and write operations per second per unit capacity (IOPS / TB, Input / Output Operations Per Second / TB), and the disaster recovery ratio is an important manifestation of security and reliability.

[0055] The "computing power infrastructure" described in the present invention refers to a new type of information infrastructure that integrates information computing power, network carrying capacity, and data storage power, and can realize the centralized computing, storage, transmission, and application of information.

[0056] The "new type of information infrastructure" described in the present invention mainly includes network infrastructures such as 5G networks, fiber broadband networks, backbone networks, international communication networks, and satellite Internet, computing power infrastructures such as data centers, general computing power centers, intelligent computing centers, and supercomputing centers, and new technology facilities such as artificial intelligence, blockchain, and quantum computing. With the emergence and popularization of new general-purpose technologies, the form of the new type of information infrastructure will be more diverse.

[0057] The "computing power" described in the present invention includes general computing power, intelligent computing power, and super computing power.

[0058] The "general computing power" described 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.

[0059] The "intelligent computing power" described in the present invention refers to a computing platform that is scaled for various artificial intelligence innovation applications based on dedicated chips such as GPU (Graphics Processing Unit), FPGA (Field Programmable Gate Array), and ASIC (Application Specific Integrated Circuit), such as natural language processing and machine vision.

[0060] The "super computing power" described in the present invention refers to: mainly 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 processes extremely complex or data-intensive problems through a dedicated operating system. It is mainly used for computing in cutting-edge scientific fields, such as planetary simulation, drug molecule design, gene analysis, etc.

[0061] The "intelligent computing center" described in the present invention refers to: a facility that provides the required computing power, data, and algorithms mainly for artificial intelligence applications (such as scenarios like artificial intelligence deep learning model development, model training, and model inference) by using large-scale heterogeneous computing power resources, including general computing power (CPU) and intelligent computing power (GPU, FPGA, ASIC, etc.). The intelligent computing center covers facilities, hardware, and software and can provide full-stack capabilities from underlying computing power to top-level application enabling.

[0062] The "intelligent computing center" described in the present invention includes but is not limited to the "intelligent computing center".

[0063] The "intelligent computing center" described in the present invention, that is, the artificial intelligence computing center, is a type of computing power infrastructure based on artificial intelligence theory, adopting an artificial intelligence computing architecture, and providing computing power services, data services, and algorithm services required for artificial intelligence applications.

[0064] The "computing power center" described in the present invention refers to: a facility mainly composed of infrastructure such as wind, fire, water, and electricity and IT software and hardware devices, with computing power, transportation power, and storage power, including general data centers, intelligent computing centers, supercomputing centers, etc.

[0065] The "supercomputing center" described in the present invention refers to: that is, the supercomputing data center, which is a data center based on supercomputers or large-scale computing clusters and can provide functions such as large-scale computing, storage, and network services, and is widely used in application scenarios such as aerospace, national defense, oil exploration, climate modeling, and genome sequencing.

[0066] The "computing power resources" described in the present invention refers to: technologies and facilities with information computing, transmission, storage, and application capabilities required for the development of the digital society, including but not limited to computing resources such as CPU and GPU, network resources such as switches and routers, storage resources such as storage arrays and distributed storage, security resources such as firewalls and intrusion detection systems, and support and guarantee resources such as wind, fire, water, and electricity.

[0067] The "inclusive computing power" described in the present invention refers to providing appropriate and effective computing power services for all social strata and groups with computing power service needs at an affordable cost based on the requirements of equal opportunity and the principle of commercial sustainability.

[0068] The "computing power cloud platform" described in the present invention refers to an online platform that provides computing power resource services based on cloud computing technology. The platform allows users to obtain and use computing power resources on demand, such as virtual machines, information computing, data storage, network transportation, etc., without having to build and maintain physical infrastructure themselves. The computing power cloud platform is widely used in various scenarios, including enterprise IT, scientific computing, game development, data analysis, etc., and provides services through the computing power infrastructure.

[0069] The "model" described in the present invention includes, but is not limited to, "large language models" and "multimodal large models".

[0070] The "large language model" described in the present invention refers to a large-scale language model (LLM), which is a language model with a relatively large number of parameters, aiming to understand and generate human language, trained with a large amount of text data, and can perform a wide range of tasks including text summarization, translation, sentiment analysis, etc.

[0071] The "multimodal large model" (Multimodal Large Models) described in the present invention refers to a model that jointly trains multimodal information such as text, images, videos, and audio, including but not limited to multimodal large language models.

[0072] Please refer to Figure 1 , Figure 1 which is a flowchart of an energy-saving method for a computing power cloud platform of an inclusive computing power intelligent computing center provided by the present invention. As Figure 1 shown, it includes the following steps:

[0073] Step S1: Obtain the index data corresponding to the computing power cloud platform at the first time point. The index data includes first load data and first system data. The first load data is used to indicate the load status of the computing power cloud platform at the first time point, and the first system data is used to indicate the hardware parameters of the computing power cloud platform corresponding to the load status at the first time point.

[0074] In the present invention, the computing power cloud platform can receive the computing power task requests of users and process the computing power tasks. During the process of processing the computing power tasks, the computing power cloud platform will generate index data. It should be noted that due to the different computing power tasks requested by users, the index data of the computing power cloud platform at different time points is also different. For example, during the peak period of processing computing power tasks, the index data will increase correspondingly, while during the trough period of computing power tasks, the index data will decrease correspondingly.

[0075] Specifically, the metric data corresponding to the first time point of the computing power cloud platform includes first load data and first system data. Among them, the first load data is used to indicate the load status of the computing power cloud platform at the first time point, such as the usage rate of the Graphics Processing Unit (GPU), specific data of computing power tasks, and so on. The first system data is used to indicate the system hardware parameters of the computing power cloud platform at the first time point, such as voltage, frequency and other data, which are not specifically limited in the present invention.

[0076] Step S2: Input the metric data into the target load prediction model to predict the load data of the computing power cloud platform at the second time point, and obtain the second load data. The second time point is a time point after the first time point.

[0077] In the present invention, the target load prediction model is a pre-trained load prediction model. Among them, the load prediction model can be a Long Short-Term Memory (LSTM). LSTM is a special type of recurrent neural network designed to solve the problems of gradient vanishing and gradient explosion faced by traditional RNNs when processing long sequence data. In the present invention, the LSTM network is used to predict the load trend of the computing power cloud platform within a certain period of time in the future.

[0078] Specifically, input the metric data into the target load prediction model to predict the load data of the computing power cloud platform at the second time point, and output the second load data. This second load data corresponds to the second time point, that is, the target load prediction model predicts the load situation of the system at the second time point. According to the prediction situation, the computing power cloud platform can make corresponding adjustments to the system hardware data.

[0079] The second time point is a time point after the first time point. Exemplarily, for example, the interval between the second time point and the first time point can be 5 minutes or 10 minutes. The specific interval can be adaptively adjusted according to the actual situation and is not specifically limited in this embodiment.

[0080] It should be noted that in the present invention, the target load prediction model also has an online learning mechanism. Through the online learning mechanism, the model parameters can be adjusted according to the predicted error value, so as to update the target load prediction model, enabling the target load prediction model to better predict the metric data.

[0081] Step S3: Determine the second system data corresponding to the computing power cloud platform at the second time point according to the second load data.

[0082] In the present invention, after determining the second load data corresponding to the second time point, the second system data corresponding to the second time point is determined according to the second load data. Specifically, the second load data can be compared with the historical data pre-stored in the system, so as to determine the system data corresponding to the generation of the second load data. Exemplarily, for example, if the second load data indicates that the current computing power task is task A and the GPU utilization rate is B, then by querying task A in the historical data and when the GPU utilization rate is 50%, the system data indicates that the current voltage is C and the frequency is D, then the current voltage C and frequency D can be determined as the second system data. It should be noted that in this embodiment, there may be a situation where the second load data does not exactly correspond to the historical data. In this case, data similar to the second load data can be selected from the historical data as a reference to generate the corresponding second system data.

[0083] Step S4: At the second time point, adjust the system data of the computing power cloud platform to the second system data.

[0084] In the present invention, after determining the second system data, at the second time point, the current system data of the computing power cloud platform is adjusted to the second system data, so as to ensure the effect of reducing the consumption of computing power resources and the power consumption cost of the data center while satisfying the normal operation of the computing power cloud platform, and realize the maximization of the cluster energy efficiency ratio.

[0085] As Figure 2 shown, Figure 2 is the process schematic diagram in this application. Taking Figure 2 as an example to illustrate the implementation process of this application. Among them, the index acquisition system collects the relevant data of the GPU to generate the index data corresponding to the first time point, and pushes the index data to the target load prediction model to predict the load data at the second time point, so as to generate the second load data. Through data comparison between the second load data and the historical data in the historical resource database, the second system data is determined. The second system data is sent to the scheduling execution layer to update the GPU system hardware data, that is, to adjust the system data of the computing power cloud platform to the second system data.

[0086] After obtaining the monitoring index data corresponding to the first time point, the present invention uses the target load prediction model to predict the load data of the computing power cloud platform at the second time point, obtains the second load data, and thus adjusts the system data to the second system data at the second time point, so that the computing power cloud platform realizes the effect of greatly reducing the consumption of computing power resources and economic costs of the computing power cloud platform on the premise of ensuring performance, which is beneficial to the wide application of inclusive computing power.

[0087] In some feasible embodiments, optionally, the first load data includes first performance data and first task data. The first performance data is the system performance data of the computing power cloud platform for processing computing power tasks at the first time point, and the first task data is the task data of the computing power cloud platform for processing computing power tasks at the first time point. The step S1 of obtaining the metric data corresponding to the computing power cloud platform at the first time point includes:

[0088] Step S11: Obtain the first performance data corresponding to the computing power cloud platform at the first time point. The first performance data includes at least one of the following: the utilization rate of the graphics processing unit (GPU) cores and the occupancy rate of the video memory bandwidth.

[0089] Step S12: Obtain the first task data corresponding to the computing power cloud platform at the first time point. The first task data includes at least one of the following: task queue information, task size information, and task processing time information. Among them, the task queue information includes multiple target computing power tasks, the task size information is the task volume size corresponding to each target computing power task among the multiple target computing power tasks, and the task processing time information is the time consumed for processing each target computing power task among the multiple target computing power tasks.

[0090] Step S13: Obtain the first system data corresponding to the computing power cloud platform at the first time point.

[0091] In this embodiment, the first load data includes first performance data and first task data. Among them, the first performance data is the system performance data of the computing power cloud platform for processing computing power tasks at the first time point. Specifically, the first performance data includes the utilization rate of the GPU cores, the occupancy rate of the video memory bandwidth, and so on.

[0092] The first task data is the task data of the computing power cloud platform for processing computing power tasks at the first time point. Specifically, the first task data includes at least one of the following: task queue information, task size, and task processing time. Among them, the task queue information includes multiple sorted target computing power tasks. The task size is the task volume size corresponding to each target computing power task among the multiple target computing power tasks. For example, the amount of calculation required for each task, such as 5G or 10G, etc. The task processing time is the time consumed for processing each target computing power task among the multiple target computing power tasks. For example, the time consumed for each task can be 5 minutes, 10 minutes, etc.

[0093] Among them, optionally, the first system data includes at least one of the following: GPU temperature, GPU power consumption, GPU voltage, and GPU frequency. The second system data includes at least one of the following: GPU temperature, GPU power consumption, GPU voltage, and GPU frequency.

[0094] Specifically, the first system data can be the current temperature of the GPU, the current voltage of the GPU, the current operating frequency of the GPU, and the power consumption of the GPU. Among them, the power consumption of the GPU can be obtained through the NVML or DCGM tool. It should be noted that NVML (NVIDIA Management Library) is a set of APIs provided by NVIDIA for managing and monitoring NVIDIA GPUs. It allows developers to obtain the status information of GPUs, manage GPU resources, and perform performance monitoring in application programs. NVML is mainly used in data centers, supercomputers, and high-performance computing (HPC) environments to ensure the efficient operation of GPUs. DCGM (DataCenter GPU Manager) is a tool provided by NVIDIA for managing and monitoring NVIDIA GPUs in data centers. It is designed for large-scale GPU deployments and can help system administrators and developers effectively manage GPU resources, monitor performance, and optimize workloads. In this embodiment, by sequentially obtaining the relevant data of the computing power cloud platform, the metric data can be accurately generated, ensuring the accuracy of the metric data when input into the target prediction model.

[0095] Optionally, step S2, inputting the metric data into the target load prediction model to predict the load data of the computing power cloud platform at the second time point to obtain the second load data, includes:

[0096] Step S21, inputting the first performance data, the first task data, and the first system data into the target load prediction model to predict the load data of the computing power cloud platform at the second time point to obtain the second load data and the target confidence level. The target confidence level is used to indicate the prediction confidence level of the target load prediction model for the second load data.

[0097] In this embodiment, the metric data includes the first performance data, the first task data, and the first system data. The first performance data, the first task data, and the first system data are input into the target load prediction model to predict the load data of the computing power cloud platform at the second time point, and the second load data and the target execution degree are output. The target confidence level can be a specific value, that is, it is used to indicate the prediction confidence level of the target load prediction model for the second load data. When the target confidence level is higher, it indicates that the prediction result of the second load data is more accurate at this time. When the target confidence level is lower, it indicates that the prediction result of the second load data is less accurate at this time.

[0098] Exemplarily, when the target confidence level is lower than a certain threshold, it can be considered that the current prediction is inaccurate, and a prompt to re - perform the prediction is given. This threshold can be set according to the actual situation and is not specifically limited in this embodiment. In this embodiment, by adding the target confidence level to the output of the target load prediction model, it is possible to better determine whether the current prediction result is accurate, thereby facilitating the subsequent determination of the second system data corresponding to the second load data.

[0099] Optionally, step S3, determining the second system data corresponding to the computing power cloud platform at the second time point according to the second load data, includes:

[0100] Step S31, based on the target confidence level and the second load data, determining the second system data corresponding to the computing power cloud platform at the second time point.

[0101] In this embodiment, after determining the target confidence level, the second system data corresponding to the second time point can be determined based on the target confidence level and the second load data. It should be noted that in this embodiment, during the query process of the second load data, it is necessary to make a judgment based on the target confidence level. Exemplarily, for example, if the value of the target confidence level is 90%, then the credibility of the second load data is 90%. Then, in the subsequent adjustment process, the accuracy of the second system data corresponds to 90%. Thus, the second system data can be corrected according to the actual situation to ensure the accuracy of its adjustment.

[0102] In this embodiment, by generating the target confidence level, it is possible to better determine whether the second system data is accurate, and thus the second system data can be corrected according to the actual situation, ensuring the accuracy of the adjustment of the system data.

[0103] Optionally, step S4, adjusting the system data of the computing power cloud platform to the second system data at the second time point, includes:

[0104] Step S41, at the second time point, determining the third system data of the computing power cloud platform, where the third system data is the system data of the computing power cloud platform at the second time point;

[0105] Step S42, determining the difference between the second system data and the third system data;

[0106] Step S43, within the target time period, adjusting the third system data multiple times according to the difference until the third system data is adjusted to the second system data. In the multiple adjustments, each adjustment adjusts the third system data by a preset percentage, and the starting time point of the target time period is the second time point.

[0107] In this embodiment, at the second time point, the third system data of the computing power cloud platform can be determined, and this third system data is the normal system data of the computing power cloud platform at the second time point. By comparing the calculated second system data with the third system data, the difference between the second system data and the third system data is determined. Thus, the third system data is adjusted multiple times by this difference until the third system data is adjusted to the second system data.

[0108] Among them, the multiple adjustments may include adjusting the third system data by a preset percentage each time, and this preset percentage can be set according to the actual situation. Exemplarily, for example, adjusting the third system data by 5% each time, and then observing the change of the current system data. Thus, the third system data is adjusted in segments within the target time period. Specifically, the starting time point of the target time period is the second time point, and the length of the target time period can be set according to the actual situation, such as 5 minutes and 10 minutes, etc., and no specific limitation is made in this embodiment.

[0109] In this embodiment, by adjusting the third system data in stages, it is possible to avoid affecting the running state of the computing power cloud platform when the system data changes too much, thereby ensuring the stability of the operation of the computing power cloud platform.

[0110] Optionally, step S4, adjusting the system data of the computing power cloud platform to the second system data at the second time point, includes:

[0111] Step S44, at the second time point, updating the GPU voltage parameter based on the second system data to update the third system data of the computing power cloud platform to the second system data, where the third system data is the system data of the computing power cloud platform at the second time point;

[0112] And / or,

[0113] Step S45, at the second time point, calling the application programming interface based on the second system data to update the third system data of the computing power cloud platform to the second system data.

[0114] In this embodiment, when adjusting the system data of the computing power cloud platform to the second system data, the GPU voltage parameter can be updated to update the third system data of the computing power cloud platform to the second system data. Exemplarily, for example, adjusting nvidia - smi - lgc <frequency> to change the system frequency. It can also directly change the system data by calling the application programming interface, such as directly calling the manufacturer's application programming interface (such as NVIDIA NVAPI) to achieve.

[0115] In this embodiment, the above two methods can be used for updating simultaneously, or one method can be selected for adjustment, which can be determined according to the actual situation.

[0116] It should be noted that the computing power cloud platform may include multiple GPUs at the same time. Then, the second system data may include data for adjusting each GPU. For example, if the computing power cloud platform includes three GPUs of different models, the second system data may include the specific system data for the three GPUs of different models. Thus, the three GPUs can be adjusted respectively according to the specific system data of the three GPUs of different models, so as to realize the synchronous adjustment of multiple GPUs and avoid resource contention caused by frequency differences in cross-card tasks. The method provided by the present invention can migrate low-priority tasks (such as development environments) to the GPU nodes with reduced frequencies, and concentrate high-frequency GPU resources to process real-time tasks during peak hours. When a new computing power task is determined, the new task is preferentially allocated to the node with the optimal energy efficiency ratio (such as the GPU with the lowest power consumption per unit of computing power).

[0117] Optionally, before the step S2 of inputting the metric data into the target load prediction model to predict the load data of the computing power cloud platform at the second time point to obtain the second load data, the method further includes:

[0118] Step S5: Obtain a sample data set from the historical resource database. The sample data set includes a plurality of historical metric data corresponding to a plurality of historical time points. Each historical metric data in the plurality of historical metric data includes historical load data and historical system data. The historical load data is used to indicate the load status of the computing power cloud platform at the historical time point, and the historical system data is used to indicate the hardware parameters of the computing power cloud platform corresponding to the load status at the historical time point;

[0119] Step S6: Train the load prediction model based on the sample data set to obtain the target load prediction model.

[0120] In this embodiment, before predicting the metric data, the load prediction model needs to be trained to obtain the target load prediction model.

[0121] Specifically, a sample data set can be obtained from the historical resource database. The sample data set includes multiple historical metric data corresponding to multiple historical time points, and each historical metric data in the multiple historical metric data includes historical load data and historical system data. Thus, by training the load prediction model with the sample data set, a target load prediction model is finally obtained. It should be further noted that this application has an online learning mechanism, which can dynamically adjust the model weights according to the prediction error, so as to update the target load prediction model, enabling the target load prediction model to better predict the metric data.

[0122] After obtaining the monitoring metric data corresponding to the first time point, the present invention uses the target load prediction model to predict the load data of the computing power cloud platform at the second time point, obtaining the second load data, and then adjusting the system data to the second system data at the second time point, so that the computing power cloud platform achieves the effect of significantly reducing the computing power resource consumption and economic cost of the computing power cloud platform on the premise of ensuring performance, which is beneficial to the wide application of inclusive computing power.

[0123] Please refer to Figure 3 , Figure 3 is the structural diagram of an energy-saving device for a computing power cloud platform of an inclusive computing power intelligent computing center provided by the present invention. As Figure 3 shown, the energy-saving device 300 for a computing power cloud platform of an inclusive computing power intelligent computing center includes:

[0124] An acquisition module 310, configured to acquire metric data corresponding to the computing power cloud platform at the first time point. The metric data includes first load data and first system data. The first load data is used to indicate the load status of the computing power cloud platform at the first time point, and the first system data is used to indicate the hardware parameters of the computing power cloud platform corresponding to the load status at the first time point;

[0125] A prediction module 320, configured to input the metric data into the target load prediction model to predict the load data of the computing power cloud platform at the second time point, obtaining the second load data. The second time point is a time point after the first time point;

[0126] A determination module 330, configured to determine the second system data corresponding to the computing power cloud platform at the second time point according to the second load data;

[0127] An adjustment module 340, configured to adjust the system data of the computing power cloud platform to the second system data at the second time point.

[0128] Optionally, the first load data includes first performance data and first task data. The first performance data is the system performance data of the computing power cloud platform for processing computing power tasks at the first time point, and the first task data is the task information data of the computing power cloud platform for processing computing power tasks at the first time point. The acquisition module 310 includes:

[0129] A first acquisition sub-module, configured to acquire the first performance data corresponding to the computing power cloud platform at the first time point. The first performance data includes at least one of the following: the utilization rate of the graphics processing unit (GPU) cores and the occupancy rate of the video memory bandwidth;

[0130] A second acquisition sub-module, configured to acquire the first task data corresponding to the computing power cloud platform at the first time point. The first task data includes at least one of the following: task queue information, task size information, and task processing time information. Among them, the task queue information includes multiple target computing power tasks, the task size information is the task volume size corresponding to each target computing power task among the multiple target computing power tasks, and the task processing time information is the time consumed for processing each target computing power task among the multiple target computing power tasks;

[0131] A third acquisition sub-module, configured to acquire the first system data corresponding to the computing power cloud platform at the first time point.

[0132] Optionally, the prediction module 320 includes:

[0133] A prediction sub-module, configured to input the first performance data, the first task data, and the first system data into a target load prediction model, predict the load data of the computing power cloud platform at the second time point, and obtain the second load data and a target confidence level. The target confidence level is used to indicate the prediction confidence level of the target load prediction model for the prediction of the second load data.

[0134] Optionally, the determination module 330 includes:

[0135] A first determination sub-module, configured to determine the second system data corresponding to the computing power cloud platform at the second time point based on the target confidence level and the second load data.

[0136] Optionally, the adjustment module 340 includes:

[0137] A second determination sub-module, configured to determine the third system data of the computing power cloud platform at the second time point. The third system data is the system data of the computing power cloud platform at the second time point;

[0138] A second determination sub-module, configured to determine the difference between the second system data and the third system data;

[0139] An adjustment sub-module, configured to adjust the third system data multiple times according to the difference within a target time period until the third system data is adjusted to the second system data. In the multiple adjustments, each adjustment adjusts the third system data by a preset percentage. The start time point of the target time period is the second time point.

[0140] Optionally, the adjustment module 340 includes:

[0141] A first update sub-module, configured to update the GPU voltage parameter based on the second system data at the second time point, so as to update the third system data of the computing power cloud platform to the second system data. The third system data is the system data of the computing power cloud platform at the second time point;

[0142] And / or,

[0143] A second update sub-module, configured to call an application programming interface based on the second system data at the second time point, so as to update the third system data of the computing power cloud platform to the second system data.

[0144] Optionally, it further includes:

[0145] A sample acquisition module, configured to acquire a sample data set from a historical resource database. The sample data set includes multiple historical metric data corresponding to multiple historical time points. Each historical metric data in the multiple historical metric data includes historical load data and historical system data. The historical load data is used to indicate the load status of the computing power cloud platform at the historical time point, and the historical system data is used to indicate the hardware parameters of the computing power cloud platform corresponding to the load status at the historical time point;

[0146] A model training module, configured to train a load prediction model based on the sample data set to obtain the target load prediction model.

[0147] Optionally, the first system data includes at least one of the following: GPU temperature, GPU power consumption, GPU voltage, and GPU frequency. The second system data includes at least one of the following: GPU temperature, GPU power consumption, GPU voltage, and GPU frequency.

[0148] After acquiring the monitoring metric data corresponding to the first time point, the present invention uses the target load prediction model to predict the load data of the computing power cloud platform at the second time point to obtain the second load data, so as to adjust the system data to the second system data at the second time point, thereby achieving the effect of greatly reducing the computing power resource consumption and economic cost of the computing power cloud platform while ensuring performance, which is beneficial to the wide application of inclusive computing power.

[0149] An embodiment of this application also provides an electronic device. Please refer to Figure 4 , the electronic device may include a processor 401, a memory 402, and a program 4021 stored in the memory 402 and executable on the processor 401.

[0150] When the program 4021 is executed by the processor 401, it can implement Figure 1 any step in the corresponding method embodiment:

[0151] Step S1: Obtain the index data corresponding to the computing power cloud platform at the first time point, where the index data includes first load data and first system data, the first load data is used to indicate the load status of the computing power cloud platform at the first time point, and the first system data is used to indicate the hardware parameters corresponding to the load status of the computing power cloud platform at the first time point;

[0152] Step S2: Input the index data into the target load prediction model to predict the load data of the computing power cloud platform at the second time point, and obtain second load data, where the second time point is a time point after the first time point;

[0153] Step S3: Determine the second system data corresponding to the computing power cloud platform at the second time point according to the second load data;

[0154] Step S4: At the second time point, adjust the system data of the computing power cloud platform to the second system data.

[0155] Optionally, the first load data includes first performance data and first task data, the first performance data is the system performance data of the computing power cloud platform for processing computing power tasks at the first time point, and the first task data is the task data of the computing power cloud platform for processing computing power tasks at the first time point. The step S1, obtaining the index data corresponding to the computing power cloud platform at the first time point, includes:

[0156] Step S11: Obtain the first performance data corresponding to the computing power cloud platform at the first time point, where the first performance data includes at least one of the following: the utilization rate of the graphics processing unit (GPU) core and the occupancy rate of the video memory bandwidth;

[0157] Step S12: Obtain the first task data corresponding to the computing power cloud platform at the first time point. The first task data includes at least one of the following: task queue information, task size information, and task processing time information. Among them, the task queue information includes multiple target computing power tasks, the task size information is the task volume size corresponding to each target computing power task among the multiple target computing power tasks, and the task processing time information is the time consumed for processing each target computing power task among the multiple target computing power tasks;

[0158] Step S13: Obtain the first system data corresponding to the computing power cloud platform at the first time point.

[0159] Optionally, step S2: Input the metric data into the target load prediction model to predict the load data of the computing power cloud platform at the second time point, and obtain the second load data, including:

[0160] Step S21: Input the first performance data, the first task data, and the first system data into the target load prediction model to predict the load data of the computing power cloud platform at the second time point, and obtain the second load data and the target confidence level. The target confidence level is used to indicate the prediction confidence level of the target load prediction model for the prediction of the second load data.

[0161] Optionally, step S3: Determine the second system data corresponding to the computing power cloud platform at the second time point according to the second load data, including:

[0162] Step S31: Based on the target confidence level and the second load data, determine the second system data corresponding to the computing power cloud platform at the second time point.

[0163] Optionally, step S4: Adjust the system data of the computing power cloud platform to the second system data at the second time point, including:

[0164] Step S41: At the second time point, determine the third system data of the computing power cloud platform. The third system data is the system data of the computing power cloud platform at the second time point;

[0165] Step S42: Determine the difference between the second system data and the third system data;

[0166] Step S43: Within the target time period, adjust the third system data multiple times according to the difference until the third system data is adjusted to the second system data. In the multiple adjustments, each adjustment adjusts the third system data by a preset percentage. The start time point of the target time period is the second time point.

[0167] Optionally, step S4 of adjusting the system data of the computing power cloud platform to the second system data at the second time point includes:

[0168] Step S44: At the second time point, update the GPU voltage parameter based on the second system data to update the third system data of the computing power cloud platform to the second system data, where the third system data is the system data of the computing power cloud platform at the second time point;

[0169] And / or

[0170] Step S45: At the second time point, call the application programming interface based on the second system data to update the third system data of the computing power cloud platform to the second system data.

[0171] Optionally, before step S2 of inputting the metric data into the target load prediction model to predict the load data of the computing power cloud platform at the second time point to obtain the second load data, the method further includes:

[0172] Step S5: Obtain a sample data set from the historical resource database. The sample data set includes multiple historical metric data corresponding to multiple historical time points. Each historical metric data in the multiple historical metric data includes historical load data and historical system data. The historical load data is used to indicate the load status of the computing power cloud platform at the historical time point, and the historical system data is used to indicate the hardware parameters of the computing power cloud platform corresponding to the load status at the historical time point;

[0173] Step S6: Train the load prediction model based on the sample data set to obtain the target load prediction model.

[0174] Optionally, the first system data includes at least one of the following: GPU temperature, GPU power consumption, GPU voltage, and GPU frequency, and the second system data includes at least one of the following: GPU temperature, GPU power consumption, GPU voltage, and GPU frequency.

[0175] After obtaining the monitoring metric data corresponding to the first time point, the present invention uses the target load prediction model to predict the load data of the computing power cloud platform at the second time point to obtain the second load data, and then adjusts the system data to the second system data at the second time point, so that the computing power cloud platform achieves the effect of significantly reducing the computing power resource consumption and economic cost of the computing power cloud platform while ensuring performance, which is beneficial to the wide application of inclusive computing power.

[0176] The embodiments of the present application further provide a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements each process of the above-mentioned energy-saving method embodiment for the computing power cloud platform of the inclusive computing power intelligent computing center, and can achieve the same technical effects. To avoid repetition, it will not be elaborated here. Among them, the computer-readable storage medium includes, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc, etc.

[0177] The embodiments of the present application further provide a computer program product. The computer program product is stored in a storage medium. The computer program product is executed by at least one processor to implement each process of the above-mentioned energy-saving method embodiment for the computing power cloud platform of the inclusive computing power intelligent computing center, and can achieve the same technical effects. To avoid repetition, it will not be elaborated here.

[0178] It should be noted that in this article, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such process, method, article or device. Without more limitations, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, article or device including that element.

[0179] 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 a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disc), and includes several instructions to enable a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0180] The embodiments of the present application have been described above in conjunction with the accompanying drawings. However, the present application is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present application, those of ordinary skill in the art can also make many forms without departing from the purpose of the present application and the scope protected by the claims, and all of them belong to the protection scope of the present application.

Claims

1. An energy-saving method for a computing power cloud platform for a universal computing power intelligent computing center, characterized in that: The method comprises: Step S1, obtaining indicator data corresponding to the computing power cloud platform at a first time point, the indicator data including first load data and first system data, the first load data being used to indicate the load state of the computing power cloud platform at the first time point, and the first system data being used to indicate the hardware parameters of the computing power cloud platform corresponding to the load state at the first time point; Step S2: input the index data into the target load prediction model to predict the load data of the computing power cloud platform at a second time point to obtain second load data, where the second time point is a time point after the first time point; Step S3: determining the second system data corresponding to the computing power cloud platform at the second time point according to the second load data; Step S4: At the second time point, the system data of the computing power cloud platform is adjusted to the second system data.

2. The method according to claim 1, characterized in that The first load data includes first performance data and first task data, the first performance data is system performance data of the computing power cloud platform processing the computing power task at the first time point, and the first task data is task information data of the computing power cloud platform processing the computing power task at the first time point. The step S1, obtaining the indicator data corresponding to the computing power cloud platform at the first time point, includes: Step S11, obtaining first performance data corresponding to the computing power cloud platform at the first time point, wherein the first performance data includes at least one of the following: a graphics processor GPU core utilization rate and a video memory bandwidth occupancy rate; Step S12: obtaining first task data corresponding to the computing power cloud platform at the first time point, wherein the first task data includes at least one of the following: task queue information, task size information, and task processing time information, wherein the task queue information includes a plurality of target computing power tasks, the task size information is the task size corresponding to each target computing power task in the plurality of target computing power tasks, and the task processing time information is the time consumed for processing each target computing power task in the plurality of target computing power tasks; Step S13: Obtain the first system data corresponding to the computing power cloud platform at the first time point.

3. The method according to claim 2, characterized in that The step S2, inputting the indicator data into the target load prediction model, predicting the load data of the computing power cloud platform at a second time point, and obtaining the second load data, includes: Step S21: input the first performance data, the first task data and the first system data into a target load prediction model, predict the load data of the computing power cloud platform at a second time point, and obtain the second load data and target confidence, wherein the target confidence is used to indicate the prediction confidence of the target load prediction model for the second load data.

4. The method according to claim 3, characterized in that The step S3, determining the second system data corresponding to the computing power cloud platform at the second time point according to the second load data, includes: Step S31: Based on the target confidence and the second load data, determine the second system data corresponding to the computing power cloud platform at the second time point.

5. The method according to claim 1, characterized in that The step S4, adjusting the system data of the computing power cloud platform to the second system data at the second time point, includes: Step S41: at the second time point, determining the third system data of the computing power cloud platform, wherein the third system data is the system data of the computing power cloud platform at the second time point; Step S42: determining a difference between the second system data and the third system data; Step S43: within the target time period, the third system data is adjusted multiple times according to the difference until the third system data is adjusted to the second system data, and in the multiple adjustments, the third system data is adjusted by a preset percentage each time, and the starting time point of the target time period is the second time point.

6. The method according to claim 1, characterized in that The step S4, adjusting the system data of the computing power cloud platform to the second system data at the second time point, includes: Step S44: at the second time point, updating the GPU voltage parameter based on the second system data, so as to update the third system data of the computing power cloud platform to the second system data, wherein the third system data is the system data of the computing power cloud platform at the second time point; and / or, Step S45: At the second time point, call the application programming interface based on the second system data to update the third system data of the computing power cloud platform to the second system data.

7. The method according to claim 1, characterized in that In step S2, inputting the indicator data into the target load prediction model to predict the load data of the computing power cloud platform at a second time point, before obtaining the second load data, the method further includes: Step S5: Acquire a sample data set in a historical resource database, the sample data set including multiple historical indicator data corresponding to multiple historical time points, each of the multiple historical indicator data including historical load data and historical system data, the historical load data is used to indicate the load state of the computing power cloud platform at the historical time point, and the historical system data is used to indicate the hardware parameters of the computing power cloud platform corresponding to the load state at the historical time point; Step S6: training the load prediction model based on the sample data set to obtain the target load prediction model.

8. The method according to any one of claims 1 to 7, characterized in that: The first system data includes at least one of the following: GPU temperature, GPU power consumption, GPU voltage and GPU frequency; the second system data includes at least one of the following: GPU temperature, GPU power consumption, GPU voltage and GPU frequency.

9. An energy-saving device for a computing power cloud platform of a universal computing power intelligent computing center, characterized in that: The device comprises: An acquisition module, used to acquire indicator data corresponding to the computing power cloud platform at a first time point, the indicator data including first load data and first system data, the first load data being used to indicate the load state of the computing power cloud platform at the first time point, and the first system data being used to indicate the hardware parameters of the computing power cloud platform corresponding to the load state at the first time point; A prediction module, used for inputting the indicator data into a target load prediction model to predict the load data of the computing power cloud platform at a second time point to obtain second load data, where the second time point is a time point after the first time point; A determination module, configured to determine second system data corresponding to the computing power cloud platform at the second time point according to the second load data; An adjustment module is used to adjust the system data of the computing power cloud platform to the second system data at the second time point.

10. 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 energy-saving method for a computing power cloud platform for a universal computing power intelligent computing center as described in any one of claims 1 to 8 are implemented.

11. 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 energy-saving method for a computing power cloud platform for a universal computing power intelligent computing center as described in any one of claims 1 to 8.

12. A computer program product, characterized in that It includes computer instructions, which, when executed by a processor, implement the steps of the energy-saving method for the computing power cloud platform of the inclusive computing power intelligent computing center as described in any one of claims 1 to 8.

Images