Super-large-scale intelligent computing center machine room cluster partition planar layout planning method

Abstract

The application belongs to the technical field of infrastructure planning, and discloses a super-large-scale intelligent computing center machine room cluster partition planar layout planning method; the method comprises the following steps: constructing a full-element digital twin base of the intelligent computing center machine room, which is used for real-time mapping of state information of equipment in the machine room; in the second step, the equipment in the machine room is clustered and divided, and according to the computing demand of the business, the business is sent to the corresponding cluster center for calculation; the state information and the environmental information of each device of the cluster center are collected in real time, and a comprehensive evaluation coefficient is generated to judge whether the device state is abnormal; when it is judged that the device state is abnormal, the equipment of the machine room is adjusted correspondingly; according to the load change of each device during calculation, the application can dynamically adjust the business computing task quantity and the parameters of each auxiliary device, and perform re-layout adjustment, so as to ensure the reasonable allocation of the computing power center resources and the overall computing efficiency of the computing power center.

Description

Technical Field

[0001] This invention relates to the field of infrastructure planning technology, and more specifically, to a method for planning the clustered and partitioned planar layout of ultra-large-scale intelligent computing center computer rooms. Background Technology

[0002] Ultra-large-scale intelligent computing centers (typically referring to those with ≥10,000 server racks) are the core infrastructure supporting computing power-intensive businesses such as artificial intelligence and big data analytics. Their server room layout directly affects resource utilization, energy consumption, and business stability.

[0003] Existing data center layout planning methods are mostly static designs, which involve a one-time partitioning layout based on initial business needs. For example, a uniform workload is planned for each data center device based on the workload of business tasks, and uniform auxiliary equipment is provided for auxiliary calculations. However, static layouts cannot adapt to dynamic changes in equipment load, which can easily lead to situations where some partitions are overloaded and other partitions have idle resources, resulting in uneven resource allocation.

[0004] In view of this, the present invention proposes a clustered partitioned planar layout planning method for ultra-large-scale intelligent computing center computer rooms to solve the above problems. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art and achieve the above objectives, the present invention provides the following technical solution: A method for planning the clustered, partitioned floor plan layout of ultra-large-scale intelligent computing center data centers, including: Step 1: Construct a full-element digital twin base for the intelligent computing center's computer room. The full-element digital twin base is used to map the status information of the equipment in the computer room in real time. Step 2: Divide the equipment in the computer room into multiple cluster centers, and send the business to the corresponding cluster center for calculation according to the business's computing needs; Step 3: Collect the status information and environmental information of each device in the clustering center in real time, analyze and generate a comprehensive evaluation coefficient based on the status information and environmental information, and determine whether the device status is abnormal based on the comprehensive evaluation coefficient. Step 4: When it is determined that there is an abnormality in the equipment status, make corresponding planning and adjustments to the equipment in the computer room.

[0006] Furthermore, the clustering centers include three types: high-performance clustering centers, medium-performance clustering centers, and low-performance clustering centers. Based on the physical spatial structure of the computer room, the computer room is divided into multiple grid blocks. Historical status information of each device within the grid block is obtained. The historical status information contains parameter values ​​of multiple status features. The parameter values ​​of each status feature are standardized to obtain the feature values ​​of each status feature. Based on the feature values ​​corresponding to the status features of each device, the performance score of each grid block is calculated. Each cluster center has a corresponding performance score interval. The calculated performance score of each grid block is compared with the performance score interval set by each cluster center. When the performance score falls into the performance score interval of the corresponding cluster center, the cluster center type corresponding to the cluster center is recorded as the cluster center of the grid block.

[0007] Furthermore, the performance score calculation method for grid blocks is as follows: The state features include positive state features and negative state features. The feature values ​​of each positive state feature are obtained and weighted and accumulated to obtain the first performance score. The feature values ​​of each negative state feature are obtained and weighted and accumulated to obtain the second performance score. The first performance score is divided by the second performance score to obtain the device score of each device. The average of the device scores of all devices in the grid block is calculated to obtain the performance score of the grid block.

[0008] Furthermore, the method for sending business calculations to the corresponding clustering center for calculation based on business calculation requirements is as follows: Obtain the computing requirements of the business, extract the quantitative values ​​of multiple judgment indicators from the computing requirements of the business, and construct a business requirement indicator vector based on the quantitative values ​​of each judgment indicator. Each cluster center has a pre-defined standard quantified value for each judgment indicator. Based on the standard quantified value of each judgment indicator, a standard indicator vector for the cluster center is constructed. The cosine similarity between the business requirement indicator vector and the standard indicator vector of the cluster center is calculated. The cluster center corresponding to the standard indicator vector of the cluster center with the largest cosine similarity is selected as the calculation center, and the business is sent to the corresponding cluster center for calculation.

[0009] Furthermore, the method for generating the comprehensive evaluation coefficient is as follows: The acquired state information is used to extract features to obtain real-time parameter values ​​of multiple state features. The real-time parameter values ​​of each state feature are then standardized to obtain actual parameter values ​​of each state feature. Based on the actual parameter values ​​of each state feature, the real-time device score of each device is obtained and recorded as the initial score of the device. The acquired environmental information is analyzed to generate environmental factors. The initial score is then corrected based on the environmental factors to obtain the final score of the device. The final score is compared with the minimum score threshold set for the device. If the final score is less than the minimum score threshold set for the device, the device is judged to be in an abnormal state.

[0010] Furthermore, the method for generating environmental correction factors is as follows: Environmental information is acquired and features are extracted to obtain parameter values ​​for multiple environmental features. Based on the ideal working environment of the computer room equipment, standard parameter values ​​for each environmental feature are determined. The absolute difference between the parameter values ​​of each environmental feature and the standard parameter values ​​is calculated. The absolute difference is normalized to obtain the deviation value of each environmental feature. The deviation values ​​of all environmental features are weighted and accumulated to obtain the environmental correction factor.

[0011] Furthermore, corresponding planning and adjustments to the equipment in the computer room include replanning and adjusting the workload of business computing tasks of each device in the computer room, as well as replanning and adjusting the parameters of each auxiliary device in the computer room.

[0012] Furthermore, the method for replanning and adjusting the workload of business computing tasks for each device in the computer room is as follows: Obtain the final score of each device in the corresponding grid block in the computer room, subtract the minimum score threshold from the final score to obtain the score difference of each device, and sort the devices in descending order according to the score difference to obtain the device sorting table. Retrieve the first A devices from the device sorting table to increase the workload of business calculations; retrieve the last B devices from the device sorting table to decrease the workload of business calculations.

[0013] Furthermore, the method for re-planning and adjusting the parameters of various auxiliary equipment in the computer room is as follows: When an abnormal device status is detected, the location of the corresponding device is obtained, the score difference of the corresponding device is obtained, the absolute value of the score difference is calculated, and the parameters of the auxiliary device at the corresponding device location are optimized based on the magnitude of the absolute score difference. The location of the device that has increased the workload of business computing tasks is obtained, and the parameters of the auxiliary device at the corresponding device location are optimized based on the number of increased workload of business computing tasks.

[0014] The technical effects and advantages of the present invention's method for clustered partitioned planar layout planning of ultra-large-scale intelligent computing center computer rooms are as follows: This invention can dynamically adjust the workload of business computing tasks and the parameters of each auxiliary device according to the load changes of each device during computing, and reconfigure the layout to ensure the rational allocation of computing center resources and the overall computing efficiency of the computing center. Attached Figure Description

[0015] Figure 1 This is a flowchart of the clustered partitioned planar layout planning method for ultra-large-scale intelligent computing center computer rooms according to the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Please see Figure 1 As shown in the figure, this embodiment discloses a method for clustered partitioned planar layout planning of ultra-large-scale intelligent computing center computer rooms. The method mainly includes: Step 1: Construct a full-element digital twin base for the intelligent computing center's computer room. The full-element digital twin base is used to map the status information of the equipment in the computer room in real time. Step 2: Divide the equipment in the computer room into multiple cluster centers, and send the business to the corresponding cluster center for calculation according to the business's computing needs; Step 3: Collect the status information and environmental information of each device in the clustering center in real time, analyze and generate a comprehensive evaluation coefficient based on the status information and environmental information, and determine whether the device status is abnormal based on the comprehensive evaluation coefficient. Step 4: When it is determined that there is an abnormality in the equipment status, make corresponding planning and adjustments to the equipment in the computer room.

[0018] Through the above scheme, this application first constructs a full-element digital twin base for the intelligent computing center's computer room. This full-element digital twin base is used to map the status information of the equipment within the computer room in real time, enabling real-time acquisition of equipment status information for subsequent analysis. Then, based on the historical performance scores of the equipment within the computer room, the equipment is clustered into multiple cluster centers. Based on the computing needs of the business, the business is sent to the corresponding cluster center for computation. This method of selecting the appropriate cluster center based on the business's computing needs allows for precise adaptation to different business requirements. During the computation, the status and environmental information of each device in the cluster centers are collected in real time. A comprehensive evaluation coefficient is generated based on the status and environmental information. This coefficient is used to determine whether the equipment status is abnormal. If an abnormality is detected, the equipment in the computer room needs to be replanned and adjusted, including adjusting the allocated workload of the business computing tasks and optimizing the parameters of auxiliary equipment in the abnormal area. This allows for dynamic adjustment of the workload of the business computing tasks and the parameters of each auxiliary device based on the load changes of each device during computation, enabling a reconfiguration and ensuring the rational allocation of computing center resources and the overall computing efficiency of the computing center.

[0019] Cluster centers are categorized into three types: high-performance, medium-performance, and low-performance. Based on the physical spatial structure of the data center, it is divided into multiple grid blocks. Historical state information of each device within a grid block is obtained. This historical state information contains parameter values ​​for multiple state features. These parameter values ​​are standardized to obtain feature values ​​for each state feature. Based on the feature values ​​corresponding to the various state features of each device, a performance score for each grid block is calculated. Each cluster center has a corresponding performance score interval. The calculated performance scores of each grid block are compared with the performance scores set for each cluster center. The performance score is compared within the performance score range. When the performance score falls within the performance score range of the corresponding cluster center, the cluster center type corresponding to the cluster center is recorded as the cluster center of the grid block. The performance score of the grid block is calculated as follows: the state features include positive state features and negative state features. The feature values ​​of each positive state feature are obtained and weighted and accumulated to obtain the first performance score. The feature values ​​of each negative state feature are obtained and weighted and accumulated to obtain the second performance score. The first performance score is divided by the second performance score to obtain the device score of each device. The average of the device scores of all devices in the grid block is calculated to obtain the performance score of the grid block.

[0020] The above scheme provides a specific method for clustering equipment within a data center. First, the cluster centers are categorized into three types: high-performance, medium-performance, and low-performance cluster centers. Then, based on the physical spatial structure of the data center, it is divided into multiple grid blocks. Historical state information for each device within each grid block is obtained. This historical state information includes the device's historical computing speed (average), historical CPU utilization (average), GPU utilization (average), memory usage (average), and historical power consumption (average). The historical state information contains parameter values ​​for multiple state features. These parameter values ​​are standardized to obtain feature values ​​for each state feature. State features include positive and negative features. For example, positive state features such as computing speed, CPU utilization, GPU utilization, and memory usage are obtained from the historical state information, while power consumption is a negative state feature. The parameter values ​​for these state features are normalized to obtain the feature values ​​corresponding to each state feature of each device. The feature values ​​of each positive state feature are then weighted and accumulated. The system first adds up the first performance score; then it acquires the feature values ​​of each negative state feature, assigns weights, and accumulates them to obtain the second performance score; the first performance score is divided by the second performance score to obtain the device score for each device. The higher the device score, the better the device performance. Therefore, the average of the device scores of all devices within a grid block is calculated to obtain the performance score of the grid block. Combining the device scores of all devices within a grid block yields the comprehensive performance score of the grid block. Each cluster center has a corresponding performance score interval. The calculated performance score of each grid block is compared with the performance score interval set by each cluster center. When the performance score falls into the performance score interval of the corresponding cluster center, the cluster center type corresponding to the cluster center is recorded as the cluster center of the grid block. By combining the grid blocks and the performance scores of the grid blocks to cluster devices in this way, physical space and performance capabilities can be strongly bound together, avoiding resource dispersion. Moreover, by combining positive and negative state features for comprehensive analysis, the clustering can be more objective, reducing subjective bias and improving the clustering accuracy.

[0021] The method for sending business calculations to the corresponding cluster centers for calculation based on business calculation requirements is as follows: Obtain the business calculation requirements; extract the quantified values ​​of multiple judgment indicators from the business calculation requirements; construct a business requirement indicator vector based on the quantified values ​​of each judgment indicator; each cluster center has preset standard quantified values ​​for each judgment indicator; construct a cluster center standard indicator vector based on the standard quantified values ​​of each judgment indicator; calculate the cosine similarity between the business requirement indicator vector and the cluster center standard indicator vector; select the cluster center corresponding to the standard indicator vector with the largest cosine similarity as the calculation center; and send the business to the corresponding cluster center for calculation.

[0022] The above technical solution provides a specific method for selecting appropriate clustering centers based on business computing needs. It involves obtaining the business's computing needs, extracting quantified values ​​of multiple judgment indicators from these needs, and constructing a business requirement indicator vector based on these quantified values. Judgment indicators can include computing intensity, business priority, data I / O requirements, and the maximum tolerable latency requirement rate for the corresponding business. The quantified values ​​of these judgment indicators are obtained by normalizing the parameters of the judgment indicators. Simultaneously, based on the experience of those in the field, standard quantified values ​​for each judgment indicator are preset in each clustering center, forming a standard indicator vector for the clustering centers. The cosine similarity between the business requirement indicator vector and the standard indicator vector of the clustering centers is calculated. A higher cosine similarity indicates a higher degree of similarity between the two vectors. Therefore, the clustering center corresponding to the standard indicator vector of the clustering center with the highest cosine similarity is selected as the computing center, and the business is sent to the corresponding clustering center for computation. This method enables precise adaptation to different business needs and improves resource utilization.

[0023] The comprehensive evaluation coefficient generation method is as follows: Features are extracted from the acquired state information to obtain real-time parameter values ​​for multiple state features. These real-time parameter values ​​are then standardized to obtain actual parameter values ​​for each state feature. Based on these actual parameter values, a real-time device score is obtained for each device, which is recorded as the initial device score. Environmental factors are generated based on the acquired environmental information. These environmental factors are then used to correct the initial score, resulting in the final device score. The final score is compared with the minimum score threshold set for each device. If the final score is less than the minimum score threshold, the device is judged to be in an abnormal state. The method for generating environmental correction factors is as follows: environmental information is acquired and features are extracted to obtain parameter values ​​of multiple environmental features. Based on the ideal working environment of the computer room equipment, standard parameter values ​​of each environmental feature are determined. The absolute difference between the parameter values ​​of each environmental feature and the standard parameter values ​​is calculated. The absolute difference is normalized to obtain the deviation value of each environmental feature. The deviation values ​​of all environmental features are weighted and accumulated to obtain the environmental correction factor.

[0024] The above technical solution provides a specific method for determining whether the device status within each grid block is abnormal. First, the acquired status information is feature-extracted to obtain real-time parameter values ​​for multiple status features. These real-time parameter values ​​are then standardized to obtain actual parameter values ​​for each status feature. Based on these actual parameter values, a real-time device score is obtained for each device, which is recorded as the initial score. The calculation method for each implemented device score is the same as the method described above. A higher implemented device score indicates better device computational efficiency. However, since the device computation process is easily affected by environmental factors such as temperature, relative humidity, and air quality, the set real-time device score needs to be corrected. Specifically: environmental information is acquired and feature-extracted to obtain parameter values ​​for multiple environmental features, such as temperature and relative humidity. The system considers temperature and air quality. Based on the ideal operating environment of the computer room equipment, standard parameter values ​​for various environmental characteristics are determined. The absolute difference between the parameter values ​​of each environmental characteristic and the standard parameter values ​​is calculated. The absolute difference is normalized to obtain the deviation value of each environmental characteristic. The larger the deviation value, the greater the difference between the current environmental characteristic and the ideal environmental characteristic, and the greater the impact. The deviation values ​​of all environmental characteristics are weighted and accumulated to obtain the environmental correction factor. Finally, the initial score is corrected in combination with the environmental factor to obtain the final score of the equipment. The final score is the initial score * (1 + environmental factor). Based on the historical scores of each device, a minimum score threshold is determined. The final score is compared with the minimum score threshold set for the corresponding device. When the final score is less than the set minimum score threshold, the device is judged to be in an abnormal state, and the device needs to be adjusted accordingly.

[0025] The corresponding planning and adjustment of the equipment in the computer room includes replanning and adjusting the workload of business computing tasks of each device in the computer room, as well as replanning and adjusting the parameters of each auxiliary device in the computer room. The method for replanning and adjusting the workload of business computing tasks for each device in the computer room is as follows: obtain the final score of each device in the corresponding grid block in the computer room, subtract the minimum score threshold from the final score to obtain the score difference of each device, sort each device in descending order according to the score difference to obtain the device sorting table; increase the workload of business computing tasks for the first A devices in the device sorting table, and decrease the workload of business computing tasks for the last B devices in the device sorting table. The method for replanning and adjusting the parameters of each auxiliary device in the computer room is as follows: When an abnormal device status is detected, the location of the corresponding device is obtained, the score difference of the corresponding device is obtained, the absolute value of the score difference is calculated, and the absolute score difference is obtained. Based on the magnitude of the absolute score difference, the parameters of the auxiliary device at the corresponding device location are optimized. The location of the device that increases the workload of business computing tasks is obtained, and the parameters of the auxiliary device at the corresponding device location are optimized based on the number of increased workload of business computing tasks.

[0026] The above technical solution provides a method for adjusting the planning of equipment in a data center when equipment malfunctions. This includes adjusting the workload of the data center equipment's business computing tasks and adjusting the auxiliary setting parameters of the data center. First, the final score of each device within the corresponding grid block in the data center is obtained. The final score is then subtracted from the minimum score threshold to obtain the score difference between each device. The larger the score difference, the better the device's computing performance, and vice versa. The devices are then sorted in descending order based on the score differences to obtain a device ranking table. The top A devices in the ranking table have their business computing tasks increased, while the bottom B devices have their business computing tasks decreased. This way, when equipment malfunctions, the corresponding business computing tasks can be reduced and transferred to the higher-ranked, higher-performance devices. This reduces the computing load on low-performance devices and prevents high-performance devices from being idle, thus more rationally combining workload with dynamic resource utilization efficiency to ensure overall computing efficiency. Simultaneously, when an abnormal equipment status is detected, the physical location of the corresponding equipment within the computer room, as well as the corresponding auxiliary equipment at that location (such as power supply equipment, air conditioning equipment, etc.), is obtained. The score difference of the corresponding equipment is acquired, and the absolute value of the score difference is calculated to obtain the absolute score difference. Based on the magnitude of the absolute score difference, the parameters of the auxiliary equipment at the corresponding equipment location are optimized. When the equipment status is abnormal, the larger the absolute score difference, the more severe the equipment's computational overload capacity. In this case, the parameters of the auxiliary equipment at that location can be optimized accordingly to improve the equipment's performance. For example, if the abnormal equipment location is in the southeast corner of grid 1, the precision air conditioning parameters in that area can be adjusted: the supply air temperature is reduced from 22℃ to 21℃, and the supply air velocity is increased from 1.5m / s to 1.7m / s. The magnitude of the auxiliary equipment parameter optimization adjustment can be... Adjustments are made based on the absolute difference in scores. For example, when the absolute difference in scores is too large, the air outlet angle of the air conditioner at the corresponding adjacent location can be changed to assist in processing and improve equipment performance. Similarly, for equipment with increased workload, the parameters at that location also need to be adjusted to ensure subsequent computing efficiency. Therefore, the number of increased workload is obtained, and the parameters of the auxiliary equipment at the corresponding equipment location are optimized accordingly. For example, the more workload increases, the higher the power limit of the dual-path power supply equipment should be to meet the computing power requirements. In this way, the workload of workload and the parameters of each auxiliary equipment can be dynamically adjusted according to the load changes of each device during calculation, and the layout can be rearranged to ensure the reasonable allocation of computing center resources and ensure computing efficiency.

[0027] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0028] All formulas in this manual are dimensionless and calculated numerically. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0029] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for clustered partitioned planar layout planning of ultra-large-scale intelligent computing center computer rooms, characterized in that: The methods include: Step 1: Construct a full-element digital twin base for the intelligent computing center's computer room. The full-element digital twin base is used to map the status information of the equipment in the computer room in real time. Step 2: Divide the equipment in the computer room into multiple cluster centers, and send the business to the corresponding cluster center for calculation according to the business's calculation requirements; Step 3: Collect the status information and environmental information of each device in the clustering center in real time, analyze and generate a comprehensive evaluation coefficient based on the status information and environmental information, and determine whether the device status is abnormal based on the comprehensive evaluation coefficient. Step 4: When it is determined that there is an abnormality in the equipment status, make corresponding planning and adjustments to the equipment in the computer room.

2. The method for clustered partitioned planar layout planning of ultra-large-scale intelligent computing center computer rooms according to claim 1, characterized in that, The cluster centers include three types: high-performance cluster centers, medium-performance cluster centers, and low-performance cluster centers. Based on the physical spatial structure of the computer room, the computer room is divided into multiple grid blocks. Historical status information of each device within the grid block is obtained. The historical status information contains parameter values ​​of multiple status features. The parameter values ​​of each status feature are standardized to obtain the feature values ​​of each status feature. Based on the feature values ​​corresponding to the status features of each device, the performance score of each grid block is calculated. Each cluster center has a corresponding performance score interval. The calculated performance score of each grid block is compared with the performance score interval set by each cluster center. When the performance score falls into the performance score interval of the corresponding cluster center, the cluster center type corresponding to the cluster center is recorded as the cluster center of the grid block.

3. The method for clustered partitioned planar layout planning of ultra-large-scale intelligent computing center computer rooms according to claim 2, characterized in that, The performance score calculation method for grid blocks is as follows: The state features include positive state features and negative state features. The feature values ​​of each positive state feature are obtained and weighted and accumulated to obtain the first performance score. The feature values ​​of each negative state feature are obtained and weighted and accumulated to obtain the second performance score. The first performance score is divided by the second performance score to obtain the device score of each device. The average of the device scores of all devices in the grid block is calculated to obtain the performance score of the grid block.

4. The method for clustered partitioned planar layout planning of ultra-large-scale intelligent computing center computer rooms according to claim 1, characterized in that, The method for sending business calculations to the corresponding clustering center for calculation based on business needs is as follows: Obtain the computing requirements of the business, extract the quantitative values ​​of multiple judgment indicators from the computing requirements of the business, and construct a business requirement indicator vector based on the quantitative values ​​of each judgment indicator. Each cluster center has a pre-defined standard quantified value for each judgment indicator. Based on the standard quantified value of each judgment indicator, a standard indicator vector for the cluster center is constructed. The cosine similarity between the business requirement indicator vector and the standard indicator vector of the cluster center is calculated. The cluster center corresponding to the standard indicator vector of the cluster center with the largest cosine similarity is selected as the calculation center, and the business is sent to the corresponding cluster center for calculation.

5. The method for clustered partitioned planar layout planning of ultra-large-scale intelligent computing center computer rooms according to claim 1, characterized in that, The method for generating the comprehensive evaluation coefficient is as follows: The acquired state information is used to extract features to obtain real-time parameter values ​​of multiple state features. The real-time parameter values ​​of each state feature are then standardized to obtain actual parameter values ​​of each state feature. Based on the actual parameter values ​​of each state feature, the real-time device score of each device is obtained and recorded as the initial score of the device. The acquired environmental information is analyzed to generate environmental factors. The initial score is then corrected based on the environmental factors to obtain the final score of the device. The final score is compared with the minimum score threshold set for the device. If the final score is less than the minimum score threshold set for the device, the device is judged to be in an abnormal state.

6. The method for clustered partitioned planar layout planning of ultra-large-scale intelligent computing center computer rooms according to claim 5, characterized in that, The method for generating environmental correction factors is as follows: Environmental information is acquired and features are extracted to obtain parameter values ​​for multiple environmental features. Based on the ideal working environment of the computer room equipment, standard parameter values ​​for each environmental feature are determined. The absolute difference between the parameter values ​​of each environmental feature and the standard parameter values ​​is calculated. The absolute difference is normalized to obtain the deviation value of each environmental feature. The deviation values ​​of all environmental features are weighted and accumulated to obtain the environmental correction factor.

7. The method for clustered partitioned planar layout planning of ultra-large-scale intelligent computing center computer rooms according to claim 1, characterized in that, The corresponding planning and adjustment of the equipment in the computer room includes replanning and adjusting the workload of business computing tasks of each device in the computer room, as well as replanning and adjusting the parameters of each auxiliary device in the computer room.

8. The method for clustered partitioned planar layout planning of ultra-large-scale intelligent computing center computer rooms according to claim 7, characterized in that, The method for replanning and adjusting the workload of various devices in the computer room is as follows: Obtain the final score of each device in the corresponding grid block in the computer room, subtract the minimum score threshold from the final score to obtain the score difference of each device, and sort the devices in descending order according to the score difference to obtain the device sorting table. Retrieve the first A devices from the device sorting table to increase the workload of business calculations; retrieve the last B devices from the device sorting table to decrease the workload of business calculations.

9. The method for clustered partitioned planar layout planning of ultra-large-scale intelligent computing center computer rooms according to claim 7, characterized in that, The method for re-planning and adjusting the parameters of various auxiliary equipment in the computer room is as follows: When an abnormal device status is detected, the location of the corresponding device is obtained, the score difference of the corresponding device is obtained, the absolute value of the score difference is calculated, the absolute score difference is obtained, and the parameters of the auxiliary device at the corresponding device location are optimized based on the magnitude of the absolute score difference. The system identifies the locations of devices that have increased workload on business computing tasks and optimizes the parameters of auxiliary devices at the corresponding device locations based on the number of increased workload on business computing tasks.

Images