A centralized management method, system and device of a data center cluster and a storage medium

Abstract

The present application relates to the technical field of data management, and particularly relates to a centralized management method, system and device of a data center cluster and a storage medium. The method comprises the following steps: obtaining environmental control data of at least one data center in the data center cluster; converting the obtained environmental control data into standardized environmental control data according to a preset environmental control data standard; performing dynamic scoring on the standardized environmental control data based on a real-time scoring model to obtain the importance of the standardized environmental control data, wherein the real-time scoring model performs multi-dimensional weighted calculation according to real-time attributes, historical behavior patterns and environmental perception characteristics of data; and classifying and storing the environmental control data according to the importance. Through the dynamic scoring mechanism, the system can preferentially process data with high importance, optimize storage resource allocation and reduce the processing overhead of data with low importance, so as to realize more efficient data management in a complex operation and maintenance environment.

Description

Technical Field

[0001] This invention relates to the field of data management technology, and in particular to a centralized management method, system, device, and storage medium for data center clusters. Background Technology

[0002] Data centers are globally collaborative network devices used to transmit, accelerate, display, compute, and store data information on Internet terminal facilities. To ensure the security and reliability of data center operations, each data center's server room is equipped with an environmental control system (ECS). Because each data center's ECS requires an interface from its vendor to transmit data, and the standardization capabilities of these interfaces vary significantly among vendors, currently, headquarters obtains data from each data center's ECS manually by staff traveling to each data center.

[0003] However, the current method of having staff travel to various data centers to transmit data from different environmental control systems requires visiting each data center to retrieve the data. Headquarters cannot automatically acquire data from each data center in real time, resulting in low efficiency in obtaining data from the environmental control systems of each data center. Furthermore, Chinese invention patent application CN112052287A discloses a centralized management method, device, and system for data center clusters. While it achieves automatic collection and standardized processing of environmental control data, it still relies on preset static classification standards to determine the importance of data. For example, it simply distinguishes environmental control performance data from alarm data into different importance levels. This approach fails to consider the dynamic characteristics of the data itself, such as real-time alarm frequency, data timeliness, and historical related events. This leads to a lack of flexibility and accuracy in importance assessment, making it unable to adapt to complex and ever-changing operational environments, thus limiting the optimization effect of data storage and management strategies.

[0004] Therefore, there is an urgent need in this field for a method that can assess the importance of data in real time and dynamically in order to improve the intelligence level and response efficiency of data center cluster management. Summary of the Invention

[0005] To address the aforementioned shortcomings, the present invention aims to propose a centralized management method, system, device, and storage medium for data center clusters, thereby resolving the problem that the management of existing data center clusters is difficult to adapt to complex and ever-changing operation and maintenance environments due to excessive reliance on preset static classification standards.

[0006] To achieve this objective, the present invention adopts the following technical solution: a centralized management method for a data center cluster, wherein the data center cluster includes multiple data centers, and the method includes:

[0007] Step S1: Obtain environmental control data from at least one data center in the data center cluster;

[0008] Step S2: Convert the acquired environmental control data into standardized environmental control data according to the preset environmental control data standard;

[0009] Step S3: Based on the real-time scoring model, dynamically score the standardized environmental control data to obtain the importance of the standardized environmental control data, wherein the real-time scoring model performs multi-dimensional weighted calculation based on the real-time attributes, historical behavior patterns and environmental perception characteristics of the data.

[0010] Step S4: Classify and store the environmental control data according to their importance.

[0011] Preferably, the steps for obtaining standardized environmental control data in step S2 are as follows:

[0012] Step S21: Based on the data source identifier of the original environmental control data, match the corresponding data parsing rules from the preset standardized configuration library, parse the unstructured original data, and decouple the basic data fields including at least the device identifier, monitoring point, data value and timestamp.

[0013] Step S22: The decoupled basic data fields are converted to units based on physical dimensions according to the physical dimensions defined in the standardized configuration library, and then normalized to a preset numerical range.

[0014] Step S23: The normalized data is appended with a globally unique data serial number and injected with the data source identifier to form the final standardized environmental control data.

[0015] Preferably, in step S3, when the real-time scoring model obtains the importance level, it performs the following steps:

[0016] Step S31: Calculate the real-time attribute sub-score based on the deviation of the current value of the standardized environmental control data from its baseline value and the trend of data change within a short time window;

[0017] Step S32: By comparing the current data pattern of standardized environmental control data with historical behavior patterns, calculate the rarity of its occurrence, continuously evaluate the continuous alarm status, and calculate the sub-score of historical behavior pattern.

[0018] Step S33: Calculate and obtain the asset value based on the predefined asset weights of the assets that generate the standardized environmental control data in the data center;

[0019] The environmental perception feature sub-score of the data center where the standardized environmental control data is located is calculated by using the asset value and the perception feature value of the standardized environmental control data.

[0020] Step S34: The real-time attribute sub-score, pattern sub-score, and environmental perception feature sub-score are fused by weighting to obtain the weighted score, which is used as the importance level.

[0021] Preferably, the formula for obtaining the real-time attribute sub-score is as follows:

[0022] ,in These are the proportionality coefficients, , These are the degree of deviation and the trend of change, respectively.

[0023] in , A curve steepness factor greater than 0. The current value of the standardized environmental control data. To standardize environmental control data, it is based on a rolling average of historical data. A sensitivity threshold greater than 1 is used to adjust the critical point of deviation;

[0024] , To standardize the slope of change in environmental control data within the most recent time window, This refers to the average or standard value of the historical change rate of standardized environmental control data.

[0025] Preferably, the historical behavior pattern sub-score acquisition step in step S32 is as follows:

[0026] Step S321: Obtain the rarity:

[0027] Construct the first vector set of the current data pattern and the second vector set of the historical data;

[0028] Obtain the mean vector and standard deviation vector of the data in the second vector set;

[0029] Calculate the standardized Euclidean distance between the current feature vector and the historical mean vector, map the standardized Euclidean distance to the interval [0,1], and obtain the rarity.

[0030] The formula for obtaining the standardized Euclidean distance is as follows:

[0031] ;

[0032] in This represents the i-th element in the first vector set. The vector representing the mean of the i-th element is... This represents the standard deviation vector of the i-th element;

[0033] Step S322: Obtain a persistence assessment of the alarm status:

[0034] The formula for obtaining the persistence assessment of alarm status is as follows:

[0035] ;

[0036] in This refers to the duration of the alarm corresponding to the environmental control data. This represents the upper limit of the alarm duration corresponding to the environmental control data;

[0037] Step S323: Obtain the historical behavior pattern sub-score by weighted fusion of the rareness and the persistence assessment of the alarm status.

[0038] Preferably, before obtaining the environmental perception feature sub-score in step S33, the following steps are also included:

[0039] Step S331: Construct multiple topology graphs using all devices in the data center as vertices and data center management dependencies or support relationships as edges;

[0040] Step S332: Assign different asset weights based on the asset value of the devices in the topology graph;

[0041] The environmental perception feature sub-score acquisition in step S33 includes the following steps:

[0042] Step S333: Calculate the data center's asset score based on the equipment asset price of the day and the asset weight corresponding to the equipment, and obtain the asset value based on the range in which the asset score falls.

[0043] Step S334: The perceived characteristic value is obtained by weighting and summing the CPU or bandwidth utilization of the device cluster in the data center, the business period flag, the number of concurrent alarms, and external environmental factors.

[0044] Step S335: Map the perceived feature value to a preset range value using a mapping function to obtain a first ratio, and multiply the first ratio by the asset value to obtain the environmental perception feature sub-score.

[0045] Preferably, the classification and storage management of the environmental control data in step S4 are marked as follows:

[0046] If the importance level is greater than or equal to the first score threshold, the environmental control data will be stored in a high-performance memory database and a notification will be triggered immediately.

[0047] If the importance level is lower than the first score threshold but higher than or equal to the second score threshold, the environmental control data will be stored in a message queue for streaming analysis.

[0048] If the importance level is lower than the second score threshold but higher than the third score threshold, then the environmental control data will be used for daily monitoring and batch persisted to the time series database.

[0049] If the importance level is lower than or equal to the third score threshold, the environmental control data will be stored in regular log data, compressed, and archived in cold storage.

[0050] A centralized management system for a data center cluster, using the aforementioned centralized management method for a data center cluster, includes a data acquisition module, a standardization module, an evaluation module, and a classified storage module.

[0051] The data acquisition module is used to acquire environmental control data from at least one data center in the data center cluster;

[0052] The standardization module is used to convert the acquired environmental control data into standardized environmental control data according to a preset environmental control data standard.

[0053] The evaluation module is used to dynamically score the standardized environmental control data based on a real-time scoring model to obtain the importance of the standardized environmental control data. The real-time scoring model performs multi-dimensional weighted calculations based on the real-time attributes, historical behavior patterns, and environmental perception characteristics of the data.

[0054] The classification and storage module is used to classify and store the environmental control data according to its importance.

[0055] An apparatus includes: a memory for storing a computer program; and a processor for executing the computer program to implement the centralized management method for a data center cluster.

[0056] A computer storage medium storing a computer program, which, when executed by a processor, implements the centralized management method for a data center cluster.

[0057] One of the above technical solutions has the following advantages or beneficial effects: Through a dynamic scoring mechanism, the system can prioritize the processing of data with high importance, optimize the allocation of storage resources, and reduce the processing overhead of data with low importance, thereby achieving more efficient data management in complex operation and maintenance environments. At the same time, it enhances the ability to perceive and respond to real-time events, and ultimately solves the problem of insufficient optimization of storage and management strategies caused by inaccurate evaluation of data importance in existing technologies. Attached Figure Description

[0058] Figure 1 This is a flowchart of one embodiment of the method of the present invention.

[0059] Figure 2 This is a schematic diagram of the structure of one embodiment of the system of the present invention. Detailed Implementation

[0060] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0061] In the description of embodiments of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of embodiments of the present invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0062] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0063] like Figures 1-2 As shown, a centralized management method for a data center cluster, the data center cluster comprising multiple data centers, the method comprising:

[0064] Step S1: Obtain environmental control data from at least one data center in the data center cluster;

[0065] Step S2: Convert the acquired environmental control data into standardized environmental control data according to the preset environmental control data standard;

[0066] Step S3: Based on the real-time scoring model, dynamically score the standardized environmental control data to obtain the importance of the standardized environmental control data, wherein the real-time scoring model performs multi-dimensional weighted calculation based on the real-time attributes, historical behavior patterns and environmental perception characteristics of the data; Step S4: According to the importance, classify the environmental control data and store and manage it.

[0067] Environmental control data is short for environmental and power monitoring data. It refers to data collected in real-time from various supporting equipment and environmental conditions within a data center to ensure the safe, stable, and efficient operation of data center infrastructure. Examples of such data include power data and temperature and humidity data. Existing technologies, such as Chinese invention patent application CN112052287A, disclose a centralized management method, device, and system for data center clusters. While this achieves automatic collection and standardized processing of environmental control data, the determination of data importance still relies on preset static classification standards. For example, simply distinguishing environmental control performance data from alarm data into different importance levels fails to consider the dynamic changes in the data itself and cannot adapt to complex and ever-changing operational environments, thus limiting the optimization effect of data storage and management strategies. Therefore, this invention introduces a real-time scoring model. This model uses multi-dimensional weighting based on real-time attributes, historical behavior patterns, and environmental perception characteristics to dynamically evaluate the importance of environmental control data. This setup can capture key changes in the data in real time, avoiding the rigidity of static classification and making the determination of importance more accurate and adaptive. Through a dynamic scoring mechanism, the system can prioritize the processing of highly important data, optimize storage resource allocation, and reduce the processing overhead of low-importance data. This enables more efficient data management in complex operation and maintenance environments, while enhancing the ability to perceive and respond to real-time events. Ultimately, it solves the problem of insufficient optimization of storage and management strategies caused by inaccurate data importance evaluation in existing technologies.

[0068] Preferably, the steps for obtaining standardized environmental control data in step S2 are as follows:

[0069] Step S21: Based on the data source identifier of the original environmental control data, match the corresponding data parsing rules from the preset standardized configuration library, parse the unstructured original data, and decouple the basic data fields including at least equipment identifier, monitoring point, data value and timestamp.

[0070] Step S22: The decoupled basic data fields are converted to units based on physical dimensions according to the physical dimensions defined in the standardized configuration library, and then normalized to a preset numerical range.

[0071] Step S23: The normalized data is appended with a globally unique data serial number and injected with the data source identifier and data confidence label to form the final standardized environmental control data.

[0072] Environmental control data in data center clusters typically originates from diverse devices and systems, exhibiting unstructured and highly heterogeneous data formats. For example, sensors from different vendors may output temperature or humidity data in different units, leading to difficulties in data integration and low analysis efficiency when processed directly. Therefore, step S21, by matching pre-defined parsing rules based on the data source identifier, can specifically decouple unstructured data and extract basic fields such as device identifiers, monitoring points, data values, and timestamps, ensuring that key information is accurately captured. Step S22, through defined unit conversion and normalization based on physical dimensions, eliminates differences between data sources, unifying data values ​​into a comparable range and improving data consistency and integrability. Step S23, by attaching a globally unique serial number, enhances data traceability and quality assessment capabilities.

[0073] Suppose a data center collects temperature data from multiple vendors, where sensor A outputs "95℉" in Fahrenheit for the first server room, and sensor B outputs "25℃" in Celsius for the second server room. The basic data field 1, parsed using the corresponding data parsing rules from the standardized configuration library, is:

[0074] Equipment identification: "Sensor A", monitoring point: "Computer Room 1 - Temperature", data value: "95℉" and timestamp: "2025-11-11, 10:00:00";

[0075] The basic data field 2 is: Device Identifier: "Sensor B", Monitoring Point: "Computer Room 2 - Temperature", Data Value: "25℃" and Timestamp: "2025-11-11, 10:00:00";

[0076] Because different manufacturers produce data in different units, errors can occur in subsequent data processing. Therefore, it is necessary to further unify the basic data fields using a preset physical dimension-based conversion to eliminate the differences caused by data units. For example, when converting temperature to degrees Celsius, the basic data field 1 after conversion in step S22 is represented as follows: Equipment identifier: "Sensor A", Monitoring point: "Room 1 - Temperature", Data value: "35℃", and Timestamp: "2025-11-11, 10:00:00". Then, it is normalized to a preset numerical range, such as 35℃ being normalized to the [0,1] range (e.g., 35℃ is mapped to 0.35). Finally, the serial number "KPL001" and "Basic Data Field 1" are added to form standardized environmental control data for Sensor A.

[0077] Preferably, in step S3, when the real-time scoring model obtains the importance level, it performs the following steps:

[0078] Step S31: Calculate the real-time attribute sub-score based on the deviation of the current value of the standardized environmental control data from its baseline value and the trend of data change within a short time window;

[0079] Step S32: By comparing the current data pattern of standardized environmental control data with historical behavior patterns, calculate the rarity of its occurrence, continuously evaluate the continuous alarm status, and calculate the sub-score of historical behavior pattern.

[0080] Step S33: Calculate and obtain the asset value based on the predefined asset weights of the assets that generate the standardized environmental control data in the data center;

[0081] The environmental perception feature sub-score of the data center where the standardized environmental control data is located is calculated by using the asset value and the perception feature value of the standardized environmental control data.

[0082] Step S34: The real-time attribute sub-score, pattern sub-score, and environmental perception feature sub-score are fused by weighting to obtain the weighted score, which is used as the importance level.

[0083] Static rules in existing technologies cannot cope with the complex and ever-changing realities of data center operations. For example, a brief, common temperature fluctuation alarm and a continuously escalating voltage anomaly from a core server are obviously of different importance, but static classification might treat them as equal. Therefore, this solution uses multi-dimensional dynamic assessment to accurately quantify data importance: Step S31, by calculating the deviation of the current value from the baseline value and short-term trends, can keenly capture anomalies that are occurring or worsening, rather than just focusing on the alarm status itself; Step S32, by analyzing the rarity of data patterns and the persistence of alarms, can identify potential risks that seem ordinary but are actually abnormal, as well as persistent, unresolved problems, thereby assessing the severity implied by their historical behavior; Step S33 introduces business logic, ensuring that data with a greater impact on core economic operations receives higher attention through predefined asset weights and associated impact ranges, thus linking data importance assessment with business value. Finally, the three parameters are integrated through weighted calculation to form a comprehensive, dynamic, and quantifiable importance score.

[0084] Preferably, the real-time attribute sub-scoring The formula for obtaining it is as follows:

[0085] ,in These are the proportionality coefficients, , These are the degree of deviation and the trend of change, respectively.

[0086] in , A curve steepness factor greater than 0. The current value of the standardized environmental control data. To standardize environmental control data, it is based on a rolling average of historical data. A sensitivity threshold greater than 1 is used to adjust the critical point of deviation;

[0087] , To standardize the slope of change in environmental control data within the most recent time window, This refers to the average or standard value of the historical change rate of standardized environmental control data.

[0088] in This is in response to concerns about the current state of standardized environmental control data. To monitor the changing trends of standardized environmental control data, real-time attribute sub-scoring is achieved by collecting... and This allows for real-time monitoring of the attributes of standardized environmental control data, providing a solid data foundation for subsequent importance assessments. It's worth noting that different standardized environmental control data sets contain varying data types, such as equipment power and ambient temperature. Since the change rates of these different data types have significantly different dimensions, the slope of change cannot be directly used for comparison during subsequent integration and comparison. Therefore, it is necessary to... The function input is mapped to [0, 1).

[0089] Preferably, the historical behavior pattern sub-score acquisition step in step S32 is as follows:

[0090] Step S321: Obtain the rarity:

[0091] Construct the first vector set of the current data pattern and the second vector set of the historical data;

[0092] Obtain the mean vector and standard deviation vector of the data in the second vector set;

[0093] Calculate the standardized Euclidean distance between the current feature vector and the historical mean vector, map the standardized Euclidean distance to the interval [0,1], and obtain the rarity.

[0094] Among them, the standardized Euclidean distance The formula for obtaining it is as follows:

[0095] ;

[0096] in This represents the i-th element in the first vector set. The vector representing the mean of the i-th element is... This represents the standard deviation vector of the i-th element;

[0097] Step S322: Obtain a persistence assessment of the alarm status:

[0098] The continuous assessment of alarm status The formula is as follows:

[0099] ;

[0100] in This refers to the duration of the alarm corresponding to the environmental control data. This represents the upper limit of the alarm duration corresponding to the environmental control data;

[0101] Step S323: Obtain the historical behavior pattern sub-score by weighted fusion of the rareness and the persistence assessment of the alarm status.

[0102] To quantify the anomaly level of standardized environmental control data and the persistence of an alarm status, this invention calculates two key indicators—"rarity" and "persistence assessment"—and weights and fuses them to achieve in-depth and dynamic analysis of historical data behavior. Specifically, it constructs a set of current and historical data vectors and calculates the standardized Euclidean distance between the current feature vector and the historical mean vector to obtain rarity. This method, compared to ordinary Euclidean distance, considers the data fluctuation characteristics of different monitoring indicators (vector dimensions) (standardized through the standard deviation vector), ensuring that the calculation result is not dominated by the large range of a single indicator value, thus more fairly and accurately measuring the rarity of the current pattern relative to historical normal patterns. Simultaneously, persistence assessment is performed by using the ratio of alarm duration to its upper limit, effectively identifying long-standing, unresolved "stubborn" alarms. These alarms often indicate potential fundamental problems or vulnerabilities in the operation and maintenance process. By weighting and integrating rarity and persistence assessments, the pattern sub-score can capture both sudden, rare anomalous patterns (which may indicate sudden failures) and persistent, unrecoverable anomalous states, thus providing profound insights into the data behavior pattern dimension for the final importance score.

[0103] A specific example is as follows: Suppose a data point is a "rack entrance temperature" monitoring point in a central computer room. The mean vector of its historical normal data (the second vector set) is 22℃, and the standard deviation vector is 1.5℃. At a certain moment, the data at this point exhibits a current pattern (the first vector), with a feature vector of 28℃. The standardized Euclidean distance between this point and the historical mean vector is calculated to be 4. Mapping this distance to the interval [0,1] yields a very low rarity (e.g., 0.3).

[0104] Meanwhile, the temperature anomaly triggered a "high temperature alarm" that lasted for 120 minutes, while the system's maximum duration for this type of alarm is 180 minutes. Therefore, the persistence assessment is 120 / 180 = 0.67. Finally, the rarity (0.95) and persistence assessment (0.67) are weighted and fused with preset weights (e.g., 0.6 and 0.4) to obtain the pattern sub-score for this data point as 0.95×0.6 + 0.67×0.4 = 0.838.

[0105] Preferably, before obtaining the environmental perception feature sub-score in step S33, the following steps are also included:

[0106] Step S331: Construct multiple topology graphs using all devices in the data center as vertices and data center management dependencies or support relationships as edges;

[0107] Step S332: Assign different asset weights based on the asset value of the devices in the topology graph;

[0108] The environmental perception feature sub-score acquisition in step S33 includes the following steps:

[0109] Step S333: Calculate the data center's asset score based on the equipment asset price of the day and the asset weight corresponding to the equipment, and obtain the asset value based on the range in which the asset score falls.

[0110] Step S334: The perceived characteristic value is obtained by weighting and summing the CPU or bandwidth utilization of the device cluster in the data center, the business period flag, the number of concurrent alarms, and external environmental factors.

[0111] The business period flag is set to 1 during peak business hours and 0 during other business hours.

[0112] External environmental factors are determined based on the operating environment conditions of the equipment cluster. If factors such as temperature and humidity fall within a suitable range, a preset value is used. If they are outside the suitable range, the preset value is adjusted based on the ratio between the operating environment and the suitable range. For example, the suitable temperature range for server operation is 15~30℃. If the sensor detects that the temperature in the server room falls within 15~30℃, the external environmental factor is set to the preset value. If the temperature in the server room is 40℃, which is 33% higher than the maximum temperature, then 67% × the preset value is used as the external environmental factor value.

[0113] Finally, the perceived characteristic value is obtained by weighting and summing the CPU or bandwidth utilization of the data center equipment cluster, the business period flag, the number of concurrent alarms, and external environmental factors.

[0114] Step S335: Map the perceived feature value to a preset range value using a mapping function to obtain a first ratio, and multiply the first ratio by the asset value to obtain the environmental perception feature sub-score.

[0115] The importance of data within data centers does not exist in isolation, but is deeply embedded in a complex physical and logical environment. Traditional methods only look at the data itself and ignore the context in which it was generated, resulting in incomplete risk assessment.

[0116] Before acquiring environmental awareness feature sub-scores, a topology map is constructed based on the connectivity relationships between devices. This abstracts the data center into an interconnected network, allowing for a formal characterization of management dependencies or support relationships between devices. For example, the failure of a core switch may cause dozens of downstream servers to lose connectivity, and each server corresponds to different devices in different regions, such as conveyor belts or robotic arms. Then, based on this topology map, differentiated asset weights are assigned to devices, ensuring that data from core business devices receives higher fundamental importance.

[0117] Then, step S333 is executed, further introducing dynamic asset prices to ensure that asset scores reflect the economic value or business criticality of equipment at a specific point in time. When acquiring the equipment asset price for the day, a big data model can be introduced. By focusing on the daily price of equipment assets through this model, the daily asset price can be easily obtained. Then, the asset score for the entire data center is obtained through the corresponding asset weights. Since the difference in asset scores can vary significantly across different data centers, severely impacting the environmental perception feature sub-score, it is necessary to obtain the asset value based on the range within which the asset score falls, thereby reducing the influence of the asset score.

[0118] In addition, this invention incorporates a perception feature value to adjust the asset value. The perception feature value integrates the data center's internal load (CPU / broadband utilization), time sensitivity (business period marker), overall system pressure (number of concurrent alarms), and even external risks (external environmental factors) to construct a comprehensive perception feature value. This perception feature value can dynamically quantify the operational level of the data center. After weighted fusion with the asset value, a clear environmental perception feature sub-score can be obtained.

[0119] Preferably, the classification and storage management of the environmental control data in step S4 are marked as follows:

[0120] If the importance level is greater than or equal to the first score threshold, the environmental control data will be stored in a high-performance in-memory database (such as Redis), and a notification will be triggered immediately.

[0121] If the importance level is lower than the first score threshold but higher than or equal to the second score threshold, the environmental control data will be stored in a message queue (such as Kafka / RabbitMQ) for streaming analysis.

[0122] If the importance level is lower than the second score threshold but higher than the third score threshold, then the environmental control data will be used for daily monitoring and batch persisted to the time series database.

[0123] If the importance level is lower than or equal to the third score threshold, the environmental control data will be stored in regular log data, compressed, and archived in cold storage.

[0124] A centralized management system for a data center cluster, using the aforementioned centralized management method for a data center cluster, includes a data acquisition module, a standardization module, an evaluation module, and a classified storage module.

[0125] The data acquisition module is used to acquire environmental control data from at least one data center in the data center cluster;

[0126] The standardization module is used to convert the acquired environmental control data into standardized environmental control data according to a preset environmental control data standard.

[0127] The evaluation module is used to dynamically score the standardized environmental control data based on a real-time scoring model to obtain the importance of the standardized environmental control data. The real-time scoring model performs multi-dimensional weighted calculations based on the real-time attributes, historical behavior patterns, and environmental perception characteristics of the data.

[0128] The classification and storage module is used to classify and store the environmental control data according to its importance.

[0129] An apparatus includes: a memory for storing a computer program; and a processor for executing the computer program to implement the centralized management method for a data center cluster.

[0130] A computer storage medium storing a computer program, which, when executed by a processor, implements the centralized management method for a data center cluster.

[0131] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0132] 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 centralized management method of a data center cluster including a plurality of data centers, characterized by, The method comprises: Step S1: obtaining the environmental control data of at least one data center in the data center cluster; Step S2: converting the obtained environmental control data into standardized environmental control data according to a preset environmental control data standard; Step S3: dynamically scoring the standardized environmental control data based on a real-time scoring model to obtain the importance of the standardized environmental control data, wherein the real-time scoring model performs multi-dimensional weighted calculation according to real-time attributes, historical behavior patterns and environmental perception characteristics of data; and Step S4: classifying and storing the environmental control data according to the importance. When obtaining the importance, the real-time scoring model in step S3 performs the following steps: Step S31: calculating a real-time attribute sub-score based on the deviation of the current value of the standardized environmental control data from the reference value thereof and the change trend of the data in a short-term time window; Step S32: calculating the rarity of the current data pattern of the standardized environmental control data by comparing it with historical behavior patterns, and continuously evaluating the alarm state to calculate a historical behavior pattern sub-score; Step S33: calculating an asset value based on the predefined asset weight of the asset generating the standardized environmental control data in the data center; An environmental perception characteristic sub-score of the data center where the standardized environmental control data is located is calculated based on the asset value and the perception characteristic value of the standardized environmental control data; Step S34: fusing the real-time attribute sub-score, the pattern sub-score and the environmental perception characteristic sub-score by weighting to obtain a weighted score, and taking the weighted score as the importance; The real-time attribute subscore The acquisition formula is as follows: wherein are respectively proportional coefficients, , are respectively a deviation degree and a variation trend; wherein , is a curve steepness factor greater than 0, is a current value of the normalized environmental data, is a rolling average value of the normalized environmental data based on historical data, is a sensitivity threshold greater than 1 for adjusting a critical point of deviation; , is the normalized change rate of the environmental control data in the recent time window, is the average or standard value of the normalized change rate of the environmental control data.

2. The method of claim 1, wherein, The step of obtaining the standardized environmental control data in step S2 is as follows: Step S21: according to the data source identifier of the original environmental control data, matching the corresponding data parsing rule from the preset standardized configuration library, parsing the unstructured original data, and decoupling at least the basic data fields including device identifier, monitoring point, data value and timestamp; Step S22: performing unit conversion based on physical dimensions on the decoupled basic data fields according to the physical dimensions defined in the standardized configuration library, and normalizing them to a preset numerical interval; Step S23: attaching a globally unique data serial number to the normalized data, and injecting the data source identifier to form the final standardized environmental control data.

3. The method of claim 1, wherein, The step of obtaining the historical behavior pattern sub-score in step S32 is as follows: Step S321: obtaining the rarity: constructing a first vector set of the current data pattern and a second vector set of historical data; obtaining the data mean vector and standard deviation vector in the second vector set; calculating the standardized Euclidean distance between the current feature vector and the historical mean vector, mapping the standardized Euclidean distance to the interval [0, 1] to obtain the rarity; where the normalized Euclidean distance The acquisition formula is as follows: ； wherein denotes the i-th element of the first vector sum, denotes the mean vector of the i-th element, said denotes the standard deviation vector of the i-th element; Step S322: obtaining the continuity evaluation of the alarm state: wherein the persistence of the alarm condition is assessed The acquisition formula is as follows: ； wherein is the duration of the alarm corresponding to the environmental control data that has continued, is the upper limit value of the alarm duration corresponding to the environmental control data. Step S323: fusing the rarity and the continuity evaluation of the alarm state by weighting to obtain the historical behavior pattern sub-score.

4. The centralized management method of a data center cluster according to claim 1, characterized in that, The step S33 further comprises the following steps before obtaining the environment perception feature sub-score: Step S331: taking all devices in the data center as vertices and taking data center management dependency or support relationship as edges to construct a plurality of topology graphs; Step S332: assigning different asset weights based on the asset value of the device in the topology graph; The step S33 comprises the following steps for obtaining the environment perception feature sub-score: Step S333: calculating the asset score of the data center based on the asset price of the device on the day and the asset weight corresponding to the device, and obtaining the asset value based on the interval in which the asset score falls; Step S334: obtaining the perception feature value by weighted addition of the CPU or bandwidth utilization of the device cluster in the data center, the business time period flag, the number of concurrent alarms, and the external environment factor; Step S335: mapping the perception feature value to a preset range value through a mapping function to obtain a first proportion, and multiplying the first proportion by the asset value to obtain the environment perception feature sub-score.

5. The method of claim 1, wherein, The annotation of the step S4 of classifying and storing the environmental control data is as follows: If the importance is higher than or equal to the first score threshold, the environmental control data is stored in a high-performance memory database, and a notification is triggered immediately; If the importance is lower than the first score threshold and higher than the second score threshold, the environmental control data is stored in a message queue for stream analysis; If the importance is lower than the second score threshold and higher than the third score threshold, the environmental control data is routine monitoring data, which is batched and persisted to a time series database; If the importance is lower than or equal to the third score threshold, the environmental control data is stored in a regular log data and is archived to cold storage after compression.

6. A centralized management system of a data center cluster, characterized by, The centralized management method of the data center cluster according to any one of claims 1-5 comprises a data acquisition module, a standardization module, an evaluation module, and a classification storage module; The data acquisition module is used to acquire environmental control data of at least one data center in the data center cluster; The standardization module is used to convert the acquired environmental control data into standardized environmental control data according to a preset environmental control data standard; The evaluation module is used to dynamically score the standardized environmental control data based on a real-time scoring model to obtain the importance of the standardized environmental control data, wherein the real-time scoring model performs multi-dimensional weighted calculation according to real-time attributes, historical behavior patterns, and environment perception features of the data; The classification storage module is used to classify and store the environmental control data according to the importance.

7. An apparatus, comprising: It comprises: a memory for storing a computer program; a processor for executing the computer program to implement the centralized management method of the data center cluster according to any one of claims 1-5.

8. A computer storage medium, characterized in that, The computer storage medium stores a computer program, and the computer program is executed by the processor to implement the centralized management method of the data center cluster according to any one of claims 1-5.

Images