IDC intelligent operation and maintenance management system based on digital twinning

Abstract

This invention discloses an IDC intelligent operation and maintenance management system based on digital twins, belonging to the field of IDC intelligent operation and maintenance management technology. The system collects equipment data and real-time operational data from the IDC to obtain equipment metadata and real-time operational metadata; it constructs a three-dimensional digital model by performing virtual-physical mapping based on the equipment metadata; it identifies and predicts latent faults of each physical entity based on the real-time operational metadata, generating fault warning messages; and it displays the three-dimensional digital model and the corresponding real-time operational data and fault warning messages for each physical entity. Through digital twin technology, the system can promptly reflect the operational status of the data center's hardware and software environment, providing comprehensive data and information, offering reliable basis for decision-makers, and solving the problem of weak correlation between current DCIM operation and maintenance platforms and real-time operational data of various devices. By using an AI algorithm module to identify, predict, and alarm on latent faults of each physical entity in the IDC, the system improves the efficiency of fault identification and repair.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance management technology for data centers (IDCs), and specifically to an intelligent operation and maintenance management system for data centers based on digital twins. Background Technology

[0002] An Internet Data Center (IDC) is a complex complex that provides an operating environment for electronic information equipment. It includes IT infrastructure such as data processing, data transmission, and network communication, as well as site infrastructure such as power, air conditioning, and transmission lines to support the IT equipment. The operation and maintenance of an IDC requires centralized management of critical data center equipment through Data Center Infrastructure Management (DCIM).

[0003] Currently, there is a weak correlation between the DCIM operation and maintenance platform and the real-time operation data of various devices. It is necessary for maintenance personnel to read meters or export historical data from the background for various devices in order to achieve a digital presentation of the 3D model. Moreover, maintenance personnel can only deal with explicit faults. This approach leads to relatively delayed fault repair. Summary of the Invention

[0004] The purpose of this invention is to solve the problems mentioned above in the background technology and to propose an IDC intelligent operation and maintenance management system based on digital twins.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] This invention provides an IDC intelligent operation and maintenance management system based on digital twins, including a physical entity data acquisition module, a big data processing module, a digital twin model, an AI algorithm module, and a visualization module; wherein:

[0007] The physical entity data acquisition module is used to collect equipment data and real-time operating data of each physical entity in the Internet Data Center (IDC).

[0008] The big data processing module is used to process the device data and the real-time operation data respectively to obtain device metadata and real-time operation metadata;

[0009] A digital twin model is constructed by mapping the virtual and physical data of the device metadata to create a three-dimensional digital model of the IDC.

[0010] The AI ​​algorithm module is used to identify and predict hidden faults of each physical entity in the IDC based on the real-time running metadata, and generate fault warning messages.

[0011] The visualization module is used to display a three-dimensional digital model of the IDC, and to display the real-time operating data and fault warning messages corresponding to each physical entity of each IDC in the three-dimensional digital model of the IDC.

[0012] Optionally, the physical entity data acquisition module includes an IoT data acquisition submodule and a background data acquisition submodule; the real-time operating data includes operating environment data and operating logs;

[0013] The IoT data acquisition submodule is used to collect the device data and operating environment data of each physical entity in the IDC; the device data includes device model, device connection relationship and device parameter settings; the operating environment data includes at least one of temperature, humidity, air conditioning fan speed, cooling tower supply and return water temperature and cooling tower supply and return water flow rate.

[0014] The background data collection submodule is used to collect the operation logs of each physical entity in the IDC.

[0015] Optionally, the system further includes a data access module; the data access module includes a resource interface submodule and a general protocol submodule;

[0016] The resource interface submodule is used to provide a unified data access interface for different devices in the IDC and the IoT data acquisition submodule;

[0017] The general protocol submodule is used to store the data transmission protocols of different devices in the IDC and the IoT data acquisition submodule.

[0018] Optionally, the big data processing module includes a first cleaning submodule and a second cleaning submodule;

[0019] The first cleaning submodule is used to clean the equipment data to unify the data format, add data tags to the data, and generate the equipment metadata;

[0020] The second cleaning submodule is used to perform scenario-based data cleaning and clustering on the data of preset items in the real-time running data, and generate real-time running metadata that can be directly calculated by the AI ​​algorithm module.

[0021] Optionally, the digital twin model includes a virtual-real mapping submodule and a 3D model construction submodule;

[0022] The virtual-physical mapping submodule is used to map the physical entity as a digital entity according to the device model and device parameter settings of each physical entity in the IDC.

[0023] The 3D model construction submodule is used to connect the digital entities according to the device connection relationship between the physical entities of the IDC to generate a 3D digital model of the IDC.

[0024] Optionally, the AI ​​algorithm module includes a region detection submodule, a first anomaly detection submodule, a second anomaly detection submodule, an early warning submodule, and a feedback submodule;

[0025] The region detection submodule is used to divide the real-time running metadata into multiple first time windows according to a first time period, divide each first time window into multiple second time windows according to a second time period, calculate the similarity of data segments belonging to the same second time window, and determine the second time window with a similarity less than a preset threshold as an unstable interval, otherwise it is determined as a stable interval.

[0026] The first anomaly detection submodule is used to calculate the SVM distance, IQR distance and MAD distance of each data point in the unstable interval, obtain the average distance of the data point, and determine whether the data point is abnormal based on the relationship between the average distance of the data point and the first real-time threshold.

[0027] The second anomaly detection submodule compares each data point within the stable interval with the second real-time threshold to determine whether the data point is abnormal.

[0028] The early warning submodule is used to identify and predict hidden faults based on abnormal data points of each physical entity in the IDC, and generate fault early warning messages.

[0029] The feedback submodule updates the first real-time threshold and the second real-time threshold according to the feedback annotation; the feedback annotation is a record of whether the judgment results of the first anomaly detection submodule and the second anomaly detection submodule are correct.

[0030] Optionally, the feedback annotation includes the number of actual positive classes predicted as positive (TP), the number of actual positive classes predicted as negative (FN), and the number of actual negative classes predicted as positive (FP).

[0031] The feedback submodule is specifically used for: adjusting the first real-time threshold to determine the updated first real-time threshold as the value that maximizes the accuracy of the first anomaly detection submodule; adjusting the second real-time threshold to determine the updated second real-time threshold as the value that maximizes the accuracy of the second anomaly detection submodule; the accuracy is...

[0032] Optionally, the system further includes an object control module; the AI ​​algorithm module further includes a device adjustment submodule and an optimization database; the optimization database contains successful optimization schemes for each device;

[0033] The device adjustment submodule is used to match the best adjustment scheme in the optimization database based on the real-time operating data, and send the adjustment control command for the target device to the object control module according to the best adjustment scheme; the target device is any physical entity in the IDC.

[0034] The object control module is used to issue adjustment control commands to the target device and monitor whether the adjustment is completed.

[0035] This invention provides an intelligent operation and maintenance management system for Internet Data Centers (IDCs) based on digital twins, comprising a physical entity data acquisition module, a big data processing module, a digital twin model, an AI algorithm module, and a visualization module. The physical entity data acquisition module collects equipment data and real-time operational data from each physical entity of the IDC. The big data processing module processes the equipment data and real-time operational data to obtain equipment metadata and real-time operational metadata. The digital twin model constructs a three-dimensional digital model of the IDC based on the equipment metadata through virtual-physical mapping. The AI ​​algorithm module identifies and predicts hidden faults in each physical entity of the IDC based on the real-time operational metadata and generates fault warning messages. The visualization module displays the three-dimensional digital model of the IDC and shows the real-time operational data and fault warning messages corresponding to each physical entity of each IDC within the three-dimensional digital model.

[0036] Digital twin technology enables timely reflection of the operational status of data center hardware and software environments, providing comprehensive data and information for data center managers, maintenance personnel, and office staff. This provides reliable data for decision-makers and solves the problem of weak correlation between current DCIM operation and maintenance platforms and real-time operational data of various devices. AI algorithm modules identify, predict, and trigger alarms for hidden faults in various physical entities within the IDC, improving the efficiency of fault identification and repair. Attached Figure Description

[0037] The invention will now be further described with reference to the accompanying drawings.

[0038] Figure 1 This is a system block diagram of an IDC intelligent operation and maintenance management system based on digital twins, provided as an embodiment of the present invention. Detailed Implementation

[0039] 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.

[0040] This invention provides an intelligent IDC (Internet Data Center) operation and maintenance management system based on digital twins. See also... Figure 1 , Figure 1 This invention provides a system block diagram of an IDC intelligent operation and maintenance management system based on digital twins. It includes a physical entity data acquisition module, a big data processing module, a digital twin model, an AI algorithm module, and a visualization module; wherein:

[0041] The physical entity data acquisition module is used to collect equipment data and real-time operating data of each physical entity in the Internet Data Center (IDC).

[0042] The big data processing module is used to process equipment data and real-time operation data respectively to obtain equipment metadata and real-time operation metadata;

[0043] Digital twin model: Construct a three-dimensional digital model of the data center by mapping virtual and physical data based on device metadata;

[0044] The AI ​​algorithm module is used to identify and predict hidden faults in various physical entities of the IDC based on real-time running metadata, and generate fault warning messages.

[0045] The visualization module is used to display the three-dimensional digital model of the IDC, and to display the real-time operating data and fault warning messages corresponding to each physical entity of each IDC in the three-dimensional digital model of the IDC.

[0046] This invention provides an intelligent IDC (Internet Data Center) operation and maintenance management system based on digital twins. Through digital twin technology, it can reflect the operational status of the data center's hardware and software environment in a timely manner, providing comprehensive data and information for data center managers, maintenance personnel, and office staff. This provides reliable data for decision-makers and solves the problem of weak correlation between current DCIM (Digital Data Center Imaging) platforms and real-time operational data of various devices. By using an AI algorithm module to identify, predict, and alert on hidden faults in various physical entities within the IDC, it improves the efficiency of fault identification and repair.

[0047] In one implementation, an automatic data acquisition mechanism is used during data collection. After the data acquisition is completed, subsequent tasks can be started directly to ensure data integrity, reduce the error rate of identification, and improve the accuracy of fault identification.

[0048] In one implementation, the visualization module can use technologies such as web, VR / AR, etc., to display the status of various devices in a data center that are accurately simulated in a one-to-one manner. The visualization module allows for intuitive observation of server hotspots, device power consumption, pipe flow and velocity, airflow patterns, and other information.

[0049] In one embodiment, the physical entity data acquisition module includes an IoT data acquisition submodule and a background data acquisition submodule; real-time runtime data includes runtime environment data and runtime logs.

[0050] The IoT data acquisition submodule is used to collect device data and operating environment data of each physical entity in the IDC. Device data includes device model, device connection relationship and device parameter settings. Operating environment data includes at least one of temperature, humidity, air conditioning fan speed, cooling tower supply and return water temperature and cooling tower supply and return water flow rate.

[0051] The background data collection submodule is used to collect the operation logs of each physical entity in the IDC.

[0052] In one implementation, the IoT data acquisition module is a plug-and-play device that collects data in real time and communicates wirelessly, avoiding the need for rewiring and disruption to data center operations. The IoT data acquisition module can be a temperature sensor, humidity sensor, flow rate sensor, and wind speed sensor, etc.

[0053] In one implementation, the physical entity data acquisition module includes functions such as automatic data re-acquisition, automatic data verification, acquisition control, and acquisition integrity analysis.

[0054] In one embodiment, the system further includes a data access module; the data access module includes a resource interface submodule and a general protocol submodule;

[0055] The resource interface submodule is used to provide a unified data access interface for different devices in the IDC and the IoT data acquisition submodule.

[0056] The general protocol submodule is used to store the data transmission protocols of different devices in the IDC and the IoT data acquisition submodule.

[0057] In one embodiment, the big data processing module includes a first cleaning submodule and a second cleaning submodule;

[0058] The first cleaning submodule is used to clean the equipment data, unify the data format, and tag the data to generate equipment metadata.

[0059] The second cleaning submodule is used to perform scenario-based data cleaning and clustering on the data of preset items in the real-time running data, and generate real-time running metadata that can be directly processed by the AI ​​algorithm module.

[0060] In one embodiment, the digital twin model includes a virtual-real mapping submodule and a 3D model construction submodule;

[0061] The virtual-physical mapping submodule is used to map each physical entity in the IDC to a digital entity based on its device model and device parameter settings.

[0062] The 3D model construction submodule is used to connect various digital entities based on the device connection relationships between the physical entities of the IDC to generate a 3D digital model of the IDC.

[0063] In one embodiment, the AI ​​algorithm module includes a region detection submodule, a first anomaly detection submodule, a second anomaly detection submodule, an early warning submodule, and a feedback submodule;

[0064] The region detection submodule is used to divide the real-time running metadata into multiple first time windows according to the first time period, divide each first time window into multiple second time windows according to the second time period, calculate the similarity of data segments belonging to the same second time window, and determine the second time window with similarity less than a preset threshold as an unstable interval, otherwise it is determined as a stable interval.

[0065] The first anomaly detection submodule is used to calculate the SVM distance, IQR distance and MAD distance of each data point in the unstable interval, obtain the average distance of the data point, and determine whether the data point is abnormal based on the relationship between the average distance of the data point and the first real-time threshold.

[0066] The second anomaly detection submodule compares each data point within the stable interval with the second real-time threshold to determine whether the data point is abnormal.

[0067] The early warning submodule is used to identify and predict hidden faults based on abnormal data points of each physical entity in the IDC, and generate fault early warning messages.

[0068] The feedback submodule updates the first and second real-time thresholds based on the feedback annotations; the feedback annotations record whether the judgment results of the first and second anomaly detection submodules are correct.

[0069] In one implementation, the first time period can be a full calendar day (24 hours), and the second time period is shorter than the first time period, such as half an hour or an hour. Calculating the similarity of data segments belonging to the same second time window is equivalent to calculating the similarity of data segments within the same time period each day. The similarity of data segments can be calculated using the Dynamic Time Warping (DTW) algorithm or the Area Under the Curve (AUC) algorithm.

[0070] In one implementation, existing SVM, IQR, and MAD algorithms are used to calculate the SVM distance, IQR distance, and MAD distance for each data point within the unstable interval, respectively, and then the SVM distance, IQR distance, and MAD distance are averaged. Using these three algorithms allows for a more comprehensive and rapid identification of outliers.

[0071] In one embodiment, the feedback annotation includes the number of actual positive classes predicted as positive (TP), the number of actual positive classes predicted as negative (FN), and the number of actual negative classes predicted as positive (FP).

[0072] The feedback submodule is specifically used for: adjusting the first real-time threshold to determine the value that maximizes the accuracy of the first anomaly detection submodule, updating the first real-time threshold; adjusting the second real-time threshold to determine the value that maximizes the accuracy of the second anomaly detection submodule, updating the second real-time threshold; the accuracy rate is...

[0073] In one embodiment, the AI ​​algorithm module includes a feature extraction submodule, a preset prediction submodel, and an alarm submodule;

[0074] The feature extraction submodule is used to extract feature data from the metadata in real time based on the type of input data of the preset prediction sub-model.

[0075] A pre-defined prediction sub-model is used to identify and predict hidden faults in each physical entity of the IDC based on feature data.

[0076] The alarm submodule is used to generate fault warning messages based on the output of the preset prediction sub-model.

[0077] In one embodiment, the preset prediction sub-model is at least one of a power supply failure prediction model, a temperature and humidity failure prediction model, and a disk storage failure prediction model; the input data for the power supply failure prediction model is the voltage value of at least one of the following: power supply and distribution system equipment, uninterruptible power supply system equipment, terminal power distribution system equipment, or power auxiliary system equipment; the input data for the temperature and humidity failure prediction model is at least one of the following: temperature, humidity, air conditioning fan speed, cooling tower supply and return water temperature, and cooling tower supply and return water flow rate; the input data for the disk storage failure prediction model is disk SMART detection parameters.

[0078] In one embodiment, a disk storage failure prediction model is used to extract features from feature data, determine preset types of disk SMART detection parameters as input parameters, and the weight corresponding to each input parameter, use a decision tree with maximum information gain to predict disk storage failures for each input parameter, and perform a weighted summation based on the weight corresponding to each input parameter and the prediction results to obtain the prediction result of disk storage failure.

[0079] In one implementation, the type and number of input parameters can be set by technicians according to the actual situation, and are not limited here. For example, the input parameters can be at least one of SMART5, SMART187, SMART196 and SMART198 in the disk SMART detection parameters.

[0080] SMART5: Reallocated Sectors. Reallocated sectors refer to bad sectors on a disk drive. Data stored in these sectors is unsafe. When a hard drive experiences a read / write error, the sector is marked as "reallocated," and the data is copied from that sector to another sector to prevent data loss or corruption. The number of reallocated sectors is the number of sectors marked as reallocated due to read or write errors; a continuously increasing count is considered a predictive parameter for hard drive failure.

[0081] SMART187: Reported uncorrectable errors. Reported uncorrectable errors refer to the number of errors that cannot be corrected using error correction codes.

[0082] SMART196: Remap operations. The remap operation count is the number of times data has been transferred from the relocation sector to other sectors of the disk. The number of successful or unsuccessful operations is stored in this parameter, which is also a bad sector indicator.

[0083] SMART198: Uncorrectable errors occurred while reading / writing sectors. This SMART parameter stores the total number of uncorrectable errors that occurred while performing write and read operations on sectors. An increase in this SMART parameter count indicates disk surface failure or any failure in the mechanical system.

[0084] In one implementation, the weights of each input parameter can be set by technicians according to the actual situation, and are not limited here.

[0085] In one embodiment, the system further includes an object control module; the AI ​​algorithm module further includes a device adjustment submodule and an optimization database; the optimization database contains successful optimization schemes for each device;

[0086] The equipment adjustment submodule is used to match the best adjustment scheme with the optimization database based on real-time operating data, and send adjustment control instructions for the target device to the object control module according to the best adjustment scheme; the target device can be any physical entity in the IDC.

[0087] The object control module is used to issue adjustment control commands to the target device and monitor whether the adjustment is completed.

[0088] In one implementation, the AI ​​algorithm module can create a database of equipment parameters and scenario-appropriate optimization experience (the aforementioned optimization database). By recording historically successful equipment optimization cases, when similar anomalies occur, the system directly matches the historical experience in the equipment parameter optimization database with the equipment's condition, providing an optimization solution. After confirmation by maintenance personnel, the instructions are directly issued to the corresponding equipment for precise control. This function plays a very good role in regulating the low-carbon energy saving of the entire data center, intelligently adjusting air conditioning fan speed, cooling tower return water temperature, flow rate, etc., to the smallest scale, thereby saving a significant amount of electricity for the data center.

[0089] In one embodiment, the AI ​​algorithm module further includes a scene simulation submodule;

[0090] The scenario simulation submodule is used to simulate scenarios when maintenance personnel perform maintenance tasks that have a significant impact on online services, such as replacing equipment or network cutovers, to evaluate reasonable operating procedures and avoid operational errors.

[0091] In one implementation, the training process of the pre-defined prediction sub-model can be customized to simulate various operational scenarios, simulating the entire process of major operational events such as equipment replacement and server cutover, and formulating safe and reliable operational plans. This avoids the previous operational mode that relied solely on personal experience, minimizing risks.

[0092] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A digital twin-based intelligent operation and maintenance management system for data centers (IDC), characterized in that: It includes a physical entity data acquisition module, a big data processing module, a digital twin model, an AI algorithm module, and a visualization module; among which: The physical entity data acquisition module is used to collect equipment data and real-time operating data of each physical entity in the Internet Data Center (IDC). The big data processing module is used to process the device data and the real-time operation data respectively to obtain device metadata and real-time operation metadata; A digital twin model is constructed by mapping the virtual and physical data of the device metadata to create a three-dimensional digital model of the IDC. The AI ​​algorithm module is used to identify and predict hidden faults of each physical entity in the IDC based on the real-time running metadata, and generate fault warning messages. The AI ​​algorithm module includes a region detection submodule, a first anomaly detection submodule, a second anomaly detection submodule, an early warning submodule, and a feedback submodule; The region detection submodule is used to divide the real-time running metadata into multiple first time windows according to a first time period, divide each first time window into multiple second time windows according to a second time period, calculate the similarity of data segments belonging to the same second time window, and determine the second time window with a similarity less than a preset threshold as an unstable interval, otherwise it is determined as a stable interval. The first anomaly detection submodule is used to calculate the SVM distance, IQR distance and MAD distance of each data point in the unstable interval, obtain the average distance of the data point, and determine whether the data point is abnormal based on the relationship between the average distance of the data point and the first real-time threshold. The second anomaly detection submodule compares each data point within the stable interval with the second real-time threshold to determine whether the data point is abnormal. The early warning submodule is used to identify and predict hidden faults based on abnormal data points of each physical entity in the IDC, and generate fault early warning messages. The feedback submodule updates the first real-time threshold and the second real-time threshold according to the feedback annotation; the feedback annotation is a record of whether the judgment results of the first anomaly detection submodule and the second anomaly detection submodule are correct. The visualization module is used to display a three-dimensional digital model of the IDC, and to display the real-time operating data and fault warning messages corresponding to each physical entity of each IDC in the three-dimensional digital model of the IDC.

2. The IDC intelligent operation and maintenance management system based on digital twin as described in claim 1, characterized in that, The physical entity data acquisition module includes an IoT data acquisition submodule and a background data acquisition submodule; The real-time operational data includes operational environment data and operational logs; The IoT data acquisition submodule is used to collect the device data and operating environment data of each physical entity in the IDC; the device data includes device model, device connection relationship and device parameter settings; the operating environment data includes at least one of temperature, humidity, air conditioning fan speed, cooling tower supply and return water temperature and cooling tower supply and return water flow rate. The background data collection submodule is used to collect the operation logs of each physical entity in the IDC.

3. The IDC intelligent operation and maintenance management system based on digital twin as described in claim 2, characterized in that, The system also includes a data access module; the data access module includes a resource interface submodule and a general protocol submodule; The resource interface submodule is used to provide a unified data access interface for different devices in the IDC and the IoT data acquisition submodule. The general protocol submodule is used to store the data transmission protocols of different devices in the IDC and the IoT data acquisition submodule.

4. A digital twin-based intelligent operation and maintenance management system for IDC (Internet Data Center) as described in claim 3, characterized in that, The big data processing module includes a first cleaning submodule and a second cleaning submodule. The first cleaning submodule is used to clean the equipment data to unify the data format, add data tags to the data, and generate the equipment metadata; The second cleaning submodule is used to perform scenario-based data cleaning and clustering on the data of preset items in the real-time running data, and generate real-time running metadata that can be directly calculated by the AI ​​algorithm module.

5. A digital twin-based intelligent operation and maintenance management system for IDC (Internet Data Center) as described in claim 4, characterized in that, The digital twin model includes a virtual-real mapping submodule and a 3D model construction submodule; The virtual-physical mapping submodule is used to map the physical entity as a digital entity according to the device model and device parameter settings of each physical entity in the IDC. The 3D model construction submodule is used to connect the digital entities according to the device connection relationship between the physical entities of the IDC to generate a 3D digital model of the IDC.

6. A digital twin-based intelligent operation and maintenance management system for data centers (IDCs) as described in claim 1, characterized in that: The feedback annotation includes the number of actual positive classes predicted as positive (TP), the number of actual positive classes predicted as negative (FN), and the number of actual negative classes predicted as positive (FP). The feedback submodule is specifically used for: adjusting the first real-time threshold to determine the value that maximizes the accuracy of the first anomaly detection submodule as the updated first real-time threshold; and adjusting the second real-time threshold to determine the value that maximizes the accuracy of the second anomaly detection submodule as the updated value of the second real-time threshold.

7. A digital twin-based intelligent operation and maintenance management system for IDC (Internet Data Center) as described in claim 1, characterized in that, The system also includes an object control module; the AI ​​algorithm module further includes a device adjustment submodule and an optimization database; the optimization database stores successful optimization schemes for each device; The device adjustment submodule is used to match the best adjustment scheme in the optimization database based on the real-time operating data, and send the adjustment control command for the target device to the object control module according to the best adjustment scheme; the target device is any physical entity in the IDC. The object control module is used to issue adjustment control commands to the target device and monitor whether the adjustment is completed.

Images