Machine room monitoring method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention discloses a machine room monitoring method and device, electronic equipment and a storage medium, and the method comprises the steps: constructing a digital twin according to the equipment information of a target machine room, the digital twin comprising a three-dimensional model of the target machine room and a three-dimensional object of each piece of equipment; acquiring environment data of the target machine room, and acquiring equipment operation data of each equipment; displaying the digital twinborn body, displaying the environment data, and mapping the equipment operation data of each equipment to a three-dimensional object for displaying; for each device, when a preset time is reached, inputting a historical data sequence of the device in a target historical time period into the time sequence prediction model, and determining a predicted device operation data sequence of the device in a target future time period; and if the abnormal prediction data exists in the predicted equipment operation data sequence, displaying the abnormal prediction data in a three-dimensional object corresponding to the equipment. According to the embodiment of the invention, visual monitoring of each device in the machine room is realized, and early warning of faults is realized.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of data processing technology, and in particular to a computer room monitoring method, device, electronic device, and storage medium. Background Art

[0002] Existing computer room monitoring technologies rely on physical cameras and manual inspections to capture information within the computer room. This information is then monitored in a tabular format through the backend monitoring systems of each device, failing to provide a comprehensive and intuitive monitoring approach. Equipment room monitoring typically generates alarms based on equipment failures or set thresholds, failing to provide early warning of failures. Consequently, there is a lack of a method for predicting failures in real time. Summary of the Invention

[0003] The embodiments of the present application provide a computer room monitoring method, device, electronic device and storage medium, which can perform intuitive monitoring of the computer room and provide fault warning.

[0004] To solve the above problems, in a first aspect, an embodiment of the present application provides a computer room monitoring method, comprising:

[0005] Constructing a digital twin of the target computer room based on the equipment information of the target computer room, wherein the digital twin includes a three-dimensional model of the target computer room and a three-dimensional object of each device in the target computer room;

[0006] Obtaining environmental data of the target computer room and equipment operation data of each equipment in the target computer room;

[0007] Displaying the digital twin, displaying the environmental data in the digital twin, and mapping the device operation data of each device to a three-dimensional object of the corresponding device in the digital twin for display;

[0008] For each device, when a preset time is reached, the historical data sequence of the device in the target historical time period is input into the time series prediction model, and the predicted device operation data sequence of the device in the target future time period is determined by the time series prediction model;

[0009] If there is abnormal prediction data in the predicted device operation data sequence, the abnormal prediction data is displayed in the three-dimensional object corresponding to the device.

[0010] In a second aspect, an embodiment of the present application provides a computer room monitoring device, comprising:

[0011] A digital twin construction module, configured to construct a digital twin of the target computer room based on the equipment information of the target computer room, wherein the digital twin includes a three-dimensional model of the target computer room and a three-dimensional object of each device in the target computer room;

[0012] A data acquisition module is used to acquire environmental data of the target computer room and equipment operation data of each device in the target computer room;

[0013] a first display module, configured to display the digital twin, display the environmental data in the digital twin, and map the device operation data of each device to a three-dimensional object of the corresponding device in the digital twin for display;

[0014] A data prediction module is configured to input, for each device, a historical data sequence of the device within a target historical time period into a time series prediction model when a preset time is reached, and determine a predicted device operation data sequence of the device within a target future time period using the time series prediction model;

[0015] The second display module is configured to display abnormal prediction data in a three-dimensional object corresponding to the device if abnormal prediction data exists in the predicted device operation data sequence.

[0016] In a third aspect, an embodiment of the present application further provides an electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the computer room monitoring method described in the embodiment of the present application when executing the computer program.

[0017] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the computer room monitoring method disclosed in the embodiment of the present application are performed.

[0018] The computer room monitoring method, device, electronic device, and storage medium provided in the embodiments of the present application construct a digital twin of the target computer room based on the equipment information of the target computer room, obtain the environmental data of the target computer room, and obtain the equipment operation data of each device in the target computer room, display the digital twin, and display the environmental data in the digital twin. The equipment operation data of each device is mapped to the three-dimensional object of the corresponding device in the digital twin for display. For each device, when a preset time is reached, the historical data sequence of the device in the target historical time period is input into a time series prediction model. The predicted equipment operation data sequence of the device in the target future time period is determined by the time series prediction model. If there is abnormal data in the predicted equipment operation data sequence, the abnormal data is displayed in the three-dimensional object corresponding to the device. By displaying the environmental data in the digital twin and the corresponding equipment operation data in the three-dimensional object of each device, real-time and intuitive monitoring of each device in the computer room is achieved. In addition, the time series prediction model can predict the equipment operation data in the target future time period. When there is abnormal data in the predicted equipment operation data, it is displayed in the three-dimensional object corresponding to the device in the digital twin, achieving early warning of faults, which can effectively reduce and prevent the occurrence of faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0020] Figure 1 This is a flow chart of a computer room monitoring method provided by an embodiment of the present application;

[0021] Figure 2 is an example diagram of the digital twin of the target computer room in the embodiment of the present application;

[0022] Figure 3 This is an example diagram for distinguishing and displaying device operation data and abnormality prediction data in the embodiment of the present application;

[0023] Figure 4 This is a schematic diagram of the Hadoop cluster construction mode in the embodiment of the present application;

[0024] Figure 5 It is a structural diagram of the computer room monitoring and early warning platform in an embodiment of the present application;

[0025] Figure 6 This is a flowchart of the operation of the intelligent analysis module in the embodiment of the present application;

[0026] Figure 7 This is a schematic diagram of the structure of a computer room monitoring device provided in an embodiment of the present application;

[0027] Figure 8 This is a structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0029] Before specifically introducing the technical solutions provided by this application, the terms involved in this application are explained as follows:

[0030] Digital Twin: A concept that transcends reality. It fully utilizes data such as physical models, sensor updates, and operational history, integrates multi-disciplinary, multi-physical, multi-scale, and multi-probability simulation processes, and completes mapping in virtual space, thereby reflecting the entire life cycle of the corresponding physical equipment. It can be regarded as a digital mapping system of one or more important, interdependent equipment systems.

[0031] Prophet (Facebook Prophet): A time series model based on time and variable values, combining time series decomposition and machine learning fitting. The modeling process considers the influence of factors such as trend lines, seasonality, cyclicality, and exogenous variables. It delivers excellent forecasting results and is highly robust to missing values, trend shifts, and large outliers, offering significant advantages over traditional time series models.

[0032] TensorFlow: An open-source machine learning library widely used for implementing various machine learning applications. It can build, train, and deploy deep learning models on various platforms. TensorFlow provides a powerful computational framework for numerical computation, especially large-scale distributed computing on large-scale graphics processing units.

[0033] Hadoop (Apache Hadoop): A distributed system infrastructure. Users can develop distributed programs without understanding the underlying details of the distributed system, fully utilizing the power of clusters for high-speed computing and storage.

[0034] HDFS (Hadoop Distributed File System): A distributed file system with high fault tolerance and designed to be deployed on low-cost hardware. It provides high throughput for accessing application data and is suitable for applications with very large data sets. Data in the file system can be accessed in a streaming manner.

[0035] YARN (Yet Another Resource Negotiator): A resource management system in Hadoop. It is a general-purpose resource management system that provides unified resource management and task scheduling for upper-layer applications. Its main advantage is that it can run various types of jobs, including batch, interactive, and containerized jobs, and it can provide flexible resource allocation during job execution.

[0036] ZooKeeper: A highly available distributed data management and coordination framework that ensures data consistency in a distributed environment. It is an important component of Hadoop and HBase.

[0037] HBase (Hadoop Database): A highly reliable, high-performance, column-oriented, scalable distributed database. Using HBase technology, large-scale structured storage clusters can be built on inexpensive PCs. Based on HDFS as its file storage system, HBase differs from general relational databases in that it is suitable for storing unstructured data.

[0038] Kafka (Apache Kafka): A distributed publish-subscribe messaging system and a powerful queue that can handle large amounts of data and is suitable for both offline and online message consumption. Kafka messages are saved on disk and replicated within the cluster to prevent data loss.

[0039] Flink (Apache Flink): An open source stream processing framework that executes arbitrary streaming programs in a data-parallel, pipelined manner. The pipeline runtime system can execute both batch and stream processing programs.

[0040] The following is an introduction to the specific scheme of the computer room monitoring method provided in the embodiment of the present application.

[0041] Figure 1 This is a flow chart of a computer room monitoring method provided by an embodiment of the present application. Figure 1 As shown, the method includes: steps 110 to 150.

[0042] Step 110: construct a digital twin of the target computer room based on the equipment information of the target computer room, wherein the digital twin includes a three-dimensional model of the target computer room and a three-dimensional object of each equipment in the target computer room.

[0043] The device information may include model, location, configuration, etc. The devices in the target computer room may include at least one of cabinets, various sensors, patch panels, network devices, data devices, transmission devices, power devices, etc.

[0044] Use a three-dimensional modeling tool to build a three-dimensional model of the target computer room based on the equipment information of the target computer room, and construct a three-dimensional object corresponding to each device in the target computer room. Based on the functional entities of each device in the equipment information (such as buttons, etc.), construct a sub-module three-dimensional object corresponding to the functional entity on the corresponding three-dimensional object to obtain a digital twin of the target computer room.

[0045] Figure 2 This is an example diagram of the digital twin of the target computer room in the embodiment of the present application, such as Figure 2 As shown in the figure, the constructed digital twin reflects information such as the location of each device in the target computer room.

[0046] Step 120: Acquire environmental data of the target computer room and obtain equipment operation data of each equipment in the target computer room.

[0047] Environmental data collection equipment is installed in the target computer room. This equipment collects environmental data from the target computer room in real time and transmits the collected data to the device executing the computer room monitoring method. The device executing the computer room monitoring method receives and stores the environmental data. The environmental data collection equipment may include sensors, cameras, and other devices. The environmental data may include information such as temperature and humidity.

[0048] Each device in the target computer room uses a device data interface to obtain device operating data and transmits this data to the device executing the computer room monitoring method. The device executing the computer room monitoring method then stores the received data. The device executing the computer room monitoring method can subsequently monitor and predict the devices in the target computer room in real time based on the stored environmental data and device operating data. The device operating data refers to data collected during device operation, and may include, for example, optical power, current, and other data.

[0049] In an embodiment of the present application, environmental data and device operation data can be obtained in the form of data streams. For example, environmental data sent by each sensor can be received through a Kafka cluster, and device operation data sent by each device through a device data interface can be received.

[0050] Step 130: display the digital twin, display the environmental data in the digital twin, and map the device operation data of each device to the three-dimensional object of the corresponding device in the digital twin for display.

[0051] The Vue framework and Three.js engine can be used to render digital twins in WEB applications, map environmental data and equipment operation data to the digital twins, and display the equipment operation data of the corresponding equipment in the three-dimensional object to achieve real-time monitoring of each device in the target computer room.

[0052] Step 140 , for each device, when a preset time is reached, the historical data sequence of the device in the target historical time period is input into a time series prediction model, and the predicted device operation data sequence of the device in the target future time period is determined by the time series prediction model.

[0053] Among them, the preset time is the pre-set trigger time for data prediction, which can be a fixed time on a certain day, for example, a fixed time every day, or a fixed time every other day, etc. The target historical time period is the historical time period before the current time (i.e., the preset time), that is, the current time is the end time of the target historical time period. The target future time period is the time period after the current time (i.e., the preset time), that is, the current time is the starting time of the target future time period. The target historical time period can be, for example, within a historical year, and the target future time period can be, for example, within the next week, etc., which can be set according to needs.

[0054] The historical data sequence for a device within a target historical time period is a chronological sequence of device operation data collected at each collection time point within the target historical time period for a predicted indicator (e.g., optical power, current, etc.). This historical data sequence is input into a time series prediction model corresponding to the device and the predicted indicator. The model then extracts data features from the historical data sequence. These features may include, for example, data growth type features, seasonal cycle features, and holiday features. Based on the extracted data features, the device operation data for the device at each time point (corresponding to the collection time point) within the target future time period is predicted to obtain a predicted device operation data sequence.

[0055] It should be noted that the historical data sequence input into the time series prediction model includes the equipment operation data within the target historical time period, and may also include the environmental data at the time corresponding to the equipment operation data, which can improve the accuracy of data prediction.

[0056] In an exemplary embodiment, the time series prediction model is a Prophet model. The Prophet model can comprehensively consider the impact of factors such as data growth type, seasonal cycles, holidays, etc., and is extremely robust to missing values, trend shifts, and a large number of outliers, thereby improving prediction accuracy.

[0057] Before inputting the historical data sequence into the time series prediction model, the historical data sequence is subjected to a differential ADF (Augmented Dickey-Fuller) test to determine whether the historical data sequence is a stationary time series. If the historical data sequence is a non-stationary time series, the historical data sequence is subjected to differential processing, and the stationary data sequence obtained after processing is input into the time series prediction model. The time series prediction model is used to determine the predicted equipment operation data sequence in the target future time period.

[0058] Among them, the n-order difference formula is as follows:

[0059] Δ n y x =Δ(Δ n-1 y x )=Δ n-1 y x+1 -Δ n-1 y x

[0060] Among them, Δ n y x Represents the n-order difference of the x-th running data in the historical data sequence, Δ n-1 y x Represents the n-1 order difference of the x-th running data in the historical data sequence, Δ n-1 y x+1 Represents the n-1th order difference of the x+1th running data in the historical data sequence.

[0061] The ADF test formula is as follows:

[0062] ADF(t)=(Y(t)-Y(t-1))-λΔY(t-1)+αt+βY(t-1)+ε(t)

[0063] Among them, ADF(t) represents the ADF test value, Y(t) represents the historical data sequence, Y(t-1) represents the lagged first-order value of the historical data sequence, ΔY(t-1) represents the first-order difference of the historical data sequence, ε(t) represents the error, and λ, α, and β are constants.

[0064] Step 150: If abnormal prediction data exists in the predicted device operation data sequence, the abnormal prediction data is displayed in a three-dimensional object corresponding to the device.

[0065] Each piece of predicted device operation data in the predicted device operation data sequence is compared with the preset normal data range corresponding to the data. If the predicted device operation data is within the preset normal data range, the predicted device operation data is determined to be normal prediction data. If the predicted device operation data is outside the preset normal data range, the predicted device operation data is determined to be abnormal prediction data. If abnormal prediction data exists in the predicted device operation data sequence, the abnormal prediction data is displayed in the three-dimensional object corresponding to the device in the digital twin, so that early warning can be issued.

[0066] In one embodiment of the present application, displaying the abnormality prediction data in the three-dimensional object corresponding to the device includes: displaying the abnormality prediction data and the device operation data separately in the three-dimensional object corresponding to the device.

[0067] In the three-dimensional object of the digital twin, the equipment operation data and abnormal prediction data of the corresponding equipment need to be displayed. In order to facilitate the distinction, the two types of data can be displayed separately. For example, the abnormal prediction data can be highlighted, or the two types of data can be displayed in different font colors, such as Figure 3 As shown, the early warning department displays abnormal prediction data.

[0068] The computer room monitoring method provided in the embodiment of the present application constructs a digital twin of the target computer room based on the equipment information of the target computer room, obtains the environmental data of the target computer room, obtains the equipment operation data of each device in the target computer room, displays the digital twin, and displays the environmental data in the digital twin. The equipment operation data of each device is mapped to the three-dimensional object of the corresponding device in the digital twin for display. For each device, when a preset time is reached, the historical data sequence of the device in the target historical time period is input into a time series prediction model. The predicted equipment operation data sequence of the device in the target future time period is determined by the time series prediction model. If there is abnormal data in the predicted equipment operation data sequence, the abnormal data is displayed in the three-dimensional object corresponding to the device. By displaying the environmental data in the digital twin and the corresponding equipment operation data in the three-dimensional object of each device, real-time and intuitive monitoring of each device in the computer room is achieved. In addition, the time series prediction model can predict the equipment operation data in the target future time period. When there is abnormal data in the predicted equipment operation data, it is displayed in the three-dimensional object corresponding to the device in the digital twin, thereby achieving early warning of faults and effectively reducing and preventing the occurrence of faults.

[0069] On the basis of the above technical solution, the computer room monitoring method is executed by a Hadoop cluster;

[0070] The Hadoop cluster includes: Zookeeper cluster, HDFS cluster, Yarn cluster, Kafka cluster, Flink cluster and Hbase cluster;

[0071] The Zookeeper cluster is used to ensure the consistency of data in the Hadoop cluster;

[0072] The HDFS cluster is used as the underlying file storage system;

[0073] The Yarn cluster is used to schedule resources for the execution of data processing tasks, wherein the data processing tasks include the task of determining the data sequence of the prediction device operation;

[0074] The Kafka cluster is used to receive and transmit data in the Hadoop cluster;

[0075] The Flink cluster is used to perform batch processing and stream processing of data in the Hadoop cluster;

[0076] The Hbase cluster is used to store the environmental data, the device operation data and the time series prediction model, and runs based on the HDFS cluster, using the HDFS cluster as the underlying file storage system.

[0077] Among them, the Zookeeper cluster, HDFS cluster, and Yarn cluster constitute the basic cluster of the Hadoop cluster.

[0078] The entire Hadoop cluster adopts a master-two-slave model and is deployed based on three servers, one of which is the master server (Master) and the other two are slave servers (Slave1 and Slave2). The construction model can be referred to Figure 4 The operating system used by each server may be CentOS (Community Enterprise Operating System).

[0079] A Zookeeper cluster can be deployed on three servers to ensure data consistency in a distributed environment. QuorumPeerMain is the startup class for the Zookeeper cluster and a key component of the cluster. It is the main entry point for the Zookeeper service and is responsible for starting and managing the Zookeeper service. An HDFS cluster can be deployed on three servers in high-availability (HA) mode, effectively avoiding the Namenode's single point of failure and improving its availability and reliability. An HDFS cluster has two types of nodes and operates in a manager-worker mode: a NameNode (manager) and multiple DataNodes (workers). The backup manager (Secondary NameNode) is responsible for regularly backing up NameNode data snapshots. A Yarn cluster can be deployed on three servers, providing unified resource management and scheduling for data processing tasks. The global resource manager (Resource Manager) is responsible for cluster-wide resource management and allocation. The node manager (Node Manager) is the resource and task manager for each node. The job history server (Job History Server) is an independent service used to store and display historical information about Yarn applications.

[0080] The Kafka cluster, deployed on three servers, receives and transmits data. For example, it receives environmental data from sensors within the target computer room and operational data from devices within the target computer room, and transmits this data to the Flink cluster. It also receives processed data returned by the Flink cluster and transmits this data to the early warning platform for display. The early warning platform is the platform that displays the digital twin.

[0081] The Flink cluster unit is used to perform batch processing and stream processing of data. It is deployed in three servers and deployed in Flink On Yarn mode. It can use the resource scheduling mechanism of the Yarn cluster to complete resource allocation, avoid the single point failure problem of the job manager (JobManager), and achieve high availability. The job manager controls the main process of the execution of an application (job). The task manager (Task Manager) is the working process in Flink. The Flink On Yarn mode includes session mode, separation (Per-Job) mode, and application mode. Among them, the session mode requires starting the Flink cluster first, applying for resources from the Yarn cluster, and then submitting tasks to the applied resources. This Flink cluster will reside in the Yarn cluster unless it is manually stopped. In separate mode, each job corresponds to a specific Flink cluster. Each submitted job independently requests resources from the Yarn cluster based on its own needs until the job is completed. The failure of one job does not affect the normal submission and execution of the next job. Each job has its own dispatcher and resource manager, which accepts resource requests on demand, making it suitable for large-scale, long-running jobs. Application mode starts a Flink cluster on the Yarn cluster. The main function of the application JAR package (the main function of the user class) will be executed on the JobManager. Once the application completes, the Flink cluster will be immediately shut down. The Flink cluster can also be stopped manually.

[0082] The HBase cluster unit is used to store data and is deployed across three servers. It runs on HDFS, using HDFS as the underlying file storage system, where data is stored. The master node (HMaster) is the primary node for the entire HBase cluster. Slave nodes (Region Servers) are responsible for data read and write operations. Backup master nodes (Backup Masters) are set up to ensure high availability of the HBase cluster. If the active master node fails, one of the backup master nodes is activated and becomes the master node.

[0083] Traditional computer room monitoring systems use a point-based data collection mechanism, with protocol interpretation performed by the monitoring host. Due to the large number of devices in the computer room and the large amount of data being processed, this results in high network traffic and bandwidth pressure, impacting system performance and stability, making it impossible to achieve true real-time monitoring or run specific tasks. However, the present embodiment enables real-time streaming and analysis of data through a Hadoop cluster comprising a Zookeeper cluster, HDFS cluster, Yarn cluster, Kafka cluster, Flink cluster, and HBase cluster.

[0084] Based on the above technical solution, the historical data sequence of the device in the target historical time period is input into the time series prediction model, and the predicted device operation data sequence of the device in the target future time period is determined by the time series prediction model, including: loading the time series prediction model from the Hbase cluster through the Flink cluster; inputting the historical data sequence into the time series prediction model through the Flink cluster, and determining the predicted device operation data sequence of the device in the target future time period by the time series prediction model.

[0085] When the preset time arrives, the Flink cluster triggers the execution of the data prediction task, loads the trained time series prediction model and the device operation data within the target historical time period from the Hbase cluster, obtains the historical data sequence, inputs the historical data sequence into the time series prediction model, and the time series prediction model extracts data features from the historical data sequence and performs data prediction based on the extracted data features to determine the predicted device operation data sequence within the target future time period.

[0086] When performing data prediction through a Flink cluster, you can generate different data prediction tasks for different prediction indicators and execute each data prediction task separately.

[0087] The Flink cluster can efficiently process streaming data, improve the accuracy of predicted data, and improve data processing efficiency.

[0088] On the basis of the above technical solution, after obtaining the equipment operation data of each device in the target computer room, it also includes: if there is abnormal data in the equipment operation data of each device, generating early warning information including the abnormal data, and sending the early warning information to the target personnel.

[0089] When acquiring the operational data of each device, the system determines whether any abnormalities are present. If so, it generates a warning message containing the abnormalities and sends it to the designated personnel. This warning message allows them to proactively perform maintenance on the device based on the warning information to prevent equipment failure. Warning messages can be sent via text messages, push notifications, or emails.

[0090] When determining whether there is abnormal data in the device operation data, the device operation data sent by the device through the device data interface will include an alarm for abnormal data. At this time, if the device operation data includes an alarm for abnormal data, it can be determined that the device operation data is abnormal data; or, when the device operation data does not include an alarm for abnormal data, the device operation data can be compared with the normal data range corresponding to the data indicator (the data indicator corresponding to the device operation data). If the device operation data is within the normal data range, it is determined that the device operation data is normal data; if the device operation data is outside the normal data range, it is determined that the device operation data is abnormal data.

[0091] On the basis of the above technical solution, after obtaining the equipment operation data of each device in the target computer room, it also includes: if there is abnormal data in the equipment operation data of each device, the time series prediction model is retrained according to the abnormal data, and the retrained time series prediction model is saved.

[0092] If, after judging the equipment operation data, it is determined that there is abnormal data in the equipment operation data, retraining of the time series prediction model is triggered, and the historical data corresponding to the abnormal data (historical data with the same data indicator as the abnormal data) and the abnormal data are used to generate a training data set, and the input data of the training samples in the training data set include the abnormal data. When generating the training data set, data with a time period length corresponding to the target historical time period is intercepted as input data, and data with a time period length corresponding to the target future time period length after the input data is intercepted from the historical data as output labeled data. The input data and the output labeled data are used as a training sample. By intercepting input data and output data from different starting times in this way, multiple training samples can be obtained, thereby obtaining a training data set.

[0093] Retrain the time series prediction model using the training dataset. This involves inputting the input data in the training dataset into the time series prediction model to be trained, obtaining predicted output data, and adjusting the parameters of the time series prediction model based on the predicted output data and the output annotation data. Training is iteratively performed in this manner until the training end conditions are met, resulting in a retrained time series prediction model. After retraining is complete, the retrained time series prediction model is saved. The next time data is predicted, the latest time series prediction model can be used for prediction, thereby improving the accuracy of data prediction.

[0094] When retraining the time series prediction model, you can use the TensorFlow framework to run it in the Flink cluster. After retraining, the newly trained time series prediction model will be saved to the HBase database. When the prediction is triggered next time, the Flink cluster can directly call the time series prediction model saved in the HBase cluster for data prediction.

[0095] When training the Prophet model, a differential ADF test is first performed on the input data of the Prophet model to determine whether the data is a stationary time series. If it is not stationary, the input data is differentially processed, and training is started only after the processed stationary data. The n-order difference and ADF test formulas are as described above and will not be repeated here.

[0096] In the Prophet model, you can customize influencing factors such as data growth type, seasonal cycle, and holidays according to the data type of different environments and device parameters. Since many hyperparameters need to be adjusted, a genetic algorithm can be introduced to optimize the hyperparameters of the Prophet model. Determining the optimized hyperparameter values ​​can improve the model prediction effect. Finally, the root mean square error (RMSE), mean absolute percentage error (MAPE), and mean absolute error (MAE) can be used to test the accuracy of the trained Prophet model. The smaller the RMSE, MAPE, and MAE, the higher the model accuracy.

[0097] The Prophet principle formula is as follows:

[0098] y(t)=g(t)+s(t)+h(t)+∈t

[0099] Among them, g(t), s(t), and h(t) represent the data growth type impact factor item, seasonal cycle impact factor item, and holiday impact factor item, respectively, and ∈t represents the adjustment item.

[0100] The formulas for RMSE, MAPE, and MAE are as follows:

[0101]

[0102] Among them, RMSE represents the root mean square error, y i Represents the i-th data in the output data sequence, Represents the average value of each data in the output data sequence, n represents the number of data in the output data sequence, Represents the i-th data in the real data sequence (output labeled data sequence) corresponding to the output data sequence.

[0103] Based on the above technical solution, when the preset time is reached, before the historical data sequence of the device within the target historical time period is input into the time series prediction model, it also includes: receiving the preset time, the target historical time period and the target future time period set by the user.

[0104] The preset time, target historical time period, and target future time period for the time series forecasting model can be customized through the early warning platform. For example, users can set the time period for historical data and forecasts, as well as the forecast interval, to obtain the user-defined target historical time period, target future time period, and the preset time for triggering the forecast. If the user does not set these, the default prediction is to use the historical data for one year to predict the data for the next week, with forecasts performed every other day.

[0105] The computer room monitoring method provided in the embodiment of the present application can be executed based on a computer room monitoring and early warning platform. Figure 5 This is a schematic diagram of the structure of the computer room monitoring and early warning platform in the embodiment of the present application. Figure 5 As shown, the computer room monitoring and early warning platform can include: digital twin module, data acquisition module, big data module, intelligent analysis module, abnormality reporting module and information management module.

[0106] The digital twin module is used to construct and render a three-dimensional model of the physical computer room, interact with the big data module, and map the environment, equipment and other data to the digital twin one by one to achieve unified, global and intuitive human-computer interaction; the data acquisition module is used to collect the computer room's environmental data, equipment operation data and other data in real time, and send the collected data to the big data module; the big data module is used to build a Hadoop big data cluster and a distributed, real-time processing framework to provide an operating environment for data processing, analysis and prediction, and the final data will be sent to the digital twin module; the intelligent analysis module is used for data processing and analysis, and uses TensorFlow and Prophet models to predict data, and runs based on the big data module; the abnormal reporting module is used to notify the operation and maintenance personnel of the abnormal alarm and record it in the alarm log; the information management module is used to store the computer room environment, equipment and user information.

[0107] The digital twin module is connected to the big data module and the information management module respectively; the data acquisition module and the big data module are connected to each other; the big data module is connected to the digital twin module, data acquisition module, intelligent analysis module, abnormality reporting module, and information management module respectively; the intelligent analysis module and the big data module are connected to each other; the abnormality reporting module and the information management module are connected to each other; and the information management module and the big data module are connected to each other.

[0108] The digital twin module can include a data twin unit and a data mapping unit. The data twin unit uses the Blender modeling tool to build and render the digital twin of the target computer room (the actual physical computer room), and interacts with the data mapping unit to perform three-dimensional modeling of the target computer room, and build corresponding three-dimensional objects for the cabinets, various sensors, distribution frames, network equipment, data equipment, transmission equipment, power equipment, etc. in the target computer room, and build corresponding sub-module three-dimensional objects according to the functional entities of the corresponding equipment to achieve operations such as controlling the equipment. The data mapping unit is used to receive data from the big data module and send the data to the digital twin unit, use the Vue framework and the Three.js engine to render the digital twin in the WEB application, map the data to the digital twin, and display the data in the three-dimensional object of the corresponding device.

[0109] The data acquisition module may include an environmental information acquisition unit and a device information acquisition unit. The environmental information acquisition unit is used to collect environmental data from the target computer room using sensors, cameras, and other methods, and transmits the data to the big data module. The device information acquisition unit is used to collect device operating data from the target computer room using device data interfaces and other methods, and transmits the data to the big data module.

[0110] The Big Data module primarily consists of a Hadoop cluster, which includes a basic cluster unit, a Kafka cluster unit, a Flink cluster unit, and an HBase cluster unit. The entire Hadoop cluster uses a master-slave model and is deployed on three CentOS servers: Master, Slave1, and Slave2.

[0111] The basic cluster unit is used to build Zookeeper, HDFS, and Yarn clusters. The Zookeeper cluster is deployed on three servers to ensure data consistency in a distributed environment. The HDFS cluster is deployed on three servers in a high-availability (HA) mode, effectively avoiding the Namenode's single point of failure and improving its availability and reliability. The Yarn cluster is deployed on three servers to provide unified resource management and scheduling for task execution.

[0112] The Kafka cluster unit is used for data reception and transmission and is deployed across three servers. The Flink cluster unit is used for batch and stream processing and is deployed across three servers. It uses the Flink on Yarn model, leveraging the Yarn cluster's resource scheduling mechanism to allocate resources, avoiding the JobManager single point of failure and achieving high availability. The HBase cluster unit is used for data storage and is deployed across three servers. It runs on HDFS, using HDFS as its underlying file storage system, and stores data in HDFS.

[0113] The intelligent analysis module can include a data processing unit and a data prediction unit, and is based on the big data module. Its operation process can be referred to Figure 6 The data processing unit is used to perform simple processing on the data collected in real time, such as data from device data interfaces, sensors, cameras, etc., and distribute it to the Flink cluster for processing. It analyzes whether the data is abnormal, and directly transmits the analyzed and processed data back to the Kafka cluster and maps it to the digital twin to facilitate real-time monitoring by users. If there is an abnormality in the data, the data will be sent to the data prediction unit and the abnormality reporting module. The data prediction unit is used to train historical data such as the environment and equipment and predict future data. By default, data within one year of history is used to predict data within the next week, and predictions are made every other day. Customization is supported, and users can set the time periods for historical data and predictions, as well as the prediction time intervals, on the platform. When there are anomalies in the real-time collected data, the Prophet model will be triggered to be retrained. After running in the Flink cluster using the TensorFlow framework, the retrained model will be saved to the Hbase database, so that the next time the prediction is triggered, there is no need to retrain the model. The Flink cluster can directly call the Prophet model for prediction. The predicted future data (predicted device operation data series) will be directly transmitted back to the Kafka cluster and mapped to the digital twin. If there are abnormal predicted data, they will be highlighted.

[0114] The exception reporting module can include an exception notification unit and a log recording unit. The exception notification unit is used to send fault information and warning information to the target personnel via mobile phone, email, etc. The log recording unit is used to save the fault information and warning information to the HBase database.

[0115] The information management module can include an environment information unit, a device information unit, and a user information unit. The environment information unit is used to store information such as the temperature, humidity, and rack of the target computer room environment; the device information unit is used to store information such as the model, location, and configuration of the equipment; and the user information unit is used to store information about the target personnel of the computer room and equipment, such as name, mobile phone number, and email address.

[0116] The embodiment of the present application adopts digital twin technology, Blender, Vue, and Three.js to perform three-dimensional modeling and rendering of the computer room environment and equipment, build a digital twin of the target computer room, and provide a global and intuitive computer room monitoring and interaction method.

[0117] The embodiment of the present application adopts the Hadoop big data framework and the Flink real-time stream processing framework to realize real-time processing and analysis of collected data, and combines it with the Prophet time series prediction model to perform real-time prediction of the status of the computer room and equipment, which can effectively reduce and prevent the occurrence of failures, improve monitoring efficiency, accuracy and reliability, and improve the work efficiency and productivity of operation and maintenance personnel.

[0118] Figure 7 This is a schematic diagram of the structure of a computer room monitoring device provided in an embodiment of the present application. Figure 7 As shown, the device includes:

[0119] A digital twin construction module 710 is configured to construct a digital twin of the target computer room based on the equipment information of the target computer room, wherein the digital twin includes a three-dimensional model of the target computer room and a three-dimensional object of each device in the target computer room;

[0120] The data acquisition module 720 is used to acquire the environmental data of the target computer room and the equipment operation data of each equipment in the target computer room;

[0121] A first display module 730 is configured to display the digital twin, display the environmental data in the digital twin, and map the device operation data of each device to a three-dimensional object of the corresponding device in the digital twin for display;

[0122] A data prediction module 740 is configured to input, for each device, a historical data sequence of the device within a target historical time period into a time series prediction model when a preset time is reached, and determine a predicted device operation data sequence of the device within a target future time period using the time series prediction model;

[0123] The second display module 750 is configured to display abnormal prediction data in a three-dimensional object corresponding to the device if abnormal prediction data exists in the predicted device operation data sequence.

[0124] Optionally, the second display module includes:

[0125] A distinguishing and displaying unit is used to distinguish and display the abnormality prediction data and the equipment operation data in the three-dimensional object corresponding to the equipment.

[0126] Optionally, the time series prediction model is a Prophet model.

[0127] Optionally, the computer room monitoring method is executed by a Hadoop cluster;

[0128] The Hadoop cluster includes: Zookeeper cluster, HDFS cluster, Yarn cluster, Kafka cluster, Flink cluster and Hbase cluster;

[0129] The Zookeeper cluster is used to ensure the consistency of data in the Hadoop cluster;

[0130] The HDFS cluster is used as the underlying file storage system;

[0131] The Yarn cluster is used to schedule resources for the execution of data processing tasks, wherein the data processing tasks include the task of determining the data sequence of the prediction device operation;

[0132] The Kafka cluster is used to receive and transmit data in the Hadoop cluster;

[0133] The Flink cluster is used to perform batch processing and stream processing of data in the Hadoop cluster;

[0134] The Hbase cluster is used to store the environmental data, the device operation data and the time series prediction model, and runs based on the HDFS cluster, using the HDFS cluster as the underlying file storage system.

[0135] Optionally, the data prediction module includes:

[0136] A model loading unit, used to load the time series prediction model from the Hbase cluster through the Flink cluster;

[0137] The data prediction unit is configured to input the historical data sequence into a time series prediction model through a Flink cluster, and determine a predicted device operation data sequence of the device within a target future time period through the time series prediction model.

[0138] Optionally, the device further includes:

[0139] The early warning sending module is used to generate early warning information including abnormal data if there is abnormal data in the equipment operation data of each device, and send the early warning information to the target personnel.

[0140] Optionally, the device further includes:

[0141] The model retraining module is used to retrain the time series prediction model according to the abnormal data if there is abnormal data in the device operation data of each device, and save the retrained time series prediction model.

[0142] Optionally, the device further includes:

[0143] The prediction setting module is used to receive the preset time, the target historical time period and the target future time period set by the user.

[0144] The computer room monitoring device provided in the embodiment of the present application is used to implement the various steps of the computer room monitoring method described in the embodiment of the present application. The specific implementation methods of each module of the device can be found in the corresponding steps and will not be repeated here.

[0145] The computer room monitoring device provided in the embodiment of the present application constructs a digital twin of the target computer room based on the equipment information of the target computer room, obtains the environmental data of the target computer room, obtains the equipment operation data of each device in the target computer room, displays the digital twin, and displays the environmental data in the digital twin. The equipment operation data of each device is mapped to the three-dimensional object of the corresponding device in the digital twin for display. For each device, when a preset time is reached, the historical data sequence of the device in the target historical time period is input into the time series prediction model, and the predicted equipment operation data sequence of the device in the target future time period is determined by the time series prediction model. If there is abnormal data in the predicted equipment operation data sequence, the abnormal data is displayed in the three-dimensional object corresponding to the device. By displaying the environmental data in the digital twin and the corresponding equipment operation data in the three-dimensional object of each device, real-time and intuitive monitoring of each device in the computer room is achieved. In addition, the time series prediction model can predict the equipment operation data in the target future time period. When there is abnormal data in the predicted equipment operation data, it is displayed in the three-dimensional object corresponding to the device in the digital twin, thereby achieving early warning of faults and effectively reducing and preventing the occurrence of faults.

[0146] Figure 8 is a structural diagram of an electronic device provided in an embodiment of the present application, such as Figure 8 As shown, the electronic device 800 may include one or more processors 810 and one or more memories 820 connected to the processors 810. The electronic device 800 may also include an input interface 830 and an output interface 840 for communicating with another device or system. The program code executed by the processor 810 may be stored in the memory 820.

[0147] The processor 810 in the electronic device 800 calls the program code stored in the memory 820 to execute the computer room monitoring method in the above embodiment.

[0148] An embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the computer room monitoring method as described in the embodiment of the present application are implemented.

[0149] The embodiment of the present application also provides a computer program product, which, when executed by a processor, implements the steps of the computer room monitoring method as described in the embodiment of the present application.

[0150] Each embodiment in this specification is described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between the various embodiments can be referred to in conjunction with each other. For the device embodiments, since they are generally similar to the method embodiments, their description is relatively simple, and for relevant parts, reference can be made to the description of the method embodiments.

[0151] The above is a detailed introduction to a computer room monitoring method, device, electronic device and storage medium provided in the embodiments of the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for general technical personnel in this field, based on the ideas of the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

[0152] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiments.

Claims

1. A computer room monitoring method, characterized in that: include: According to the equipment information of the target computer room, a digital twin of the target computer room is constructed, wherein the digital twin includes a three-dimensional model of the target computer room and a three-dimensional object of each equipment in the target computer room; Acquire environmental data of the target computer room, and acquire equipment operation data of each equipment in the target computer room; Displaying the digital twin, displaying the environmental data in the digital twin, and mapping the device operation data of each device to a three-dimensional object of the corresponding device in the digital twin for display; For each device, when a preset time is reached, the historical data sequence of the device in the target historical time period is input into the time series prediction model, and the predicted device operation data sequence of the device in the target future time period is determined by the time series prediction model; If there is abnormal prediction data in the predicted device operation data sequence, the abnormal prediction data is displayed in the three-dimensional object corresponding to the device.

2. The method according to claim 1, characterized in that The displaying of the abnormal prediction data in the three-dimensional object corresponding to the device includes: The abnormal prediction data and the equipment operation data are distinguished and displayed in the three-dimensional object corresponding to the equipment.

3. The method according to claim 1, characterized in that The time series prediction model is the Prophet model.

4. The method according to any one of claims 1 to 3, characterized in that: The computer room monitoring method is executed by a Hadoop cluster; The Hadoop cluster includes: Zookeeper cluster, HDFS cluster, Yarn cluster, Kafka cluster, Flink cluster and Hbase cluster; The Zookeeper cluster is used to ensure the consistency of data in the Hadoop cluster; The HDFS cluster is used as an underlying file storage system; The Yarn cluster is used to schedule resources for the execution of data processing tasks, wherein the data processing tasks include the task of determining the data sequence of the prediction device operation; The Kafka cluster is used to receive and transmit data in the Hadoop cluster; The Flink cluster is used to perform batch processing and stream processing of data in the Hadoop cluster; The Hbase cluster is used to store the environmental data, the equipment operation data and the time series prediction model, and runs based on the HDFS cluster, using the HDFS cluster as the underlying file storage system.

5. The method according to claim 4, characterized in that The step of inputting the historical data sequence of the device in the target historical time period into the time series prediction model, and determining the predicted device operation data sequence of the device in the target future time period by using the time series prediction model, comprises: Load the time series prediction model from the Hbase cluster through the Flink cluster; The historical data sequence is input into a time series prediction model through a Flink cluster, and a predicted device operation data sequence of the device in a target future time period is determined through the time series prediction model.

6. The method according to any one of claims 1 to 3, characterized in that: After acquiring the equipment operation data of each equipment in the target computer room, the method further includes: If there is abnormal data in the equipment operation data of each device, an early warning message including the abnormal data is generated and sent to a target person.

7. The method according to any one of claims 1 to 3, characterized in that: After acquiring the equipment operation data of each equipment in the target computer room, the method further includes: If there are abnormal data in the device operation data of each device, the time series prediction model is retrained according to the abnormal data, and the retrained time series prediction model is saved.

8. The method according to any one of claims 1 to 3, characterized in that: When the preset time is reached, before the historical data sequence of the device in the target historical time period is input into the time series prediction model, the method further includes: The preset time, the target historical time period and the target future time period set by the user are received.

9. A computer room monitoring device, characterized in that: include: A digital twin construction module, used to construct a digital twin of the target computer room according to the equipment information of the target computer room, wherein the digital twin includes a three-dimensional model of the target computer room and a three-dimensional object of each equipment in the target computer room; A data acquisition module, used to acquire environmental data of the target computer room and equipment operation data of each equipment in the target computer room; A first display module is used to display the digital twin, display the environmental data in the digital twin, and map the device operation data of each device to a three-dimensional object of the corresponding device in the digital twin for display; A data prediction module is used to input the historical data sequence of each device in a target historical time period into a time series prediction model when a preset time is reached, and determine the predicted device operation data sequence of the device in a target future time period through the time series prediction model; The second display module is used to display the abnormal prediction data in the three-dimensional object corresponding to the device if there is abnormal prediction data in the predicted device operation data sequence.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the computer room monitoring method according to any one of claims 1 to 8 is implemented.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the computer room monitoring method described in any one of claims 1 to 8 are implemented.

Images