A machine room server hosting intelligent monitoring management system and method

Abstract

The application relates to the technical field of monitoring management, in particular to a machine room server hosting intelligent monitoring management system and method; the method comprises the following steps: taking a real-time collected multi-channel acoustic signal as input, reconstructing a sound field intensity distribution nephogram in a three-dimensional space of a machine room; from the reconstructed three-dimensional sound field intensity distribution nephogram, extracting a sound field fingerprint of each server in the three-dimensional space region of the machine room as the server; comparing the real-time sound field fingerprint of each server with a baseline sound field fingerprint library of the server in a normal running state in a frequency domain, when the amplitude deviation of a certain frequency band exceeds a preset threshold, determining that the server has abnormal sound, and calculating the three-dimensional space coordinates of the abnormal sound source inside the server; by establishing the sound field fingerprint of each server and comparing with the baseline, non-intrusive automatic tracing and three-dimensional space positioning of abnormal sound are realized.

Description

Technical Field

[0001] This invention relates to the field of monitoring and management technology, and in particular to an intelligent monitoring and management system and method for hosting data center servers. Background Technology

[0002] Data center hosting facilities are the core infrastructure hosting a large number of servers. Early detection of server mechanical failures is crucial in data center operation and maintenance management. Current technologies for detecting abnormal server noise primarily include manual inspection and listening, scanning each server individually with handheld ultrasonic detectors, and monitoring environmental noise by installing single or a small number of sound pressure level sensors within the data center. Manual inspection relies on the experience of maintenance personnel and cannot accurately distinguish the faint abnormal noises of specific servers amidst the background noise of the data center. Handheld detectors require maintenance personnel to operate each server at close range, which is inefficient and cannot achieve 24 / 7 real-time monitoring. Fixed sound pressure level sensors can only measure the overall volume in decibels of the data center, triggering an alarm when the overall volume exceeds a preset threshold, but they cannot distinguish which server is the source of the sound, nor can they pinpoint the specific location of the abnormal noise within that server.

[0003] However, in the aforementioned existing technologies, in the complex acoustic environment of a hosting data center with a large amount of background noise from fans, air conditioners, etc., it is impossible to automatically separate, identify, and three-dimensionally locate abnormal mechanical sounds generated inside a specific server from the mixed acoustic field. Faults in internal server components, such as worn fan bearings, abnormal hard drive head noise, and capacitor squealing, produce sounds with specific frequency characteristics in their early stages, but the sound intensity is extremely low and easily masked by the background noise of the data center. Existing decibel threshold alarm methods cannot distinguish the sources of sound at different frequencies, nor can they associate sound signals with specific physical servers and their internal spatial locations. Therefore, there is an urgent need for an intelligent monitoring and management method that can reconstruct the acoustic field in the three-dimensional space of the data center and establish acoustic fingerprints for each server, thereby achieving automatic source tracing and three-dimensional localization of abnormal sounds. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent monitoring and management system and method for hosting servers in a data center, which aims to solve the technical problem in the prior art that it is impossible to automatically separate, identify and three-dimensionally locate abnormal mechanical sounds generated inside a specific server in a complex acoustic environment such as a hosting data center with a large number of fans, air conditioners and other background noise.

[0005] To achieve the above objectives, the present invention employs an intelligent monitoring and management method for data center server hosting, comprising the following steps: Deploy a distributed microphone array within the hosting data center to collect multi-channel acoustic signals from the data center in real time. The multi-channel acoustic signals acquired in real time are used as input, and the sound field intensity distribution cloud map in the three-dimensional space of the computer room is reconstructed through beamforming algorithm or acoustic holography algorithm. From the reconstructed three-dimensional sound field intensity distribution cloud map, the corresponding sound field frequency domain feature vector of each server in the computer room is extracted as the sound field fingerprint of the server. The real-time sound field fingerprint of each server is compared with the baseline sound field fingerprint library pre-calibrated under normal operating conditions in the frequency domain. When the amplitude deviation of a certain frequency band exceeds the preset threshold, it is determined that the server has abnormal sound. Based on the peak position of the beamforming spatial spectrum, the three-dimensional spatial coordinates of the abnormal sound source inside the server are calculated. The server that is producing abnormal noise and its estimated internal 3D coordinates will be highlighted on the 3D visualization model of the data center for maintenance personnel to locate and check.

[0006] Among the steps, deploying a distributed microphone array within the hosting data center to synchronously collect multi-channel acoustic signals from the data center in real time includes: The distributed microphone array consists of multiple microphone units arranged in a spherical array, planar array, or linear array on the top of the computer room, the side of the rack, or in front of the server. Each microphone unit is synchronized at the nanosecond level via wired or wireless means, with a sampling rate of no less than 48 kHz to ensure effective acquisition of high-frequency abnormal sounds inside the server. The acquired multi-channel acoustic signals are pre-amplified and anti-aliasing filtered before being transmitted to the edge computing unit via Ethernet or USB interface.

[0007] In the step of reconstructing the sound field intensity distribution cloud map in the three-dimensional space of the computer room by taking the real-time acquired multi-channel acoustic signals as input and using beamforming algorithms or acoustic holography algorithms: The beamforming algorithm employs either delay summation beamforming or minimum variance distortion-free response beamforming, which discretizes the three-dimensional space into multiple imaging grid points and calculates the weighted sum of the signals from each channel for each grid point as the sound source intensity at that point. The spatial resolution of the sound field reconstruction is no less than 0.5 meters, and the reconstruction frequency range is from 500 Hz to 20,000 Hz, covering the frequency bands of common abnormal sounds such as server fan bearing wear, hard drive head noise, and capacitor squealing.

[0008] In the step of extracting the corresponding sound field frequency domain feature vector from the three-dimensional spatial region where each server is located in the computer room from the reconstructed three-dimensional sound field intensity distribution cloud map, as the sound field fingerprint of that server: For each server's corresponding three-dimensional spatial region, the time series of sound field intensity of all grid points in that region is spatially averaged to obtain the server's sound field time signal. Perform short-time Fourier transform or wavelet transform on the sound field time signal to extract the energy distribution of each frequency band as the sound field frequency domain feature vector; The acoustic fingerprint is updated every predetermined time interval, which is between ten and sixty seconds.

[0009] Among them, the process of comparing the real-time sound field fingerprint of each server with the pre-calibrated baseline sound field fingerprint library of the server under normal operating conditions in the frequency domain, and determining that the server has abnormal sound when the amplitude deviation of a certain frequency band exceeds a preset threshold, and calculating the three-dimensional spatial coordinates of the abnormal sound source inside the server based on the peak position of the beamforming spatial spectrum: The baseline acoustic field fingerprint database is obtained by statistically modeling the long-term acoustic field fingerprint data of the server under different time periods and different load conditions, including the mean and standard deviation of the amplitude for each frequency band. The amplitude deviation is calculated by subtracting the baseline mean from the real-time acoustic field fingerprint amplitude and then dividing by the baseline standard deviation to obtain the standard deviation score for that frequency band. The preset threshold is set to an absolute value of the deviation standard score greater than 2.5, that is, when the deviation from the baseline exceeds 2.5 times the standard deviation, it is judged as abnormal.

[0010] Among them, the process of comparing the real-time sound field fingerprint of each server with the pre-calibrated baseline sound field fingerprint library of the server under normal operating conditions in the frequency domain, and determining that the server has abnormal sound when the amplitude deviation of a certain frequency band exceeds a preset threshold, and calculating the three-dimensional spatial coordinates of the abnormal sound source inside the server based on the peak position of the beamforming spatial spectrum: When the acoustic field fingerprints collected multiple times in a row exceed the preset threshold in the same frequency band, it is judged as a persistent anomaly and the anomaly level is increased. If the threshold is exceeded only momentarily in a single data collection and then recovers, it is judged as a transient disturbance and the abnormality level is reduced.

[0011] Among the steps, highlighting the server that is generating abnormal noise and its estimated internal 3D coordinates on the 3D visualization model of the data center for maintenance personnel to locate and inspect: The 3D visualization model of the data center is pre-built using laser scanning or CAD drawing import, and includes the 3D structural model of server racks, servers, and network equipment. The anomaly markers use a pulsed bright flashing effect and color to distinguish the anomaly level. Red indicates a serious anomaly that requires immediate action, while yellow indicates a general anomaly that requires planned maintenance. Maintenance personnel can click on the anomaly marker on the display terminal to view the server's historical sound field fingerprint trend curve, abnormal frequency band, and duration information.

[0012] The process involves comparing the real-time sound field fingerprint of each server with a pre-calibrated baseline sound field fingerprint database for that server under normal operating conditions in the frequency domain. If the amplitude deviation of a certain frequency band exceeds a preset threshold, it is determined that the server has abnormal sound. Then, based on the peak position of the beamforming spatial spectrum, the three-dimensional spatial coordinates of the abnormal sound source inside the server are calculated: Based on the frequency range where abnormal amplitude deviations occur, a matching is performed with a pre-established fault acoustic feature library; The fault acoustic feature library includes low-to-mid frequency continuous spectrum features corresponding to fan bearing wear, high frequency pulse features corresponding to hard disk head noise, and narrowband single frequency features corresponding to capacitor whistling. Once a match is successful, the abnormal sound source type will be pushed to the display terminal and displayed together with the abnormal marker.

[0013] Before the step of deploying a distributed microphone array in the hosting data center to synchronously collect multi-channel acoustic signals in real time: When the system is first deployed and all servers in the computer room are confirmed to be running normally, multi-channel acoustic signals are continuously collected for no less than 24 hours to complete the element position calibration of the distributed microphone array. At the same time, an initial baseline acoustic field fingerprint database is established for each server. The baseline acoustic field fingerprint database is updated regularly using data from periods when no anomalies are confirmed during the continuous operation of the system.

[0014] This invention also provides an intelligent monitoring and management system for server hosting in a data center, comprising an acoustic acquisition module, a sound field reconstruction module, a sound field fingerprint extraction module, an anomaly comparison and location module, and a visualization and marking module, wherein: The acoustic acquisition module: deploys a distributed microphone array in the hosting room to collect multi-channel acoustic signals in the room in real time. The sound field reconstruction module takes the real-time acquired multi-channel acoustic signals as input and reconstructs the sound field intensity distribution cloud map in the three-dimensional space of the computer room through beamforming algorithm or acoustic holography algorithm. The sound field fingerprint extraction module: From the reconstructed three-dimensional sound field intensity distribution cloud map, extracts the corresponding sound field frequency domain feature vector of the three-dimensional spatial region where each server is located in the computer room, as the sound field fingerprint of the server. The anomaly comparison and localization module compares the real-time sound field fingerprint of each server with the baseline sound field fingerprint library pre-calibrated under normal operating conditions in the frequency domain. When the amplitude deviation of a certain frequency band exceeds a preset threshold, it determines that the server has abnormal sound. Based on the peak position of the beamforming spatial spectrum, it calculates the three-dimensional spatial coordinates of the abnormal sound source inside the server. The visualization marking module highlights the server that generates abnormal noise and its estimated internal 3D coordinates on the 3D visualization model of the data center, allowing maintenance personnel to locate and view the location.

[0015] This invention discloses an intelligent monitoring and management system and method for hosting servers in a data center. A distributed microphone array is deployed within the hosting data center to synchronously collect multi-channel acoustic signals in real time. These real-time collected multi-channel acoustic signals are used as input, and a sound field intensity distribution cloud map in the three-dimensional space of the data center is reconstructed using beamforming or acoustic holography algorithms. From the reconstructed three-dimensional sound field intensity distribution cloud map, the corresponding sound field frequency domain feature vector is extracted for each server's three-dimensional spatial region, serving as the server's sound field fingerprint. The real-time sound field fingerprint of each server is compared in the frequency domain with a pre-calibrated baseline sound field fingerprint database for that server under normal operating conditions. When the amplitude deviation of a certain frequency band exceeds a preset threshold, the server is determined to have abnormal sound. Based on the peak position of the beamforming spatial spectrum, the three-dimensional spatial coordinates of the abnormal sound source within the server are calculated. The server generating the abnormal sound and its estimated internal three-dimensional coordinates are highlighted on a three-dimensional visualization model of the data center for maintenance personnel to locate and view. This invention achieves real-time reconstruction of the three-dimensional sound field of a hosting data center through a distributed microphone array and beamforming or acoustic holography algorithms. This overcomes the limitations of traditional single-point sound pressure level sensors, which can only measure the overall volume and cannot distinguish the sound source. By extracting the frequency domain feature vector of the sound field for each server as a unique sound field fingerprint and comparing it with a baseline sound field fingerprint database established through long-term statistical modeling, the invention can sensitively identify faint abnormal sounds caused by early mechanical failures such as fan bearing wear or hard drive head noise within a specific server, even in noisy background noise. It also automatically calculates the three-dimensional spatial coordinates of the abnormal sound source inside the server. Finally, the abnormal information is highlighted on a three-dimensional visualization model of the data center, allowing maintenance personnel to intuitively locate faulty servers and their internal components without manual inspection. This represents a leap from traditional passive listening to active acoustic imaging and three-dimensional positioning, significantly improving the intelligence level and early fault detection capabilities of hosting data center operations. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of the steps of the intelligent monitoring and management method for server hosting in the data center according to the present invention.

[0018] Figure 2 This is a flowchart of steps S100 of the present invention.

[0019] Figure 3 This is a flowchart of steps S200 of the present invention.

[0020] Figure 4 This is a flowchart of steps S300 of the present invention.

[0021] Figure 5 This is a flowchart of steps S400 of the present invention.

[0022] Figure 6 This is a flowchart of steps S500 of the present invention.

[0023] Figure 7 This is a schematic diagram of the intelligent monitoring and management system for server hosting in the computer room according to the present invention.

[0024] 601-Acoustic Acquisition Module, 602-Sound Field Reconstruction Module, 603-Sound Field Fingerprint Extraction Module, 604-Anomaly Comparison and Localization Module, 605-Visual Marking Module. Detailed Implementation

[0025] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0026] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0027] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0028] Please see Figures 1-6 This invention provides an intelligent monitoring and management method for server hosting in a data center, comprising the following steps: S100: Deploys a distributed microphone array in the hosting room to collect multi-channel acoustic signals in the room in real time. In this embodiment, a distributed microphone array is deployed within the hosting data center to synchronously acquire multi-channel acoustic signals within the data center in real time. Specifically, this includes: S101: When the system is initially deployed and all servers in the computer room are confirmed to be running normally, continuously collect multi-channel acoustic signals for no less than 24 hours to complete the element position calibration of the distributed microphone array. S102: At the same time, an initial baseline sound field fingerprint database is established for each server. The baseline sound field fingerprint database is updated periodically during the continuous operation of the system using data from time periods when no anomalies are confirmed. S103: The distributed microphone array consists of multiple microphone units arranged in a spherical array, planar array, or linear array on the top of the computer room, the side of the server rack, or in front of the server. S104: Each microphone unit is synchronized at the nanosecond level via wired or wireless means, with a sampling rate of no less than 48 kHz, to ensure effective acquisition of high-frequency abnormal sounds inside the server. S105: The acquired multi-channel acoustic signals are pre-amplified and anti-aliasing filtered before being transmitted to the edge computing unit via Ethernet or USB interface.

[0029] In the hosting data center, microphone units are suspended in a spherical array from the center of the ceiling, based on the space size and server rack layout. Linear arrays are also deployed along the sides of each rack to ensure comprehensive 3D sound field coverage for each server. After initial power-on, the system performs at least 24 hours of continuous data acquisition. This phase assumes all servers in the data center are operating normally and without abnormal noise. The collected data serves two purposes: first, by using the correlation of ambient background noise received by each microphone unit, the relative 3D position coordinates of each array element are determined through a cross-correlation delay estimation algorithm, eliminating the need for manual measurement; second, an initial baseline sound field fingerprint database is established for each server. This database records the statistical distribution of sound energy across frequency bands under normal operating conditions. During subsequent system operation, data from periods when the sound field fingerprint does not trigger anomalies is automatically selected daily, and the baseline mean and standard deviation are updated using a sliding mechanism to adapt to normal sound field drift caused by seasonal temperature and humidity changes or regular changes in server load. The microphone unit sampling rate is set to 48 kHz to ensure Nyquist sampling requirements for high-frequency abnormal sounds such as capacitor howling at frequencies up to 20 kHz within the server. The multi-channel signals are amplified by a preamplifier with a fixed gain, then passed through an anti-aliasing filter with a cutoff frequency of 20 kHz, and finally packaged and transmitted to the edge computing unit via a real-time streaming protocol through the Ethernet interface.

[0030] S200: It takes the real-time acquired multi-channel acoustic signals as input and reconstructs the sound field intensity distribution cloud map in the three-dimensional space of the computer room through beamforming algorithm or acoustic holography algorithm. In this embodiment, multi-channel acoustic signals acquired in real time are used as input, and a sound field intensity distribution cloud map in the three-dimensional space of the computer room is reconstructed using beamforming or acoustic holography algorithms. Specifically, this includes: S201: The beamforming algorithm adopts delayed summation beamforming or minimum variance distortion-free response beamforming, which discretizes the three-dimensional space into multiple imaging grid points, and calculates the weighted sum of the signals of each channel for each grid point as the sound source intensity at that point. S202: The spatial resolution of the sound field reconstruction is no less than 0.5 meters, and the reconstruction frequency range is from 500 Hz to 20,000 Hz, covering the frequency bands of common abnormal sounds such as server fan bearing wear, hard drive head noise, and capacitor squealing.

[0031] The edge computing unit, upon receiving synchronized, multi-channel data, first discretizes the three-dimensional space of the computer room, using a 0.2-meter cubic grid as the imaging unit to cover the entire space. For each grid point, a delay-summing beamforming algorithm is employed: based on the spatial distance from the grid point to each microphone unit, the corresponding time delay of sound wave propagation is calculated, and the sampled values ​​at the corresponding moment are extracted from the signal buffers of each channel, summed to obtain the instantaneous sound source intensity at that grid point. When higher resolution imaging of sound sources in strongly coherent noise environments is required, the algorithm switches to minimum variance distortion-free response beamforming, using adaptive weight vectors to suppress interference noise in directions other than grid points. Each imaging process handles a one-second signal frame, generating a three-dimensional sound field intensity distribution cloud map. The spatial resolution is 0.5 meters, and the frequency coverage is limited to 500 Hz to 20 kHz, encompassing all common mechanical fault sound bands, from low-frequency fan vibration to high-frequency hard drive noise.

[0032] S300: From the reconstructed three-dimensional sound field intensity distribution cloud map, extract the corresponding sound field frequency domain feature vector of the three-dimensional spatial region where each server is located in the computer room, as the sound field fingerprint of the server. In this embodiment, from the reconstructed three-dimensional sound field intensity distribution cloud map, the corresponding sound field frequency domain feature vector of each server in the data center is extracted as the sound field fingerprint of that server. Specifically, this includes: S301: For each server's corresponding three-dimensional spatial region, spatially average the sound field intensity time series of all grid points in that region to obtain the server's sound field time signal. S302: Perform short-time Fourier transform or wavelet transform on the sound field time signal to extract the energy distribution of each frequency band as the sound field frequency domain feature vector; S303: The acoustic field fingerprint is updated once every predetermined time interval, which is ten to sixty seconds.

[0033] In the 3D sound field intensity distribution cloud map, the voxel region corresponding to each server is delineated based on the 3D bounding box occupied by each server in the pre-imported 3D visualization model of the data center. The spatial arithmetic mean of the sound field intensity time series of all voxels within this region is taken to obtain a single-channel time signal representing the overall acoustic response of the server. This time signal is then transformed into a time-frequency domain spectrum using a short-time Fourier transform with a window length of 1024 points and 50% overlap. On the frequency axis, frequency bands are divided at 100 Hz intervals, and the energy integral of each frequency band within a time window is calculated, ultimately forming a sound field frequency domain feature vector with a dimension equal to the number of frequency bands, which is the sound field fingerprint of the server. The update interval of the sound field fingerprint is set to ten seconds, which can capture sudden abnormal noises in a timely manner while avoiding the computational overhead caused by overly frequent baseline comparisons.

[0034] S400: The real-time sound field fingerprint of each server is compared with the baseline sound field fingerprint library pre-calibrated under normal operating conditions in the frequency domain. When the amplitude deviation of a certain frequency band exceeds the preset threshold, it is determined that there is abnormal sound in the server. Based on the peak position of the beamforming spatial spectrum, the three-dimensional spatial coordinates of the abnormal sound source inside the server are calculated. In this embodiment, the real-time sound field fingerprint of each server is compared in the frequency domain with a pre-calibrated baseline sound field fingerprint database for that server under normal operating conditions. When the amplitude deviation of a certain frequency band exceeds a preset threshold, it is determined that the server has abnormal sound, and the three-dimensional spatial coordinates of the abnormal sound source inside the server are calculated based on the peak position of the beamforming spatial spectrum. Specifically, this includes: S401: The baseline acoustic fingerprint database is obtained by statistically modeling the long-term acoustic fingerprint data of the server under different time periods and different load conditions, including the mean and standard deviation of the amplitude of each frequency band. S402: Amplitude deviation is calculated by subtracting the baseline mean from the real-time acoustic field fingerprint amplitude and then dividing by the baseline standard deviation to obtain the standard deviation score for that frequency band. S403: The preset threshold is set to an absolute value of the deviation standard score greater than 2.5, that is, when the deviation from the baseline exceeds 2.5 times the standard deviation, it is judged as abnormal; S404: When the acoustic field fingerprint collected multiple times continuously exceeds the preset threshold in the same frequency band, it is judged as a persistent anomaly and the anomaly level is increased. S405: When the threshold is exceeded only momentarily in a single data acquisition and then recovers, it is judged as a transient disturbance and the abnormality level is reduced; S406: Match the frequency band range where the abnormal amplitude deviation occurs with a pre-established fault acoustic feature library; S407: The fault acoustic feature library includes low-to-medium frequency continuous spectrum features corresponding to fan bearing wear, high-frequency pulse features corresponding to hard disk head noise, and narrowband single-frequency features corresponding to capacitor whistling. S408: After a successful match, the abnormal sound source type will be pushed to the display terminal and displayed together with the abnormal marker.

[0035] The baseline acoustic fingerprint database used for comparison contains the mean and standard deviation of energy for each frequency band, obtained by fitting a Gaussian distribution to historical normal data of the server under different time periods and load conditions. Upon arrival of the real-time acoustic fingerprint, a standard deviation score is calculated for each frequency band, which is the real-time amplitude minus the baseline mean and then divided by the baseline standard deviation. When the absolute value of the standard deviation score for any frequency band exceeds 2.5, an anomaly is determined for that frequency band. At this point, the beamforming spatial spectrum corresponding to the anomaly frame is traced back, and the coordinates of the three grid points with the strongest energy contribution in that frequency band are extracted. Their spatial geometric centroids are used as the three-dimensional predicted coordinates of the anomaly source within the server. Furthermore, if anomalies are detected in the same frequency band for five consecutive acoustic fingerprint update cycles, it is determined to be a persistent anomaly, and the alarm level is upgraded; if it only occurs in a single cycle and then recovers, it is recorded as a transient disturbance, and no alarm upgrade is performed. Once the abnormal frequency band is identified, it is matched against the fault acoustic feature database: continuous broadband spectrum in the mid-to-low frequency band corresponds to fan bearing wear, periodic pulses in the high frequency band correspond to abnormal noise from the hard drive head, and narrowband single-frequency spikes correspond to capacitor whistling. Upon successful matching, the fault type label is output along with the alarm.

[0036] S500: Highlights the server that is producing abnormal noise and its estimated internal 3D coordinates on the 3D visualization model of the data center for maintenance personnel to locate and check.

[0037] In this implementation, the server generating the abnormal sound and its estimated internal 3D coordinates are highlighted on the 3D visualization model of the data center for maintenance personnel to locate and inspect. Specifically, this includes: S501: The 3D visualization model of the data center is pre-built through laser scanning or CAD drawing import, including the 3D structural model of server racks, servers, and network equipment. S502: The anomaly marker uses a pulsed high-brightness flashing effect and distinguishes the anomaly level by color. Red indicates a serious anomaly that requires immediate handling, while yellow indicates a general anomaly that requires planned maintenance. S503: Maintenance personnel can click on the anomaly marker on the display terminal to view the server's historical sound field fingerprint trend curve, abnormal frequency band, and duration information.

[0038] The 3D visualization model of the data center is generated by pre-collecting point cloud data of the data center using laser scanning equipment and then reconstructing it into a grid. Each server, rack, and network device in the model has an independent identifier and real spatial coordinates. When a server is detected to have abnormal noise, a pulsed, brightly flashing spherical marker is generated within the server model corresponding to the estimated 3D coordinates. A red ball indicates a serious anomaly, meaning the deviation standard score exceeds five and the condition is persistent, requiring immediate on-site handling by maintenance personnel. A yellow ball indicates a general anomaly, meaning the deviation standard score is between 2.5 and 5, and the server can be included in the planned maintenance list. Maintenance personnel can directly click on the marker on the display terminal, and the system will pop up an information panel displaying the server's historical sound field fingerprint trend curve for the past 24 hours, the specific frequency values ​​of the abnormal frequency band, the peak time of the deviation standard score, and the matched fault types. The display terminal supports rotation, scaling, and sectioning of the 3D model, allowing users to view the spatial correspondence between the marker position and the server's internal structure from any angle.

[0039] Corresponding to the aforementioned embodiments of the intelligent monitoring and management method for hosting data center servers, this application also provides embodiments of the intelligent monitoring and management system for hosting data center servers.

[0040] Figure 7 This is a block diagram illustrating an intelligent monitoring and management system for server hosting in a data center, according to an exemplary embodiment. (Refer to...) Figure 7 The system may include: an acoustic acquisition module 601, a sound field reconstruction module 602, a sound field fingerprint extraction module 603, an anomaly comparison and localization module 604, and a visualization and marking module 605, wherein: The acoustic acquisition module 601: deploys a distributed microphone array in the hosting room to acquire multi-channel acoustic signals in the room in real time. The sound field reconstruction module 602: takes the real-time acquired multi-channel acoustic signals as input and reconstructs the sound field intensity distribution cloud map in the three-dimensional space of the computer room through beamforming algorithm or acoustic holography algorithm. The sound field fingerprint extraction module 603: From the reconstructed three-dimensional sound field intensity distribution cloud map, extracts the corresponding sound field frequency domain feature vector of the three-dimensional spatial region where each server is located in the computer room, as the sound field fingerprint of the server. The anomaly comparison and positioning module 604 compares the real-time sound field fingerprint of each server with the baseline sound field fingerprint library pre-calibrated under normal operating conditions in the frequency domain. When the amplitude deviation of a certain frequency band exceeds a preset threshold, it determines that there is an abnormal sound in the server, and calculates the three-dimensional spatial coordinates of the abnormal sound source inside the server based on the peak position of the beamforming spatial spectrum. The visualization marking module 605 highlights the server that generates abnormal noise and its estimated internal 3D coordinates on the 3D visualization model of the computer room, so that maintenance personnel can locate and view it.

[0041] In this embodiment, the acoustic acquisition module 601 deploys a distributed microphone array within the hosting data center to synchronously acquire multi-channel acoustic signals from the data center in real time; the sound field reconstruction module 602 takes the real-time acquired multi-channel acoustic signals as input and reconstructs a sound field intensity distribution cloud map in the three-dimensional space of the data center using beamforming or acoustic holography algorithms; the sound field fingerprint extraction module 603 extracts the corresponding sound field frequency domain feature vector from the reconstructed three-dimensional sound field intensity distribution cloud map for the three-dimensional spatial region where each server is located in the data center, using it as the sound field of that server. The fingerprint; the anomaly comparison and positioning module 604: compares the real-time sound field fingerprint of each server with the baseline sound field fingerprint library pre-calibrated under normal operating conditions in the frequency domain. When the amplitude deviation of a certain frequency band exceeds a preset threshold, it is determined that the server has abnormal sound, and the three-dimensional spatial coordinates of the abnormal sound source inside the server are calculated according to the peak position of the beamforming spatial spectrum; the visualization marking module 605: highlights the server that generates abnormal sound and its estimated internal three-dimensional coordinates on the three-dimensional visualization model of the computer room for operation and maintenance personnel to locate and view.

[0042] Regarding the system in the above embodiments, the specific ways in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0043] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

[0044] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A method for intelligent monitoring and management of server hosting in a data center, characterized in that, Includes the following steps: A distributed microphone array is deployed in the hosting data center to collect multi-channel acoustic signals in real time. The multi-channel acoustic signals acquired in real time are used as input, and the sound field intensity distribution cloud map in the three-dimensional space of the computer room is reconstructed through beamforming algorithm or acoustic holography algorithm. From the reconstructed three-dimensional sound field intensity distribution cloud map, the corresponding sound field frequency domain feature vector of each server in the computer room is extracted as the sound field fingerprint of the server. The real-time sound field fingerprint of each server is compared with the baseline sound field fingerprint library pre-calibrated under normal operating conditions in the frequency domain. When the amplitude deviation of a certain frequency band exceeds the preset threshold, it is determined that the server has abnormal sound. Based on the peak position of the beamforming spatial spectrum, the three-dimensional spatial coordinates of the abnormal sound source inside the server are calculated. The server that is producing abnormal noise and its estimated internal 3D coordinates will be highlighted on the 3D visualization model of the data center for maintenance personnel to locate and check.

2. The intelligent monitoring and management method for server hosting in a data center as described in claim 1, characterized in that, In the step of deploying a distributed microphone array in the hosting data center to synchronously acquire multi-channel acoustic signals in the data center in real time: The distributed microphone array consists of multiple microphone units arranged in a spherical array, planar array, or linear array on the top of the computer room, the side of the rack, or in front of the server. Each microphone unit is synchronized at the nanosecond level via wired or wireless means, with a sampling rate of no less than 48 kHz to ensure effective acquisition of high-frequency abnormal sounds inside the server. The acquired multi-channel acoustic signals are pre-amplified and anti-aliasing filtered before being transmitted to the edge computing unit via Ethernet or USB interface.

3. The intelligent monitoring and management method for server hosting in a data center as described in claim 1, characterized in that, In the step of reconstructing the sound field intensity distribution cloud map in the three-dimensional space of the computer room by using real-time acquired multi-channel acoustic signals as input and employing beamforming or acoustic holography algorithms: The beamforming algorithm employs either delay summation beamforming or minimum variance distortion-free response beamforming, which discretizes the three-dimensional space into multiple imaging grid points and calculates the weighted sum of the signals from each channel for each grid point as the sound source intensity at that point. The spatial resolution of the sound field reconstruction is no less than 0.5 meters, and the reconstruction frequency range is from 500 Hz to 20,000 Hz, covering the frequency bands of common abnormal sounds such as server fan bearing wear, hard drive head noise, and capacitor squealing.

4. The intelligent monitoring and management method for server hosting in a data center as described in claim 1, characterized in that, In the step of extracting the corresponding sound field frequency domain feature vector from the three-dimensional spatial region where each server is located in the computer room, based on the reconstructed three-dimensional sound field intensity distribution cloud map, as the sound field fingerprint of that server: For each server's corresponding three-dimensional spatial region, the time series of sound field intensity of all grid points in that region is spatially averaged to obtain the server's sound field time signal. Perform short-time Fourier transform or wavelet transform on the sound field time signal to extract the energy distribution of each frequency band as the sound field frequency domain feature vector; The acoustic fingerprint is updated every predetermined time interval, which is between ten and sixty seconds.

5. The intelligent monitoring and management method for server hosting in a data center as described in claim 1, characterized in that, In the step of comparing the real-time sound field fingerprint of each server with the pre-calibrated baseline sound field fingerprint library of the server under normal operating conditions in the frequency domain, when the amplitude deviation of a certain frequency band exceeds a preset threshold, it is determined that the server has abnormal sound, and the three-dimensional spatial coordinates of the abnormal sound source inside the server are calculated based on the peak position of the beamforming spatial spectrum: The baseline acoustic field fingerprint database is obtained by statistically modeling the long-term acoustic field fingerprint data of the server under different time periods and different load conditions, including the mean and standard deviation of the amplitude for each frequency band. The amplitude deviation is calculated by subtracting the baseline mean from the real-time acoustic field fingerprint amplitude and then dividing by the baseline standard deviation to obtain the standard deviation score for that frequency band. The preset threshold is set to an absolute value of the deviation standard score greater than 2.5, that is, when the deviation from the baseline exceeds 2.5 times the standard deviation, it is judged as abnormal.

6. The intelligent monitoring and management method for server hosting in a data center as described in claim 5, characterized in that, In the step of comparing the real-time sound field fingerprint of each server with the pre-calibrated baseline sound field fingerprint library of the server under normal operating conditions in the frequency domain, when the amplitude deviation of a certain frequency band exceeds a preset threshold, it is determined that the server has abnormal sound, and the three-dimensional spatial coordinates of the abnormal sound source inside the server are calculated based on the peak position of the beamforming spatial spectrum: When the acoustic field fingerprints collected multiple times in a row exceed the preset threshold in the same frequency band, it is judged as a persistent anomaly and the anomaly level is increased. If the threshold is exceeded only momentarily in a single data collection and then recovers, it is judged as a transient disturbance and the abnormality level is reduced.

7. The intelligent monitoring and management method for server hosting in a data center as described in claim 1, characterized in that, In the step of highlighting the server that is generating abnormal noise and its estimated internal 3D coordinates on the 3D visualization model of the data center for maintenance personnel to locate and inspect: The 3D visualization model of the data center is pre-built using laser scanning or CAD drawing import, and includes the 3D structural model of server racks, servers, and network equipment. The anomaly markers use a pulsed bright flashing effect and color to distinguish the anomaly level. Red indicates a serious anomaly that requires immediate action, while yellow indicates a general anomaly that requires planned maintenance. Maintenance personnel can click on the anomaly marker on the display terminal to view the server's historical sound field fingerprint trend curve, abnormal frequency band, and duration information.

8. The intelligent monitoring and management method for server hosting in a data center as described in claim 1, characterized in that, After comparing the real-time sound field fingerprint of each server with the pre-calibrated baseline sound field fingerprint library of the server under normal operating conditions in the frequency domain, and determining that the server has abnormal sound when the amplitude deviation of a certain frequency band exceeds a preset threshold, and calculating the three-dimensional spatial coordinates of the abnormal sound source inside the server based on the peak position of the beamforming spatial spectrum: Based on the frequency range where abnormal amplitude deviations occur, a matching is performed with a pre-established fault acoustic feature library; The fault acoustic feature library includes low-to-mid frequency continuous spectrum features corresponding to fan bearing wear, high frequency pulse features corresponding to hard disk head noise, and narrowband single frequency features corresponding to capacitor whistling. Once a match is successful, the abnormal sound source type will be pushed to the display terminal and displayed together with the abnormal marker.

9. The intelligent monitoring and management method for server hosting in a data center as described in claim 1, characterized in that, Before deploying a distributed microphone array within the hosting data center to synchronously acquire multi-channel acoustic signals from the data center in real time: When the system is first deployed and all servers in the computer room are confirmed to be running normally, multi-channel acoustic signals are continuously collected for no less than 24 hours to complete the element position calibration of the distributed microphone array. At the same time, an initial baseline acoustic field fingerprint database is established for each server. The baseline acoustic field fingerprint database is updated regularly using data from periods when no anomalies are confirmed during the continuous operation of the system.

10. A smart monitoring and management system for server hosting in a data center, employing the smart monitoring and management method for server hosting in a data center as described in claim 1, characterized in that, It includes an acoustic acquisition module, a sound field reconstruction module, a sound field fingerprint extraction module, an anomaly comparison and localization module, and a visualization and marking module, among which: The acoustic acquisition module: deploys a distributed microphone array in the hosting room to collect multi-channel acoustic signals in the room in real time. The sound field reconstruction module takes the real-time acquired multi-channel acoustic signals as input and reconstructs the sound field intensity distribution cloud map in the three-dimensional space of the computer room through beamforming algorithm or acoustic holography algorithm. The sound field fingerprint extraction module: From the reconstructed three-dimensional sound field intensity distribution cloud map, extracts the corresponding sound field frequency domain feature vector of the three-dimensional spatial region where each server is located in the computer room, as the sound field fingerprint of the server. The anomaly comparison and localization module compares the real-time sound field fingerprint of each server with the baseline sound field fingerprint library pre-calibrated under normal operating conditions in the frequency domain. When the amplitude deviation of a certain frequency band exceeds a preset threshold, it determines that the server has abnormal sound. Based on the peak position of the beamforming spatial spectrum, it calculates the three-dimensional spatial coordinates of the abnormal sound source inside the server. The visualization marking module highlights the server that generates abnormal noise and its estimated internal 3D coordinates on the 3D visualization model of the data center, allowing maintenance personnel to locate and view the location.

Images