A data center operating temperature monitoring method and monitoring system

Abstract

The application discloses a data center operation temperature monitoring method and a monitoring system, relates to the technical field of temperature monitoring, and is used for solving the abnormal temperature misjudgment problem caused by the temperature adaptive change of the same data center in different positions. The device combination participating in data storage in the data center is identified, the device storage and temperature information are extracted, the core device is screened, and the temperature thereof is collected. The overall trend temperature of the device combination is calculated in combination with the temperature, operation storage proportion and calling duration of each marked device. The operation temperature characteristics of the current period are extracted by using principal component analysis, and the temperature time sequence deviation is calculated based on a sliding time window. Meanwhile, the ectopic data of the same device combination under different coordinates is acquired, an ectopic temperature characteristic model is constructed, the temperature position sequence deviation of the current coordinate and the ectopic characteristic mean value is calculated, and finally the probability of temperature abnormality of the target coordinate in the future time is predicted by fusing the temperature time sequence deviation and the temperature position sequence deviation, so that it is judged whether the early warning is triggered.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of temperature monitoring, more particularly, the present application relates to a data center operation temperature monitoring method and monitoring system. BACKGROUND

[0002] Temperature monitoring technology is a technology for real-time or periodic detection and recording of environmental, equipment or material temperature, and when applied to data center operation temperature monitoring, it can speed up temperature monitoring response and improve data center security.

[0003] The prior art has the following disadvantages:

[0004] Traditional data center temperature monitoring is performed by monitoring the temperature of the core heating device of the equipment or the overall equipment, without considering that the heating condition of the data center changes over time and space, so when two data centers with the same equipment combination are placed in two different environmental locations, the temperature requirements will change, resulting in abnormal temperature misjudgment and data damage. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a data center operation temperature monitoring method, which analyzes the actual temperature requirements of the data center with the same setting combination at different times and coordinates to determine whether to issue a warning, thereby solving the problems raised in the background art.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical scheme, a data center operation temperature monitoring method, comprising the following steps:

[0007] Step S1: Obtain the equipment combination participating in data storage, count the data storage items of each device in the equipment combination, select the device with the largest data storage item for temperature detection, and record the core device temperature of the equipment combination;

[0008] Step S2: Classify and mark each device in the equipment combination, detect the device temperature of each marked device, collect the running storage ratio and data calling time length of each marked device, set the marking weight, and calculate the overall trend temperature of the equipment combination by comprehensively marking the device temperature and marking weight of the device;

[0009] Step S3: Analyze the operation temperature characteristics of the data center according to the core device temperature and overall trend temperature of the equipment combination, call the historical data, calculate the historical temperature characteristics of the data center using the historical data, and analyze the temperature time sequence deviation of the data center in combination with the operation temperature characteristics;

[0010] Step S4: Obtain the ectopic data of the same device combination, calculate the ectopic temperature characteristics of the data center using the ectopic data, analyze the temperature position sequence deviation of the data center in combination with the operating temperature characteristics, and integrate the operating temperature characteristics of the data center by comprehensively considering the temperature time sequence deviation and the temperature position sequence deviation.

[0011] In a preferred embodiment, in step S1, all devices participating in data storage are identified as a device combination, data storage item information is read from each device, and the data storage item information is the used storage space of each device; the device with the largest value is selected as the core device;

[0012] The temperature data collected by the internal temperature sensor of the core device in real time is called as the core device temperature C of the device combination c .

[0013] In a preferred embodiment, in step S2, the running storage ratio refers to the ratio of the device running storage space to the total storage capacity of the device;

[0014] The start and end time stamps of the storage data calls in each marked device are merged to accumulate the actual data read / write operation duration of each marked device as the data call duration.

[0015] In a preferred embodiment, in step S2, the data call duration is standardized to generate a data call coefficient, and the marking weight of the corresponding marked device is set according to the data call coefficient and the running storage ratio: w i = α × S i + β × T i , wherein w i is the marking weight of each marked device, S i is the running storage ratio of each marked device, T i is the data call duration ratio in unit time, α is the importance coefficient of the running storage ratio, and β is the importance coefficient of the data call duration ratio in unit time;

[0016] The real-time temperature of each marked device is weighted and averaged with the marking weight to obtain the overall trend temperature of the device combination.

[0017] In a preferred embodiment, in step S3, the principal component analysis method is used to analyze the operating temperature characteristics of the data center according to the core device stability and the overall trend temperature of the device combination, and the specific steps are as follows:

[0018] Data preparation: Set the core device temperature sequence and the overall trend temperature sequence of the current device combination in the time window to form a two-dimensional matrix, and generate a standardized sample matrix after standardizing the two-dimensional matrix;

[0019] Covariance matrix calculation: calculate the covariance matrix of the standardized sample matrix;

[0020] Eigenvalue decomposition: perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues and eigenvectors, and select the eigenvector with the largest eigenvalue as the principal component direction;

[0021] Eigenvalue projection: project the standardized temperature sample at the current time point t onto the principal component direction to obtain the operating temperature eigenvalue.

[0022] In a preferred embodiment, in step S3, a historical period is selected as a historical analysis time, and a plurality of time points are set within the historical analysis time to call historical data, which is the core device temperature and overall trend temperature of the device combination at each time point within the historical analysis time; a plurality of historical temperature features are generated based on the principal component analysis method and combined into a historical temperature feature sequence.

[0023] In a preferred embodiment, in step S3, a sliding window with a length of k is set, and the historical temperature feature mean value at each time point is calculated within the sliding window. The temperature time series deviation of the current time point t is obtained by taking the absolute value of the difference between the current time point temperature eigenvalue and the historical temperature feature mean value.

[0024] In a preferred embodiment, in step S4, the ex situ data is the operating data of the current device combination under other operating coordinates, and the ex situ temperature features of the data center are calculated using the ex situ data. The core device temperature and overall trend temperature under different operating coordinates are constructed into a matrix and standardized, and the standardized results are used to generate a plurality of ex situ temperature features using the principal component analysis method;

[0025] The temperature position sequence deviation is calculated by calculating the moving average deviation of the current operating coordinate L0 temperature eigenvalue relative to the ex situ temperature feature mean value.

[0026] In a preferred embodiment, in step S4, the temperature time series deviation and the temperature position sequence deviation are used as input features to construct a Bayesian binary classification inference model to integrate and predict the operating temperature features of the data center. The specific steps are as follows:

[0027] Define abnormal event: set temperature anomaly value Y, when temperature anomaly value Y = 1, it is determined that the temperature is abnormal; when temperature anomaly value Y = 0, it is determined that the temperature is normal;

[0028] Calculate probability: define the temperature time series deviation and the temperature position sequence deviation as ΔC (tf) and ΔC (L0) , respectively, combine them into a deviation data pair (ΔC (tf) , ΔC (L0) ), and calculate the probability value of the temperature anomaly according to the deviation data pair;

[0029] Abnormal prediction: when the deviation data pair calculated temperature abnormal probability value is lower than the preset optimal decision boundary, no temperature abnormality warning is performed; when the deviation data pair calculated temperature abnormal probability value is higher than the preset optimal decision boundary, temperature abnormality warning is performed.

[0030] A data center operation temperature monitoring system comprises a device identification module, a label modeling module, a temperature feature extraction module and an abnormal prediction module.

[0031] The device identification module is used for identifying all devices participating in data storage, counting the used storage space of each device in the combination, screening out core devices and obtaining the temperature readings of the internal temperature sensors of the core devices.

[0032] The label modeling module is used for classifying and labeling non-core devices in the device combination, collecting the real-time temperature, operation storage ratio and data calling duration of each device in a unit period, and calculating the overall trend temperature of the device combination.

[0033] The temperature feature extraction module takes the temperature of the core device and the overall trend temperature as input to form the current operation temperature feature value, and calls the temperature data in the historical time window to calculate the sliding average difference between the current operation temperature feature and the historical average value, thereby forming the temperature time series deviation index of the data center.

[0034] The abnormal prediction module retrieves monitoring records with the same device combination identifier but different operation coordinates, extracts the temperature data under the corresponding environment, calculates the temperature feature difference between the current coordinate and the historical multiple operation coordinates, and judges whether to trigger the temperature abnormality warning by comprehensively considering the temperature time series deviation and the temperature position sequence deviation.

[0035] The technical effects and advantages of the present application are as follows:

[0036] The present application identifies the device combination participating in data storage in the data center, extracts the device storage and temperature information, screens the core devices and collects their temperatures, calculates the overall trend temperature of the device combination by combining the temperature, operation storage ratio and calling duration of each labeled device, reduces the difference between the overall trend temperature and the temperature of the core device, extracts the current period operation temperature feature by principal component analysis, and calculates the temperature time series deviation based on the sliding time window. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 A flow implementation diagram of a data center operation temperature monitoring method of the present application.

[0038] Figure 2 A module schematic diagram of a data center operation temperature monitoring system of the present application. DETAILED DESCRIPTION

[0039] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0040] The present application identifies the combination of devices participating in data storage in the data center, extracts device storage and temperature information, selects core devices and collects their temperatures, combines the temperatures of each labeled device, the running storage ratio and the calling duration, calculates the overall trend temperature of the device combination, extracts the running temperature characteristics of the current period using principal component analysis, calculates the temperature time sequence deviation based on a sliding time window, simultaneously obtains the ectopic data of the same device combination under different coordinates, constructs an ectopic temperature feature model, calculates the temperature position sequence deviation of the current coordinate and the mean value of the ectopic features, and finally fuses the temperature time sequence deviation and the temperature position sequence deviation to predict the probability of temperature anomaly of the target coordinate in the future time through Bayesian inference, judges whether to trigger an alarm, so as to realize dynamic modeling and anomaly prediction of the operation temperature of the data center, improve the intelligent level of temperature control, and ensure the safe and energy-saving operation of the system.

[0041] Embodiment 1, a data center operation temperature monitoring method, as shown in Figure 1 The method comprises the following steps:

[0042] Step S1: Obtain the combination of devices participating in data storage, count the data storage items of each device in the device combination, select the device with the largest data storage item for temperature detection, and record the temperature of the core device of the device combination;

[0043] Step S2: Classify and label each device in the device combination, detect the device temperature of each labeled device, collect the running storage ratio and data calling duration of each labeled device, set the label weight, and calculate the overall trend temperature of the device combination by comprehensively considering the device temperature and label weight of the labeled device;

[0044] Step S3: Analyze the operating temperature characteristics of the data center based on the core equipment temperature and overall trend temperature of the equipment combination, call up historical data, use the historical data to calculate the historical temperature characteristics of the data center, and combine the operating temperature characteristics to analyze the temperature time series deviation of the data center.

[0045] Step S4: Obtain out-of-location data of the same equipment combination, use the out-of-location data to calculate the out-of-location temperature characteristics of the data center, combine the operating temperature characteristics to analyze the temperature sequence deviation of the data center, and integrate and predict the operating temperature characteristics of the data center by combining the temperature time sequence deviation and the temperature sequence deviation.

[0046] The specific implementation is as follows:

[0047] In step S1, the device management platform connected to the data center identifies all devices involved in data storage as a device group, and reads data storage item information from each device. The data storage item information is the used storage space of each device.

[0048] The used storage space values ​​of each device are compared, and the device with the largest value is selected as the core device. The temperature data collected in real time by the internal temperature sensor of the core device is retrieved and recorded as the core device temperature C of the device combination. c .

[0049] It should be noted that the device management platform is a systematic software used to centrally manage and monitor various devices in the data center; the temperature sensor is a physical detection element integrated inside the device to sense the temperature of the device and obtain temperature values.

[0050] In step S2, each device in the device assembly is labeled as d. i (i = 1, 2, ..., n), retrieve the real-time temperature data collected by the internal temperature sensors of each device and record it as C. i Simultaneously, it reads the running storage space from each marked device and calculates the running storage percentage. The running storage percentage refers to the ratio of the device's running storage space to the device's total storage capacity, calculated using the following formula: In the formula, S i R represents the percentage of runtime storage. i For device operating storage space, D i This represents the total storage capacity of the device.

[0051] By merging the start and end timestamps of each data call, the total duration of actual data read and write operations on each marked device is accumulated as the data call duration. The standardized calculation of the data call duration is based on the proportion of data retrieval time within a fixed monitoring period T. The specific calculation formula is as follows: Among them, T i This is a standardized value for the data retrieval time. is the start time stamp, is the end time stamp, i represents the device number, j represents the number of event calls of the device in the period, m is the number of data calls of the device in the current monitoring period, and T is a fixed monitoring period.

[0052] After completing the standardization calculation of the data call duration of each marking device, it is input into the marking weight calculation process of each marking device together with the synchronously collected running storage proportion. The specific marking weight calculation formula is: i = α × S i + β × T i In the formula, w i is the marking weight of each marking device, S i is the running storage proportion of each marking device, T i is the proportion of the data call duration of each marking device in a unit time, α is the importance coefficient of the running storage proportion, β is the importance coefficient of the proportion of the data call duration in a unit time, and α and β can be adjusted according to actual conditions.

[0053] Based on the temperature data and marking weight of each marking device, the overall trend temperature of the device combination is calculated by using a weighted average model. The specific calculation formula is as follows: Wherein, C g is the overall trend temperature of the device combination, n is the number of marking devices participating in the calculation in the current device combination, C gi is the temperature data of the marking device, and w is the marking weight of the marking device.

[0054] It should be noted that the time stamp refers to the absolute time identification of the occurrence time of a specified event, which is used to accurately identify the time point of the occurrence of a specific event. The parameters of the device combination (including device identification, core device, temperature of each device, used storage space, call duration, and device running coordinates) are continuously collected during the running period and uniformly accessed to the historical data storage platform of the data center. When performing historical backtracking analysis of temperature characteristics, the unique identification of the device combination and the specified time window are used.

[0055] Through the above weighted calculation, the running heat load characteristics of each marking device are converged into a unified overall trend temperature index, which provides a quantifiable data basis for subsequent data center running temperature characteristic modeling, historical characteristic comparison analysis, and temperature abnormality early warning.

[0056] In step S3, the principal component analysis (PCA) method is used to comprehensively analyze the running temperature characteristics of the device combination based on the core device temperature and the overall trend temperature of the device combination, and a quantifiable running temperature characteristic index is obtained. The specific operation process is as follows:

[0057] First, let The core equipment temperature sequence collected by the current equipment combination in a time window, The two types of sequence data are formed into a two-dimensional matrix for the overall trend temperature sequence in the corresponding time window, denoted as: Wherein, is the nth time point data of the core equipment temperature, is the nth time point data of the overall trend temperature, and n is the sample number.

[0058] The input matrix is standardized to obtain a standardized sample matrix, and the calculation formula is Wherein, X ij is the jth temperature data of the ith time point (j = 1 represents the core equipment temperature of the equipment combination, and j = 2 represents the overall trend temperature of the equipment combination), μ j is the sample mean of the jth type, and the calculation formula is σ j is the standard deviation of the jth column sample, and the calculation formula is

[0059] Then the covariance matrix of the standardized sample matrix Z is calculated to obtain the joint variation relationship between the temperature dimensions, and the calculation formula is Wherein, Σ is the standardized covariance matrix, Z T is the transpose of the standardized matrix.

[0060] The eigenvalue decomposition of the covariance matrix Σ is carried out, and according to the relationship formula Σ × V k = λ k × V k (k = 1, 2), and and the corresponding eigenvalue λ k and the eigenvector V k are obtained, V1 is the eigenvector with the largest eigenvalue, representing the maximum variation direction (i.e. the principal component direction), and the first principal component vector V1 is selected as the main direction of the operating temperature feature.

[0061] For the standardized temperature sample of the current time point t Project it onto the principal component direction to obtain the operating temperature feature value: Wherein, C (t) is the operating temperature feature value of the current time point t, is the standardized core temperature at time point t, is the standardized overall trend temperature at time point t, V 1ct and V 1gt are the weight coefficients of the principal component vectors corresponding to the core temperature and the overall trend temperature, respectively.

[0062] A historical analysis time period is selected as a historical analysis time, historical data at multiple time points in the historical analysis time is called, and the historical data is core device temperature and overall trend temperature of a device combination at each time point in the historical analysis time; a plurality of historical temperature features are generated based on a principal component analysis method and are combined into a historical temperature feature sequence.

[0063] Based on the historical temperature feature value sequence generated by the principal component analysis, a sliding time window model is constructed to evaluate the dynamic deviation of different time points in temperature. Let {C (t-k) ,C (t-k+1) ,...,C (t-1)} be the temperature feature values of the first k time points in the sliding window with a length of k. The historical temperature feature values in the sliding window are calculated by mean value calculation, and the specific formula is Where a represents the a-th time point back from the current time point t.

[0064] Then, the sliding mean difference of the temperature feature value of the current time point relative to the window mean value is calculated, which is used to represent the temperature time series deviation, and the calculation formula is Where, ΔC (t) is the temperature time series deviation of the current time point t.

[0065] After completing the running temperature feature extraction and temperature time series deviation calculation, a spatial dimension temperature comparison mechanism is further introduced to realize the global steady-state analysis of the data center running temperature.

[0066] In step S4, let the unique identifier of the current analyzed device combination be G id , and the running coordinate be L0. The monitoring data records consistent with the device combination identifier but different in running coordinates (such as L1, L2,..., L m ) are retrieved from the historical data storage platform, and the running data of the current device combination under other running coordinates, i.e. “allopatric data”, are obtained. Based on the principal component analysis (PCA) method, the core device temperature and the overall trend temperature of the device combination under different running coordinates are constructed into a matrix Y’. Where, is the core device temperature of the device combination at position L m , is the overall trend temperature of the device combination at position L m , and m is the number of analyzed running coordinates.

[0067] The matrix Y’ is standardized to obtain a standardized sample matrix, and the calculation formula is Where, Z ij * is the standardized sample matrix, Y ijis the ith running coordinate jth temperature data of the equipment combination (j = 1 represents the core temperature of the equipment combination, j = 2 represents the overall trend temperature of the equipment combination), μ j * is the mean value of the jth sample, and the calculation formula is σ j * is the standard deviation of the jth column sample, and the calculation formula is

[0068] The standardized sample matrix Z * is subjected to covariance matrix calculation to obtain the joint variation relationship between the temperature dimensions, and the calculation formula is wherein, Σ is the standardized covariance matrix, Z *Τ is the transpose of the standardized matrix.

[0069] The eigenvalue decomposition is performed on the covariance matrix Σ, and according to the relationship formula Σ × V = × V, the eigenvector V corresponding to the maximum eigenvalue is taken as the main direction of the ectopic temperature change.

[0070] The standardized temperature sample of the current running coordinate is projected to the principal component direction to obtain the ectopic running temperature eigenvalue of the current running coordinate L0, and the specific calculation formula is wherein is the ectopic running temperature eigenvalue, is the standardized core temperature of the running coordinate L0, is the standardized overall trend temperature of the running coordinate L0, V cp and V gp are the weight coefficients corresponding to the core temperature and the overall trend temperature in the principal component vector, respectively.

[0071] Based on the ectopic temperature eigenvalue sequence generated by the principal component analysis, let be the average value of the temperature eigenvalue of the equipment combination under all running coordinates participating in the PCA modeling, that is, the ectopic temperature eigenvalue mean, and the specific calculation formula is wherein is the ectopic temperature eigenvalue mean, and b represents the bth running coordinate backtracking from the current running coordinate L0.

[0072] The sliding average deviation of the temperature eigenvalue of the current running coordinate L0 relative to the ectopic temperature eigenvalue mean is calculated, which is used to represent the temperature rank deviation, and the calculation formula is wherein, is the temperature rank deviation of the current running coordinate L0.

[0073] Based on the temperature time series deviation ΔC (t) and the temperature rank deviation the calculation results, two types of deviation data are taken as input characteristics, a Bayesian binary classification inference model is constructed to estimate the temperature anomaly probability of the future time t f .

[0074] Let the temperature anomaly event be denoted as a binary random variable Y∈{0,1}, where Y=1 represents that the temperature anomaly occurs at time t f , and Y=0 represents no anomaly. For any deviation data pair , the corresponding temperature anomaly probability value is: , where is the deviation data pair , P(Y=1) is the prior probability of the anomaly event in the training sample, is the conditional probability of the combination of the deviation data pair occurring given that the anomaly occurs, is the marginal probability of the deviation combination in the overall sample.

[0075] Under the assumption of feature conditional independence, the joint conditional probability is decomposed as: Further, a fitting model based on Gaussian distribution is established for the temperature time series deviation and the bit sequence deviation, respectively, and is expressed as follows: where y∈{0,1} represents the non-anomaly and anomaly situations respectively, represents the mean and standard deviation of in the sample with label y, is the mean and standard deviation of . For the current monitoring point , the prediction probability of the temperature anomaly is calculated as follows:

[0076] A series of candidate threshold sets {θ j}, j=1,2,…,M, are set. For each threshold, the performance indicator under the model prediction is calculated, and the harmonic mean F1 score of the precision and recall is calculated. The accuracy rate refers to the proportion of actual anomalies in the samples predicted as anomalies, which is marked as S; the recall rate refers to the proportion of actual anomalies successfully predicted, which is marked as R; the calculation formula of F1 is The threshold that maximizes F1(θ j ) is taken as the optimal decision boundary of the system, denoted as θ.

[0077] According to the probability threshold θ, if , no temperature anomaly warning is performed, indicating that the risk of temperature anomaly at the running coordinate L0 and time t f is low; if , temperature anomaly warning is performed, indicating that the risk of temperature anomaly at the running coordinate L0 and time t fThe risk of temperature anomaly is low.

[0078] Embodiment 2, a data center operating temperature monitoring system, as shown in Figure 2 As shown, a data center operating temperature monitoring method is implemented, including the following modules: device identification module, label modeling module, temperature feature extraction module and anomaly prediction module, and the functions of each module are as follows:

[0079] Device identification module: used to identify all devices participating in data storage, count the used storage space of each device in the combination, select the core device and obtain the temperature reading of the internal temperature sensor of the core device;

[0080] Label modeling module: classifies and labels non-core devices in the device combination, respectively collects the real-time temperature, running storage ratio and data calling duration of each device in a unit period, and calculates the overall trend temperature of the device combination;

[0081] Temperature feature extraction module: takes the temperature of the core device and the overall trend temperature as input to form the current operating temperature feature value, and calls the temperature data in the historical time window to calculate the sliding average difference between the current operating temperature feature and the historical average value, thereby forming the temperature time sequence deviation index of the data center;

[0082] Anomaly prediction module: retrieves monitoring records with the same device combination identifier but different operating coordinates, extracts the temperature data under the corresponding environment, calculates the temperature feature difference between the current coordinate and the historical multiple operating coordinates, and judges whether to trigger the temperature anomaly warning by comprehensively considering the temperature time sequence deviation and the temperature position sequence deviation.

[0083] The above formulas are all dimensionless values, and the formulas are obtained by software simulation of a large amount of data to obtain a formula closest to the actual situation. The preset parameters in the formula are set by the person skilled in the art according to the actual situation.

[0084] The above-described embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented in software, the above-described embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are wholly or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center through a wired or wireless (such as infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. containing one or more available medium collections. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state disk.

[0085] It should be understood that the size of the sequence number of each process described above in various embodiments of the present application does not mean the order of execution, and the execution order of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0086] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be realized in electronic hardware, or a combination of computer software and electronic hardware. Whether the functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0087] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0088] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other manners. For example, the described device embodiments are merely schematic. The division of the units is merely logical function division. There can be other division manners in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0089] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0090] In addition, each functional unit in the various embodiments of the present application can be integrated into a processing unit, or each unit can be a physically independent unit, or two or more units can be integrated into one unit.

[0091] If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application can be embodied in the form of a software product, and the computer software product is stored in a storage medium, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program code storage media.

[0092] The above is merely specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for monitoring operating temperature in a data center, characterized in that: Includes the following steps: Step S1: Obtain the device combination participating in data storage, count the data storage items of each device in the device combination, filter out the device with the largest data storage item for temperature detection, and record the temperature of the core device in the device combination; Step S2: Classify and label each device in the device group, detect the device temperature of each labeled device, collect the operating storage ratio and data access time of each labeled device to set the label weight, and calculate the overall trend temperature of the device group by combining the device temperature of the labeled devices and the label weight; Step S3: Analyze the operating temperature characteristics of the data center based on the core equipment temperature and overall trend temperature of the equipment combination, call up historical data, use the historical data to calculate the historical temperature characteristics of the data center, and combine the operating temperature characteristics to analyze the temperature time series deviation of the data center. Step S4: Obtain out-of-location data of the same equipment combination, use the out-of-location data to calculate the out-of-location temperature characteristics of the data center, combine the operating temperature characteristics to analyze the temperature sequence deviation of the data center, and integrate and predict the operating temperature characteristics of the data center by combining the temperature time sequence deviation and the temperature sequence deviation.

2. The data center operating temperature monitoring method according to claim 1, characterized in that: In step S1, all devices involved in data storage are identified as a device group, and data storage item information is read from each device. The data storage item information is the used storage space of each device. The device with the largest value is selected as the core device. Retrieve real-time temperature data collected by the internal temperature sensors of the core equipment and record it as the core equipment temperature C of the equipment assembly. c .

3. The data center operating temperature monitoring method according to claim 1, characterized in that: In step S2, the running storage ratio refers to the ratio of the device's running storage space to the device's total storage capacity; The start and end timestamps of data access in each tagging device are merged, and the total duration of actual data read and write operations in each tagging device is used as the data access duration.

4. The data center operating temperature monitoring method according to claim 3, characterized in that: In step S2, the data retrieval duration is standardized to generate data retrieval coefficients. Based on these coefficients and the percentage of memory used for operation, the corresponding labeling weights for the labeled devices are set: w i =α×S i +β×T i , where w i S represents the labeling weight for each labeling device. i T represents the percentage of runtime storage for each tagging device. i The percentage of data access time for each marked device within a unit of time is denoted by α, which is the importance coefficient of the percentage of runtime storage, and β is the importance coefficient of the percentage of data access time within a unit of time. The overall trend temperature of the equipment combination is obtained by weighting the real-time temperature of each marking device with the marking weight.

5. The data center operating temperature monitoring method according to claim 4, characterized in that: In step S3, principal component analysis is used to analyze the operating temperature characteristics of the data center based on the stability of the core equipment and the overall trend temperature of the equipment combination. The specific steps are as follows: Data preparation: Set the core equipment temperature sequence and overall trend temperature sequence of the current equipment combination within the time window, form a two-dimensional matrix, and generate a standardized sample matrix after standardizing the two-dimensional matrix; Covariance matrix calculation: Calculate the covariance matrix of the standardized sample matrix; Eigenvalue decomposition: Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors, and select the eigenvector with the largest eigenvalue as the principal component direction; Eigenvalue projection: Projecting the standardized temperature sample at the current time point t onto the principal component direction to obtain the operating temperature eigenvalues.

6. The data center operating temperature monitoring method according to claim 5, characterized in that: In step S3, a historical period is selected as the historical analysis time. Multiple time points are set within the historical analysis time to call historical data. The historical data consists of the core equipment temperature and overall trend temperature of the equipment combination at each time point within the historical analysis time. Multiple historical temperature features are generated based on the principal component analysis method and merged into a historical temperature feature sequence.

7. A method for monitoring the operating temperature of a data center according to claim 6, characterized in that: In step S3, a sliding window of length k is set, and the historical temperature characteristic mean value of each time point is calculated within the sliding window. The absolute value of the difference between the current time point temperature characteristic value and the historical temperature characteristic mean value is taken as the temperature time series deviation of the current time point t.

8. A method for monitoring the operating temperature of a data center according to claim 7, characterized in that: In step S4, the out-of-location data is the operating data of the current equipment combination under other operating coordinates. The out-of-location temperature characteristics of the data center are calculated using the out-of-location data. The core equipment temperature and the overall trend temperature under different operating coordinates are combined into a matrix and standardized. The standardized results are used to generate multiple out-of-location temperature characteristics using the principal component analysis method. The moving average difference between the current running coordinate L0 temperature characteristic value and the mean of the out-of-place temperature characteristic value is calculated as the temperature positional deviation.

9. A method for monitoring the operating temperature of a data center according to claim 8, characterized in that: In step S4, the temperature time series deviation and temperature positional deviation are used as input features to construct a Bayesian binary classification inference model to integrate and predict the operating temperature characteristics of the data center. The specific steps are as follows: Define abnormal events: Set an abnormal temperature value Y. When the abnormal temperature value Y = 1, the temperature is determined to be abnormal; when the abnormal temperature value Y = 0, the temperature is determined to be normal. Calculate the probability: Define the temperature time sequence deviation and temperature positional deviation as ΔC respectively. (tf) and ΔC (L0) Combine them into a deviation data pair (ΔC) (tf) ,ΔC (L0) ), calculate the probability value of temperature anomalies based on the deviation data; Anomaly prediction: If the calculated probability value of temperature anomaly based on the deviation data is lower than the preset optimal judgment boundary, no temperature anomaly warning will be issued. If the calculated probability value of temperature anomaly from the deviation data is higher than the preset optimal judgment boundary, a temperature anomaly warning will be issued.

10. A data center operating temperature monitoring system, based on the data center operating temperature monitoring method according to any one of claims 1-9, characterized in that: It includes a device identification module, a tag modeling module, a temperature feature extraction module, and an anomaly prediction module; Device identification module: used to identify all devices involved in data storage, count the used storage space of each device in the combination, filter out the core devices and obtain the temperature readings of their internal temperature sensors; The labeling and modeling module classifies and labels non-core devices within the device portfolio, collects real-time temperature, operating storage ratio, and data retrieval time per unit period for each device, and calculates the overall trend temperature of the device portfolio. Temperature feature extraction module: Taking the core equipment temperature and the overall trend temperature as input, it forms the current operating temperature feature value, and calls the temperature data within the historical time window to calculate the moving average difference between the current operating temperature feature and the historical average value, thereby forming the temperature time series deviation index of the data center. Anomaly prediction module: Retrieves monitoring records with the same equipment combination identifier but different operating coordinates, extracts temperature data under the corresponding environment, calculates the temperature characteristic differences between the current coordinate and multiple historical operating coordinates, and determines whether to trigger a temperature anomaly warning by combining temperature time sequence deviation and temperature position sequence deviation.

Images