IDC equipment intelligent detection and classification management method and system

Abstract

The application relates to the technical field of electrical equipment testing, in particular to an IDC equipment intelligent detection and classification management method and system, which comprises the following steps: collecting the current, voltage, power consumption and temperature of a server in an IDC cabinet, synchronously recording the CPU occupancy rate, GPU occupancy rate, storage I / O load and network throughput rate, extracting server log entries and storing the server log entries into an equipment exception log library, and obtaining IDC equipment running state data.In the application, the current, voltage, power consumption and temperature of the IDC equipment are synchronously monitored, and server running data is collected, a new scheme optimizes the comprehensiveness of data and the real-time performance of monitoring, the monitoring strategy provides more fine risk assessment and classification management by dynamically analyzing the state fluctuation of the equipment, the accuracy of fault prediction and the efficiency of maintenance are improved, the prediction capability for potential problems is enhanced by comprehensively utilizing the analysis of abnormal logs and running data, and the system downtime rate and maintenance cost are effectively reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electrical equipment testing, and in particular to an IDC equipment intelligent detection and classification management method and system. BACKGROUND

[0002] The technical field of electrical equipment testing includes technical means for measuring and evaluating the performance, electrical parameters, and fault conditions of various electrical equipment. The core content of this technical field includes the detection of current, voltage, resistance, capacitance, inductance, and power quality parameters, and covers signal acquisition, data analysis, and diagnostic methods. The overall technical system mainly involves the application of measuring instruments and sensors, data acquisition and transmission methods, electrical fault diagnosis methods, and related intelligent analysis means. Electrical equipment testing technology is widely used in power systems, industrial automation, communication equipment, and computer hardware fields to ensure the stability, safety, and reliability of equipment.

[0003] Among them, the IDC equipment intelligent detection and classification management method refers to the technical method for monitoring the state of servers, storage devices, and network devices in IDC machine rooms through a preset detection process and classification rules to complete the automatic classification management of equipment state. This method covers real-time monitoring of equipment operating state, identification of abnormal data, setting of equipment classification rules, and storage management of classification results. Specifically, it uses detection methods based on current, voltage, and power consumption changes to collect equipment operating data, combines feature parameter matching methods to complete equipment classification, and uses database systems to manage classification results.

[0004] The monitoring and diagnosis in the prior art focus on a single parameter, lack real-time analysis of multiple data dimensions, and limit the timeliness and effectiveness of fault prevention and equipment maintenance. This technical limitation leads to incomplete equipment state evaluation, inability to achieve early risk identification and classification, and increases the instability and maintenance difficulty of the system. The lack of effective data integration and intelligent analysis results in low efficiency of maintenance strategies and resource allocation, affecting the cost-effectiveness of overall equipment management. SUMMARY

[0005] The purpose of the present application is to solve the shortcomings in the prior art and to provide an IDC equipment intelligent detection and classification management method and system.

[0006] To achieve the above purpose, the present application adopts the following technical solution: an IDC equipment intelligent detection and classification management method, comprising the following steps:

[0007] S1: Collecting the current, voltage, power consumption, and temperature of servers in an IDC cabinet, synchronously recording the CPU occupancy rate, GPU occupancy rate, storage I / O load, and network throughput rate, extracting server log entries and storing them in a device exception log library to obtain IDC equipment operating state data;

[0008] S2: call the IDC equipment running state data, analyze server load change rate, calculate temperature, current, voltage fluctuation degree, count abnormal log occurrence frequency, obtain IDC equipment state fluctuation degree;

[0009] S3: based on the IDC equipment state fluctuation degree, calculate running stability score, analyze load fluctuation, electrical parameter, abnormal log, compare stability and health score, classify risk and stable running equipment, generate IDC equipment running classification label;

[0010] S4: according to the IDC equipment running classification label, analyze temperature, current, load, abnormal log trend, predict load peak value change, establish IDC equipment abnormal trend prediction value;

[0011] S5: for the IDC equipment abnormal trend prediction value, calculate maintenance priority, adjust equipment maintenance order, generate IDC equipment maintenance priority order.

[0012] As a further scheme of the application, the IDC equipment running state data is specifically server current, server voltage, network throughput rate change rate, device abnormal log library storage content, the IDC equipment state fluctuation degree includes server load change rate, current difference value, voltage difference value, temperature difference value, abnormal log weight, the IDC equipment running classification label specifically refers to high-risk IDC equipment, stable running IDC equipment, the IDC equipment abnormal trend prediction value specifically is server temperature trend, current trend, load peak value interval change, abnormal log quantity trend, the IDC equipment maintenance priority order includes server health score, abnormal early warning level, equipment business importance score.

[0013] As a further scheme of the application, the step of obtaining the IDC equipment running state data is specifically:

[0014] S101: collect the current, voltage, power consumption and temperature data of the servers in the IDC cabinet, obtain the CPU occupancy rate, GPU occupancy rate, storage I / O load and network throughput rate, record the server log entries, and store them to the device abnormal log library, and generate the device running parameter data set;

[0015] S102: based on the device running parameter data set, normalize the server current, voltage and power consumption data, calculate the power consumption mean value and fluctuation amplitude in multiple time windows, and calculate the heat dissipation state change amount combined with the server temperature data, using the formula:

[0016]

[0017] Calculate server temperature fluctuation values ​​and generate server heat dissipation status parameters;

[0018] Among them, S temp T represents the server temperature fluctuation value. i T represents the server temperature in the i-th time window. avg Represents the average temperature, n represents the total number of time windows, and P max P represents the maximum power consumption value. min V represents the minimum power consumption value. avg Represents the average voltage;

[0019] S103: Based on the server heat dissipation status parameters, call the CPU utilization rate, GPU utilization rate, storage I / O load and network throughput data, calculate the changing trend of device power consumption and heat dissipation parameters under different load conditions, extract server abnormal state characteristics, and generate IDC device operating status data.

[0020] As a further aspect of the present invention, the step of obtaining the state fluctuation of the IDC device specifically includes:

[0021] S201: Call the IDC device operation status data, calculate the server load change rate within the differentiated time window, calculate the change rate of CPU utilization, GPU utilization and storage I / O load based on adjacent time periods, and obtain the average and fluctuation range to obtain the server load change rate.

[0022] S202: Calculate the fluctuation values ​​of temperature, current, and voltage based on the server load change rate, monitor changes in adjacent data, and use the following formula:

[0023]

[0024] Calculate the degree of fluctuation in the operating parameters of the equipment;

[0025] Among them, Y dev T represents the degree of fluctuation in the equipment's operating parameters. j T represents the temperature data for time period j. avg I represents the average temperature. j I represents the current data for time period j. avg Represents the average current, V j V represents the voltage data for time period j. avg represents the average voltage, and m represents the number of time windows;

[0026] S203: Based on the fluctuation of the device operating parameters, count the abnormal log entries in the device abnormal log library, calculate the number of times abnormal logs occur within the differentiated time window, and combine the fluctuation of temperature, current and voltage to summarize the fluctuation trend of the differentiated time period and obtain the IDC device status fluctuation.

[0027] As a further scheme of the present application, the obtaining step of the IDC equipment running classification label is specifically:

[0028] S301: Based on the IDC equipment state fluctuation degree, the stability score of multi-equipment running is calculated, the change of load fluctuation, electrical parameter and abnormal log in the difference time window is counted, and the stability weight of multiple indexes is calculated to obtain the equipment stability score;

[0029] S302: Based on the equipment stability score, the load fluctuation, electrical parameter and abnormal log frequency of multi-equipment are analyzed, the mean value and dispersion degree of stability score are calculated, and the formula is:

[0030]

[0031] The equipment stability deviation value is calculated to obtain the stability and health score;

[0032] Among them, S stab represents the equipment stability deviation value, W k represents the stability weight of the kth equipment, L k represents the load fluctuation value of the kth equipment, represents the load fluctuation mean value, E k represents the abnormal log frequency of the kth equipment, Q k represents the electrical parameter value of the kth equipment, represents the electrical parameter mean value, p represents the total number of equipment, I dev represents the current fluctuation degree;

[0033] S303: Based on the stability and health score, the equipment is divided into high-risk equipment, low-risk equipment and stable running equipment according to the score range, and its state category is marked to obtain the IDC equipment running classification label.

[0034] As a further scheme of the present application, the obtaining step of the IDC equipment abnormal trend prediction value is specifically:

[0035] S401: According to the IDC equipment running classification label, the temperature, current, load and abnormal log trend in multiple time windows are analyzed, the state change rate of multiple equipment is calculated, and the equipment state trend parameter is obtained;

[0036] S402: Based on the equipment state trend parameter, the load peak change of multiple equipment is analyzed, the load peak growth rate and fluctuation range of difference time period are calculated, and the formula is:

[0037]

[0038] Calculate the load peak variation parameter, integrate the load peak variation trend;

[0039] Wherein, P peak represents the load peak variation parameter, T s represents the temperature data of the s period, T avg represents the temperature mean value, I s represents the current data of the s period, L' s represents the load data of the s period, represents the load data mean value, r represents the total number of time windows, E α represents the frequency of abnormal log occurrence;

[0040] S403: Based on the load peak variation trend, the abnormal device state change mode is induced, the abnormal change amplitude of the future time window is calculated, and the IDC device abnormal trend prediction value is obtained.

[0041] As a further scheme of the application, the obtaining step of the IDC device maintenance priority ranking is specifically:

[0042] S501: Based on the IDC device abnormal trend prediction value, the maintenance priority of multiple devices is calculated, the device maintenance priority parameter is obtained according to the abnormal trend change rate and the operation failure rate of the device;

[0043] S502: Based on the device maintenance priority parameter, the emergency degree of multiple device maintenance is analyzed, the maintenance demand index is calculated, and the formula is adopted:

[0044]

[0045] The maintenance demand index is calculated, and the device maintenance order adjustment value is obtained;

[0046] Wherein, M req represents the maintenance demand index, U q represents the abnormal trend change rate of the qth device, R q represents the mean value of all device abnormal trends, W q represents the cumulative failure influence factor of the qth device, X q represents the load stability index of the qth device, p represents the total number of devices, Y δ represents the device state fluctuation parameter;

[0047] S503: Based on the device maintenance order adjustment value, the device maintenance order is optimized and adjusted, and the IDC device maintenance priority ranking is obtained according to the maintenance priority ranking.

[0048] An IDC equipment intelligent detection and classification management system for performing the IDC equipment intelligent detection and classification management method, the system comprising:

[0049] The device running state monitoring module collects server current, voltage, power consumption, temperature data, obtains CPU / GPU occupation rate, storage I / O load, network throughput rate, and stores them to a data monitoring system; error codes and key parameters are extracted from detection logs, and the characters are filtered and stored to an abnormal log library; the adjacent change rates of current, voltage, and network throughput rate are calculated, mutation points are marked and stored, and device running state data is obtained;

[0050] The load fluctuation and abnormality analysis module calculates the server load change rate based on the device running state data, normalizes the CPU / GPU load, storage I / O load, and network throughput, calculates the first-order difference value of the server temperature, current, and voltage, extracts abnormal log categories, counts the number of log occurrences, calculates the abnormal weight, and obtains the IDC equipment state fluctuation degree;

[0051] The server running stability evaluation module calls the IDC equipment state fluctuation degree, calculates the server running stability score, calculates the current state score based on the load fluctuation, electrical parameter change, and abnormal log weight, calculates the health state score in combination with the number of power failures, current and voltage stability, and fault frequency, classifies or stably runs the equipment, and obtains the IDC equipment running classification label;

[0052] The IDC equipment abnormal trend prediction module extracts the temperature, current, load, and abnormal log quantity of the past target period based on the IDC equipment running classification label, calculates the time series trend, judges whether the trend curve is rising, calculates the load peak interval change, and if the interval is shortened, the warning level is increased, and the IDC equipment abnormal trend prediction value is obtained;

[0053] The equipment maintenance priority sorting module calls the IDC equipment abnormal trend prediction value, calculates the maintenance priority, generates a maintenance score in combination with the health score, abnormal warning level, and equipment business criticality score, arranges the maintenance plan according to the score, and establishes the IDC equipment maintenance priority sorting.

[0054] Compared with the prior art, the advantages and positive effects of the present application are that:

[0055] In the present application, by synchronously monitoring the current, voltage, power consumption and temperature of the IDC equipment and collecting server running data, the new scheme optimizes the comprehensiveness of data and the real-time of monitoring. This monitoring strategy provides more detailed risk assessment and classification management by dynamically analyzing the state fluctuation of the equipment, thereby improving the accuracy of fault prediction and the efficiency of maintenance. The analysis of abnormal logs and running data enhances the prediction ability of potential problems, effectively reducing the system downtime rate and maintenance costs. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 is a workflow schematic diagram of the present application;

[0057] Figure 2 is a flowchart of the step of obtaining the running state data of the IDC equipment of the present application;

[0058] Figure 3 is a flowchart of the step of obtaining the state fluctuation degree of the IDC equipment of the present application;

[0059] Figure 4 is a flowchart of the step of obtaining the running classification label of the IDC equipment of the present application;

[0060] Figure 5 is a flowchart of the step of obtaining the abnormal trend prediction value of the IDC equipment of the present application;

[0061] Figure 6 is a flowchart of the step of obtaining the maintenance priority ranking of the IDC equipment of the present application. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and do not limit the present application.

[0063] In the description of the present application, it should be understood that the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation of the present application. In addition, in the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise specifically limited.

[0064] Example one

[0065] Please refer to Figure 1The application provides a technical scheme: an IDC equipment intelligent detection and classification management method, comprising the following steps:

[0066] S1: collect electrical parameters such as server current, voltage, power consumption and temperature in the IDC cabinet, record the collected values at a set sampling interval, collect device running state parameters such as CPU occupancy rate, GPU occupancy rate, storage I / O load and network throughput rate, store them into an IDC data monitoring system, detect server logs, extract log entry quantity, error code and key parameters, perform character filtering, store them into a device exception log library, calculate the change rate of adjacent sampling points of the server current, voltage and network throughput rate, mark the mutation point and store it; obtain IDC equipment running state data;

[0067] S2: based on the IDC equipment running state data, calculate the server load change rate, normalize the CPU / GPU load, storage I / O load and network throughput, judge whether the rate exceeds a set threshold, calculate the first-order difference value of the server temperature, current and voltage in the IDC cabinet, analyze the change trend of adjacent time points, calculate the mean value, judge the fluctuation degree, extract the abnormal log category, count the number of occurrences of the category of logs in the past N periods, and calculate the abnormal weight; obtain the IDC equipment state fluctuation degree;

[0068] S3: call the IDC equipment state fluctuation degree, calculate the server running stability score, use the load fluctuation degree, electrical parameter change rate and abnormal log weight to calculate the current state score, calculate the server health state score, calculate the score according to the number of server power failures, current and voltage stability and historical fault times, and normalize the processing, compare the running stability score and the health state score in the current period, if the running stability score is lower than the set threshold and the health state score is lower, then classify it as a high-risk IDC equipment, if the load change rate is normal, the electrical parameters are not over limited, and the health state score is higher, then classify it as a stable running IDC equipment; generate an IDC equipment running classification label;

[0069] S4: based on the IDC equipment running classification label, extract the server temperature, current, load and abnormal log quantity in the past N collection periods, calculate the time series trend, judge whether the trend curve is rising, calculate the server load peak interval change, if the peak interval gradually shortens, then increase the abnormal warning level; obtain the IDC equipment abnormal trend prediction value;

[0070] S5: call the IDC equipment abnormal trend prediction value, calculate the server maintenance priority, generate a comprehensive maintenance score by using the server health score, abnormal warning level and device business importance score, arrange the maintenance plan according to the score; establish an IDC equipment maintenance priority order.

[0071] The IDC equipment running state data specifically refers to server current, server voltage, network throughput rate change rate, and device exception log library storage content. The IDC equipment state fluctuation degree includes server load change rate, current difference value, voltage difference value, temperature difference value, and abnormal log weight. The IDC equipment running classification label specifically refers to high-risk IDC equipment and stable running IDC equipment. The IDC equipment abnormal trend prediction value specifically refers to server temperature trend, current trend, load peak interval change condition, and abnormal log quantity trend. The IDC equipment maintenance priority ranking includes server health score, abnormal early warning level, and device business importance score.

[0072] Please refer to Figure 2 The acquisition step of the IDC equipment running state data specifically includes:

[0073] S101: Collecting current, voltage, power consumption and temperature data of the servers in the IDC cabinet, acquiring CPU occupancy rate, GPU occupancy rate, storage I / O load and network throughput rate, recording server log entries, and storing them to the device exception log library to generate device running parameter data set;

[0074] The current, voltage, power consumption and temperature data of the servers in the IDC cabinet are collected through high-precision current sensors, voltage sensors installed on the server power supply circuit and temperature sensors in the cabinet. The current sensor uses a Hall effect sensor with a sampling frequency set to 100 Hz to ensure the capture of high-frequency current fluctuations. The voltage sensor uses a voltage division measurement circuit and an A / D conversion module to convert the analog voltage signal to a digital signal for storage. The power consumption data is obtained by the real-time multiplication of the current and voltage signals and stored in the database. The temperature data is obtained by multiple temperature probes in the cabinet and uses a moving average filtering algorithm to eliminate transient fluctuations. In addition, the data of CPU utilization, GPU utilization, storage I / O load and network throughput rate are obtained by relying on the monitoring module of the server operating system. CPU utilization is obtained by calling the task manager interface, recorded once per second, and the average utilization rate within one minute is calculated. GPU utilization is obtained by NVIDIAsMI or AMD Radeon Metrics tools regularly and stored in JSON format. Storage I / O load is obtained by Iostat tools or built-in system monitoring, recorded once every 10 seconds and stored in the database. Network throughput rate is obtained by the network card traffic statistics module, including uplink and downlink traffic, recorded in Mbps. Server log entries are obtained from system logs (Syslog) or Windows event logs (EventViewer), including error codes, warning information and operation records. Abnormal logs are stored in the device exception log library according to the timestamp and retrieved using indexing to facilitate subsequent calls. Finally, after data cleaning, these data form a complete device running parameter data set, which contains server status information at each time, as shown in Table 1.

[0075] Table 1: Example of device running parameter data set

[0076]

[0077] As shown in Table 1, the data set contains time stamp, current, voltage, power consumption, temperature and other key parameters, which can be used for further analysis of server status.

[0078] S102: Based on the device running parameter data set, the server current, voltage and power consumption data are normalized, the power consumption average and fluctuation amplitude in multiple time windows are calculated, and the heat dissipation state change is calculated combined with the server temperature data, using the formula:

[0079]

[0080] Calculate the server temperature fluctuation value and generate the server heat dissipation state parameter;

[0081] where Stemp representing the server temperature fluctuation value, T i representing the server temperature of the i-th time window, T avg representing the temperature mean value, n representing the total number of time windows, P max representing the maximum power consumption value, P min representing the minimum power consumption value, V avg representing the voltage mean value;

[0082] Based on the device operating parameter data set, first, the current, voltage and power consumption data of the server are normalized. The normalization method adopts the minimum-maximum normalization, and the calculation method is as follows:

[0083]

[0084] wherein X is the original data, X min and X max respectively represent the minimum value and the maximum value of the parameter in the data set. For example, assuming that the power consumption data range is 2000W to 3000W, for a power consumption value of 2415W at a certain time, the normalized value is:

[0085]

[0086] Next, the power consumption mean value and fluctuation amplitude are calculated in multiple time windows. The time window is set to 1 minute, 5 minutes and 10 minutes, and the power consumption mean values and standard deviations σ P are calculated respectively. Assuming that the power consumption data in the 1-minute window is {2400, 2450, 2500, 2420, 2415}W, then:

[0087]

[0088] Then, the heat dissipation state change amount S temp is calculated in combination with the server temperature data:

[0089]

[0090] Assuming that the temperature data in 5 minutes is {35, 36, 37, 36, 35}℃, the maximum power consumption P max = 2500W, the minimum power consumption P min = 2400W, and the voltage mean value V avg = 230V, then:

[0091]

[0092] This value represents the server temperature fluctuation value S temp is low, and in this case, the server heat dissipation state parameter is in the normal range.

[0093] S103: Based on the server heat dissipation state parameter, the CPU occupancy rate, GPU occupancy rate, storage I / O load and network throughput rate data are called to calculate the change trend of device power consumption and heat dissipation parameter under different load states, extract the server abnormal state characteristics, and generate IDC device running state data.

[0094] Based on the calculated server heat dissipation state parameter S temp , the change trend of power consumption and heat dissipation parameter under different load states is further analyzed. The specific method is to first filter out the CPU occupancy rate, GPU occupancy rate, storage I / O load and network throughput rate data in different time periods, and calculate the correlation between them and power consumption and heat dissipation state. For example, in a 5-minute time window, the correlation coefficient of each load parameter and power consumption is calculated:

[0095]

[0096] Suppose the calculated r P,CPU = 0.85, which indicates that there is a strong positive correlation between CPU load and power consumption. Through data analysis under different load states, the server abnormal state characteristics can be extracted. For example, if the power consumption is still higher than 2500W under low load (CPU occupancy rate < 20%), it indicates that the server may have hardware failure or resource leakage problem. Finally, all data is summarized to generate IDC device running state data for abnormal detection and intelligent control.

[0097] Please refer to Figure 3 , the steps to obtain the IDC device state fluctuation degree are as follows:

[0098] S201: Call IDC device running state data, calculate server load change rate in different time windows, calculate rate change based on CPU occupancy rate, GPU occupancy rate and storage I / O load in adjacent time periods, and take the mean and fluctuation amplitude to obtain the server load change rate;

[0099] Call IDC device running state data, first extract CPU occupancy rate, GPU occupancy rate and storage I / O load data at different times from the database, and divide them according to the set time window. The time window can be set to 1 minute, 5 minutes and 10 minutes to analyze the load change under different time granularity. Calculate the load change rate in adjacent time windows, that is, between time t and t+1, calculate the CPU load change rate R CPU and GPU load change rate R GPU , which are calculated as follows:

[0100]

[0101] Assuming that the CPU occupancy rate data is {30, 35, 40, 38, 42}%, and the GPU occupancy rate data is {20, 22, 25, 23, 26}%, within a 5-minute time window, then:

[0102]

[0103] After calculating the complete data set according to the above method, the change rate in all time windows is obtained, and the average value is calculated and the standard deviation σ R The load fluctuation amplitude is calculated, and if the standard deviation σ R exceeds the set threshold (such as 10%), it indicates that the server load change rate is large. Through the above process, the server load change rate data can be obtained, as shown in Table 1.

[0104] Table 1 Server load change rate data

[0105]

[0106] As shown in Table 1, the data is used for subsequent calculation of device state fluctuation.

[0107] S202: Calculate the fluctuation value of temperature, current, and voltage based on the server load change rate, monitor adjacent data changes, and use the formula:

[0108]

[0109] Calculate the device operating parameter fluctuation degree;

[0110] Where Y dev represents the device operating parameter fluctuation degree, T j represents the temperature data in the jth period, T avg represents the temperature mean value, I j represents the current data in the jth period, I avg represents the current mean value, V j represents the voltage data in the jth period, V avg represents the voltage mean value, and m represents the number of time windows.

[0111] Based on the server load change rate data, the temperature, current, and voltage data in the same time window are extracted, and the fluctuation value of each parameter in the adjacent time window is calculated. First, the mean value in each time window is calculated:

[0112]

[0113] Then, the device operating parameter fluctuation degree Y dev is calculated:

[0114]

[0115] Assuming the temperature data in 5 minutes is {35, 36, 37, 36, 35}℃, the current data is

[0116] {10.5, 10.8, 10.2, 10.6, 10.4}A, and the voltage data is {229, 230, 231, 229, 230}V, then:

[0117]

[0118] Calculate the fluctuations of temperature, current, and voltage:

[0119]

[0120] This value represents the degree of fluctuation of the device operating parameters.

[0121] S203: Based on the degree of fluctuation of the device operating parameters, count the abnormal log entries in the device abnormal log library, calculate the number of abnormal logs in the differentiated time window, and combine the fluctuations of temperature, current, and voltage to summarize the fluctuation trend of the differentiated time period, and obtain the IDC device state fluctuation degree.

[0122] Based on the calculated Y dev , count the abnormal log entries in the abnormal log library within the corresponding time window, and calculate the number of occurrences, and filter out the number of logs in the same time window, as shown in Table 2.

[0123] Table 2 Abnormal log statistics

[0124]

[0125] Combine the degree of device state fluctuation to determine whether the number of abnormal logs exceeds the set threshold (e.g., if the number of abnormal logs exceeds 5 times within a 5-minute window, it is determined to be an abnormal state), if it exceeds the set threshold, it indicates that the IDC device state fluctuation is large, further summarize the fluctuation trend of the differentiated time period, assuming that the statistical result shows that the fluctuation degree increases significantly under high temperature state (T avg >40℃), it can be judged that the high temperature environment has an impact on the stability of the device state, and finally the IDC device state fluctuation degree data is obtained.

[0126] Please refer to Figure 4 , the steps for obtaining the IDC device operation classification label are as follows:

[0127] S301: Based on the IDC device state fluctuation degree, calculate the stability score of multiple device operation, count the changes of load fluctuation, electrical parameters and abnormal logs within the differentiated time window, and calculate the stability weight of multiple indicators, and obtain the device stability score;

[0128] Based on the IDC equipment state fluctuation degree, first extract the running state data of multiple devices in different time windows, including load fluctuation, electrical parameters (current, voltage, power consumption) and abnormal log occurrence, for each device, calculate the load fluctuation amplitude σ L , electrical parameter change amplitude σ Q and abnormal log occurrence E k in each time window, then determine the distribution of each parameter in all devices, and calculate the stability weight W k , the determination of the stability weight is based on the historical operation performance of the device, the higher the weight value represents the smaller the historical operation fluctuation of the device, and the weight calculation method is:

[0129]

[0130] Assuming that the load fluctuation σ L of a certain device is 5%, the electrical parameter fluctuation σ Q is 3V, and the abnormal log number E k is 2, then:

[0131]

[0132] Then, based on the stability weight W k of each device, calculate its stability score S score , the score calculation method is:

[0133] S score = W k ×(1-σ L )×(1-σ Q )×(1-E k );

[0134] Assuming that the load fluctuation σ L of a device is 5%, the electrical parameter fluctuation σ Q is 3V, and the abnormal log number E k is 2, calculate its stability score:

[0135] S score = 0.16 × (1-0.05) × (1-0.03) × (1-0.02) = 0.16 × 0.95 × 0.97 × 0.98 = 0.142;

[0136] Finally, aggregate the stability scores of all devices to obtain the overall device running stability score, as shown in Table 3.

[0137] Table 3 Device stability score

[0138]

[0139] The score can be used to further calculate the stability deviation value, as shown in Table 3.

[0140] S302: Based on the device stability score, analyze the load fluctuation, electrical parameter and abnormal log frequency of the multi-device, calculate the mean value and dispersion degree of the stability score, and use the formula:

[0141]

[0142] Calculate the device stability deviation value to obtain the stability and health score;

[0143] Wherein, S stab represents the device stability deviation value, W k represents the stability weight of the kth device, L k represents the load fluctuation value of the kth device, represents the load fluctuation mean value, E k represents the abnormal log frequency of the kth device, Q k represents the electrical parameter value of the kth device, represents the electrical parameter mean value, p represents the total number of devices, I dev represents the current fluctuation degree;

[0144] Based on the device stability score, extract the score data of all devices and calculate the score mean value and dispersion degree, the score mean value is calculated as follows:

[0145]

[0146] Suppose the stability scores of devices A1, A2 and A3 are 0.142, 0.228 and 0.072 respectively, then:

[0147]

[0148] Then, calculate the device stability deviation value S stab :

[0149]

[0150] Suppose the load fluctuation mean value the electrical parameter mean value the current fluctuation degree I dev = 1.5, then calculate:

[0151]

[0152] Calculate the second part:

[0153]

[0154] The final device stability deviation value S is obtained stab = 1.61.

[0155] S303: Based on the stability and health scores, the devices are divided into high-risk devices, low-risk devices and stable operation devices according to the score range, and the state categories are marked, and the IDC device operation classification label is obtained.

[0156] Based on the stability and health scores, first define the classification label of the device, and set the classification standard as follows:

[0157] High-risk device: S stab > 2.0;

[0158] Low-risk device: 1.0 stab ≤ 2.0;

[0159] Stable operation device: S stab < 1.0;

[0160] According to the calculation result of S302, the device stability deviation value S stab = 1.61, falls into the low-risk interval, so the device is classified as a low-risk device. For all devices, calculate the stability deviation value, and mark the device operation classification label according to the classification standard, as shown in Table 4.

[0161] Table 4 Device operation classification label

[0162]

[0163] As shown in Table 4, this classification is used to identify the running state of different devices, and finally obtain the IDC device operation classification label.

[0164] Please refer to Figure 5 , the steps of obtaining the IDC device abnormal trend prediction value are as follows:

[0165] S401: According to the IDC device operation classification label, analyze the temperature, current, load and abnormal log trends in multiple time windows, calculate the state change rate of multiple devices, and obtain the device state trend parameter;

[0166] According to the IDC device operation classification label, first filter the historical operation data of devices of all categories, including the time series data of temperature, current, load and abnormal log, and divide the data set according to the set time window (1 minute, 5 minutes, 10 minutes). For each time window, extract the temperature data T s , current data I s , load data L s and the number of abnormal log occurrences E α , calculate the state change rate of each device, wherein the load change rate ALs The calculation is as follows:

[0167]

[0168] Assuming that the 5-minute time window load data of a certain device is {55, 60, 58, 62, 65}%, the change rate is calculated as follows:

[0169]

[0170] The state trend parameter of the device is calculated as follows:

[0171]

[0172] For current change rate ΔI s and temperature change rate ΔT s , the calculation is the same, for example, current data is {12.5, 12.8, 12.7, 13.0, 13.2} A, and temperature data is {32, 33, 34, 34, 35} ℃, then:

[0173]

[0174] Finally, the state change rate of all devices is calculated, and the device state trend parameter is obtained, as shown in Table 5.

[0175] Table 5 Device state trend parameter

[0176]

[0177] As shown in Table 5, this data is used to further calculate the load peak change parameter.

[0178] S402: Based on the device state trend parameter, analyze the load peak change of multiple devices, calculate the load peak growth rate and fluctuation range of the differentiated time period, and use the formula:

[0179]

[0180] Calculate the load peak change parameter and integrate the load peak change trend;

[0181] Where P peak represents the load peak change parameter, T s represents the temperature data of the s period, T avg represents the average temperature, I s represents the current data of the s period, L' s represents the load data of the s period, represents the average load data, r represents the total number of time windows, and E α represents the frequency of abnormal log occurrence;

[0182] Based on the device state trend parameters, the temperature, current and load data of different time windows are extracted, and the load peak change parameter P is calculated peak The calculation method is as follows:

[0183]

[0184] Wherein, assuming that the 5-minute window temperature data of a certain device is {32, 33, 34, 34, 35}℃, the current data is {12.5, 12.8, 12.7, 13.0, 13.2}A, the load data is {55, 60, 58, 62, 65}%, the number of abnormal logs E α =3, the calculation is as follows:

[0185]

[0186] The load fluctuation is calculated as follows:

[0187]

[0188] Finally, the load peak change parameters of all devices are summarized, as shown in Table 6.

[0189] Table 6 Load peak change parameter

[0190]

[0191]

[0192] As shown in Table 6, the data is used to predict future abnormal trends.

[0193] S403: Based on the load peak change trend, the abnormal device state change mode is summarized, and the abnormal change amplitude of the future time window is calculated to obtain the IDC device abnormal trend prediction value.

[0194] Based on the load peak change trend, the abnormal device state change mode is summarized, and the abnormal change amplitude of the future time window is calculated. First, the historical load fluctuation data of high-risk and low-risk devices is extracted, and the growth trend is calculated. The load growth rate between time windows t and t+1 is set as:

[0195]

[0196] Assuming that the load data of a certain device is {55, 60, 65, 68, 72}%, then:

[0197]

[0198] The future load growth trend is calculated as follows:

[0199]

[0200] The trend is extrapolated to the next time window:

[0201]

[0202] Finally, the IDC device anomaly trend prediction value is obtained, as shown in Table 7.

[0203] Table 7 Prediction of load in future time window

[0204]

[0205] As shown in Table 7, the data is used to identify abnormal trends in advance, and finally the IDC device anomaly trend prediction value is obtained.

[0206] See Figure 6 , the steps for obtaining the IDC device maintenance priority ranking are as follows:

[0207] S501: Based on the IDC device anomaly trend prediction value, calculate the maintenance priority of multiple devices, and obtain the device maintenance priority parameter according to the abnormal trend change rate and the historical operation failure rate of the device;

[0208] Based on the IDC device anomaly trend prediction value, first extract the abnormal trend change rate U q and the historical operation failure rate of all devices, and calculate the maintenance priority parameter of each device. For each device, the abnormal trend change rate U q is determined by calculating the load growth rate ΔL and the abnormal log growth rate ΔE of adjacent time windows, and the calculation method is as follows:

[0209]

[0210] Assume that the load data of device A1 in the time window is {65, 70, 75, 80, 85}%, and the number of abnormal logs is {3, 4, 5, 6, 7}, the change rate is calculated as follows:

[0211] ΔL1 = 70-65 = 5, ΔL2 = 75-70 = 5;

[0212] ΔE1 = 4-3 = 1, ΔE2 = 5-4 = 1;

[0213]

[0214] Calculate the abnormal trend change rate of all devices, and combine the historical operation failure rate to obtain the device maintenance priority parameter, as shown in Table 8.

[0215] Table 8 Device maintenance priority parameter

[0216]

[0217] The data is used to further calculate the maintenance demand index, as shown in Table 8.

[0218] S502: Based on the equipment maintenance priority parameter, analyze the emergency degree of multi-equipment maintenance, calculate the maintenance demand index, and use the formula:

[0219]

[0220] Calculate the maintenance demand index to obtain the equipment maintenance order adjustment value;

[0221] Wherein, M req represents the maintenance demand index, U q represents the abnormal trend change rate of the qth equipment, R q represents the average of all equipment abnormal trends, W q represents the cumulative failure influence factor of the qth equipment, X q represents the load stability index of the qth equipment, p represents the total number of equipment, Y δ represents the equipment state fluctuation parameter;

[0222] Based on the equipment maintenance priority parameter, extract the abnormal trend change rate U q and the operating failure rate of all equipment, and calculate the maintenance demand index M req , the calculation method is as follows:

[0223]

[0224] Wherein, R q represents the average of all equipment abnormal trends:

[0225]

[0226] Suppose the abnormal trend change rates of equipment A1, A2, A3 are {6.0, 7.2, 5.8}:

[0227]

[0228] Calculate the error term:

[0229]

[0230] Calculate the cumulative failure influence factor W q and the load stability index X q of the equipment:

[0231]

[0232] Set the equipment state fluctuation parameter Y δ = 0.8, and calculate the maintenance demand index:

[0233]

[0234] Finally, the maintenance requirement index of all devices is summarized, as shown in Table 9.

[0235] Table 9 Device Maintenance Requirement Index

[0236]

[0237]

[0238] As shown in Table 9, the data is used to optimize the device maintenance order.

[0239] S503: Based on the device maintenance order adjustment value, the device maintenance order is optimized and adjusted, and the IDC device maintenance priority order is obtained according to the maintenance priority.

[0240] Based on the device maintenance order adjustment value, the devices are sorted according to the maintenance requirement index, and the maintenance priority sorting rule is as follows:

[0241] High priority (need immediate maintenance): M > 0.25; req

[0242] Medium priority (short-term maintenance): 0.20≤M < 0.25; req

[0243] Low priority (routine inspection): M < 0.20; req

[0244] According to the calculation result of S502, the maintenance requirement index of device A2 is 0.26, which belongs to high priority, the maintenance requirement index of device A1 is 0.21, which belongs to medium priority, and the maintenance requirement index of device A3 is 0.18, which belongs to low priority. Finally, the IDC device maintenance priority order is obtained, as shown in Table 10.

[0245] Table 10 Device Maintenance Priority Order

[0246]

[0247] As shown in Table 10, the sorting result is used to determine the maintenance plan of the IDC device, and the IDC device maintenance priority order is finally obtained.

[0248] An IDC device intelligent detection and classification management system, the IDC device intelligent detection and classification management system is used to execute the IDC device intelligent detection and classification management method, and the system comprises:

[0249] ​​​The device running state monitoring module collects server current, voltage, power consumption, temperature data, obtains CPU / GPU occupancy, storage I / O load, network throughput, stores to the data monitoring system, detects log extracts error code, key parameters, character filtering is stored to the abnormal log library, calculates the adjacent change rate of current, voltage, network throughput, marks the mutation point and stores, obtains the device running state data;

[0250] The load fluctuation and abnormality analysis module calculates the server load change rate based on the device running state data, normalizes the CPU / GPU load, storage I / O load, network throughput, calculates the first-order difference value of the server temperature, current, voltage, extracts the abnormal log category, counts the log occurrence times, calculates the abnormal weight, and obtains the IDC device state fluctuation degree;

[0251] The server running stability evaluation module calls the IDC device state fluctuation degree, calculates the server running stability score, calculates the current state score according to the load fluctuation, electrical parameter change, and abnormal log weight, calculates the health state score combined with the power-off times, current and voltage stability, and fault times, classifies or stably runs the device, and obtains the IDC device running classification label;

[0252] The IDC device abnormal trend prediction module extracts the temperature, current, load, and abnormal log quantity of the past target period based on the IDC device running classification label, calculates the time series trend, judges whether the trend curve rises, calculates the load peak interval change, and if the interval is shortened, the warning level is increased, and the IDC device abnormal trend prediction value is obtained.

[0253] The device maintenance priority sorting module calls the IDC device abnormal trend prediction value, calculates the maintenance priority, generates a maintenance score combined with the health score, abnormal warning level, and device business criticality score, arranges the maintenance plan according to the score, and establishes the IDC device maintenance priority sorting.

[0254] The above is only a preferred embodiment of the present application, and is not intended to limit the present application in other forms. Any skilled person in the art can modify or change the above disclosed technical content to equivalent embodiments applied to other fields, but any simple modification, equivalent change and modification made according to the technical essence of the present application to the above embodiments shall fall within the protection scope of the present application.

Claims

1. An IDC device intelligent detection and classification management method, characterized in that, Comprise the following steps: S1: Collecting the IDC cabinet server current, voltage, power consumption, temperature, synchronously recording CPU occupancy, GPU occupancy, storage I / O load, network throughput rate, extracting server log entries and storing to device exception log library, obtaining IDC device running state data; S2: Calling the IDC device running state data, analyzing the server load change rate, calculating the fluctuation degree of temperature, current and voltage, counting the number of abnormal log occurrences, obtaining the IDC device state fluctuation degree; S3: Based on the IDC device state fluctuation degree, the running stability score is calculated, the load fluctuation, electrical parameter and abnormal log are analyzed, the stability and health score are compared, the risk and stable running equipment are classified, and the IDC equipment running classification label is generated; S4: According to the IDC equipment running classification label, the temperature, current, load and abnormal log trend are analyzed, the load peak value change is predicted, and the IDC equipment abnormal trend prediction value is established; S5: According to the IDC equipment abnormal trend prediction value, the maintenance priority is calculated, the device maintenance order is adjusted, and the IDC equipment maintenance priority ranking is generated; The acquisition step of the IDC device state fluctuation degree is specifically: S201: Calling the IDC device running state data, calculating the server load change rate in the differential time window, calculating the rate change based on the CPU occupancy, GPU occupancy and storage I / O load of adjacent time period, and obtaining the mean value and fluctuation amplitude, obtaining the server load change rate; S202: Based on the server load change rate, the fluctuation value of temperature, current and voltage is calculated, the adjacent data change is monitored, and the formula is used: ； The fluctuation degree of device running parameters is calculated; wherein, representing a degree of fluctuation of a device operating parameter, representing a first period temperature data, representing a temperature mean, representing a first period current data, representing a current mean, representing a first period voltage data, representing a voltage mean, representing a number of time windows; S203: Based on the fluctuation degree of device running parameters, the abnormal log entries in the device abnormal log library are counted, the number of abnormal log in the differential time window is calculated, the fluctuation trend of differential time period is summarized combined with the fluctuation of temperature, current and voltage, and the IDC device state fluctuation degree is obtained. 2.The IDC equipment intelligent detection and classification management method of claim 1, characterized in that, The IDC device running state data is specifically server current, server voltage, network throughput rate change rate, device abnormal log library storage content, the IDC device state fluctuation degree includes server load change rate, current difference value, voltage difference value, temperature difference value, abnormal log weight, the IDC equipment running classification label specifically refers to high-risk IDC equipment and stable running IDC equipment, the IDC equipment abnormal trend prediction value specifically refers to server temperature trend, current trend, load peak interval change and abnormal log quantity trend, and the IDC equipment maintenance priority ranking includes server health score, abnormal warning level and device business importance score. 3.The IDC equipment intelligent detection and classification management method of claim 2, characterized in that, The acquisition step of the IDC device running state data is specifically: S101: Collecting the current, voltage, power consumption and temperature data of the server in the IDC cabinet, obtaining the CPU occupancy, GPU occupancy, storage I / O load and network throughput rate, recording the server log entries, and storing to the device abnormal log library, generating the device running parameter data set; S102: Based on the device operating parameter data set, normalize the server current, voltage and power consumption data, calculate the power consumption mean and fluctuation amplitude in multiple time windows, and calculate the heat dissipation state change quantity combined with the server temperature data, using the formula: ; A server temperature fluctuation value is calculated to generate a server heat dissipation state parameter; wherein, representing a server temperature fluctuation value, representing a server temperature of a first time window, representing a temperature mean value, representing a total number of time windows, representing a maximum power consumption value, representing a minimum power consumption value, representing a voltage mean value; S103: Based on the server heat dissipation state parameter, CPU occupancy rate, GPU occupancy rate, storage I / O load and network throughput rate data are called to calculate the change trend of device power consumption and heat dissipation parameters under a differentiated load state, extract server abnormal state features, and generate IDC device running state data.

4. The IDC equipment intelligent detection and classification management method of claim 1, wherein, The obtaining step of the IDC device running classification label is specifically: S301: Based on the IDC device state fluctuation degree, the stability score of multi-device running is calculated, the change of load fluctuation, electrical parameters and abnormal logs in the differentiated time window is counted, and the stability weight of multiple indexes is calculated to obtain the device stability score; S302: Based on the device stability score, the load fluctuation, electrical parameters and abnormal log frequency of multi-device are analyzed, the mean and dispersion degree of the stability score are calculated, and the formula is used: ； The device stability deviation value is calculated to obtain the stability and health score; wherein, representing a device stability deviation value, representing a first device stability weight, representing a first device load fluctuation value, representing a load fluctuation mean value, representing a first device abnormal log frequency, representing a first device electrical parameter value, representing an electrical parameter mean value, representing a total number of devices, representing a current fluctuation degree; S303: Based on the stability and health score, the device is divided into high-risk device, low-risk device and stable running device according to the score range, and its state category is marked to obtain the IDC device running classification label.

5. The IDC equipment intelligent detection and classification management method of claim 1, wherein, The obtaining step of the IDC device abnormal trend prediction value is specifically: S401: According to the IDC device running classification label, the temperature, current, load and abnormal log trend in multiple time windows are analyzed, the multi-device state change rate is calculated, and the device state trend parameter is obtained; S402: Based on the device state trend parameter, the load peak value change of multi-device is analyzed, the load peak value growth rate and fluctuation range of differentiated time period are calculated, and the formula is used: ； The load peak value change parameter is calculated to integrate the load peak value change trend; wherein, represents a load peak variation parameter, represents a first period temperature data, represents a temperature mean value, represents a first period current data, represents a first period load data, represents a load data mean value, represents a total number of time windows, represents an abnormal log occurrence frequency; S403: Based on the load peak value change trend, the abnormal device state change mode is summarized, and the abnormal change amplitude of future time window is calculated to obtain the IDC device abnormal trend prediction value.

6. The IDC equipment intelligent detection and classification management method of claim 1, wherein, The obtaining step of the IDC device maintenance priority sorting is specifically: S501: Based on the IDC device abnormal trend prediction value, the maintenance priority of multi-device is calculated, and the device maintenance priority parameter is obtained according to the abnormal trend change rate and running failure rate of the device; S502: Based on the device maintenance priority parameter, the maintenance urgency of multi-device is analyzed, the maintenance demand index is calculated, and the formula is used: ； The maintenance demand index is calculated to obtain the device maintenance order adjustment value; wherein, represents a maintenance demand index, represents the first represents a device abnormal trend change rate, represents an average of all device abnormal trends, represents the first represents a device cumulative failure impact factor, represents the first represents a device load stability index, represents a total number of devices, represents a device state fluctuation parameter; S503: Based on the device maintenance order adjustment value, the device maintenance order is optimized and adjusted, and the IDC device maintenance priority sorting is obtained according to the maintenance priority.

7. An IDC equipment intelligent detection and classification management system, characterized in that, The IDC device intelligent detection and classification management method according to any one of claims 1-6, the system comprises: The device running state monitoring module collects server current, voltage, power consumption, temperature data, obtains CPU / GPU occupancy, storage I / O load, network throughput, stores to the data monitoring system, detects error code, key parameters, character filtering after storing to the abnormal log library, calculates the adjacent change rate of current, voltage, network throughput, marks the mutation point and stores, obtains the device running state data; The load fluctuation and abnormality analysis module calculates the server load change rate based on the device running state data, normalizes the CPU / GPU load, storage I / O load, network throughput, calculates the first-order difference value of the server temperature, current and voltage, extracts the abnormal log category, counts the log occurrence number, calculates the abnormal weight, and obtains the IDC device state fluctuation degree; The server running stability evaluation module calls the IDC device state fluctuation degree, calculates the server running stability score, calculates the current state score according to the load fluctuation, electrical parameter change and abnormal log weight, calculates the health state score combined with the power-off number, current and voltage stability and fault number, classifies or stably runs the device, and obtains the IDC device running classification label; The IDC device abnormal trend prediction module extracts the temperature, current, load and abnormal log number of the past target period based on the IDC device running classification label, calculates the time series trend, judges whether the trend curve rises, calculates the load peak interval change, and if the interval is shortened, the warning level is increased, and the IDC device abnormal trend prediction value is obtained; The device maintenance priority sorting module calls the IDC device abnormal trend prediction value, calculates the maintenance priority, generates the maintenance score combined with the health score, abnormal warning level and device business criticality score, arranges the maintenance plan according to the score, and establishes the IDC device maintenance priority sorting.

Images