Monitoring, operation and maintenance management method and system for cloud data center

Abstract

The present invention discloses a monitoring, operation and maintenance management method and system for a cloud data center, which relates to the field of cloud data center monitoring, operation and maintenance, and includes: obtaining historical data of cloud data monitoring metrics and establishing a cloud data monitoring database; based on wavelet transform, obtaining short-term fluctuations, medium-term fluctuations and long-term trends of each cloud data monitoring threshold; establishing a hardware performance index evaluation model to obtain the impact of hardware performance degradation on cloud data monitoring thresholds; obtaining the correlation between each cloud data monitoring metric and establishing an adaptive adjustment and alarm model for cloud data monitoring thresholds; determining the dynamic response time of each cloud data monitoring metric and setting the alarm level of cloud data monitoring thresholds; establishing a cloud data monitoring threshold alarm frequency model to dynamically adjust the number of cloud data monitoring threshold alarms per unit time. It effectively reduces the high false alarm rate caused by business complexity, improves the relevance and environmental adaptability of each cloud data monitoring threshold, and enhances the comprehensive quality of monitoring, operation and maintenance.

Description

Technical Field

[0001] The present invention relates to the field of cloud data center monitoring and operation and maintenance, and specifically to a monitoring and operation and maintenance management method and system for a cloud data center. Background Art

[0002] As the "heart" of the modern enterprise IT architecture, the cloud data center bears a huge amount of business applications, data storage and computing tasks. Its monitoring and operation and maintenance capabilities directly determine business continuity, user experience and cost control. In the process of monitoring and operation and maintenance management of the cloud data center, for the cloud data monitoring threshold, static threshold alarms are generally used. However, in the monitoring of the cloud data center, high false alarm rates are often caused by environmental dynamics and business complexity. Especially in peak business periods, sudden traffic or complex dependency scenarios, the problem is particularly prominent. The operation and maintenance personnel handle it fatiguedly, and real faults are submerged in the noise. Along with repetitive invalid alarms, the trust of the team in the alarms is reduced. Secondly, due to the independence of alarms between various cloud data monitoring indicators, independent alarms cannot reflect the real reasons, misleading the direction of fault troubleshooting and generating repeated events. At the same time, as the hardware performance declines, it causes the cloud data monitoring threshold to exceed the limit, and the operation and maintenance efficiency decreases, resulting in a decline in the comprehensive quality of the monitoring and operation and maintenance of the cloud data center.

[0003] The existing cloud data monitoring threshold processing technology uses a single time window, lacking hierarchical management of cloud data monitoring data and unable to separate the influences of short-term fluctuations, medium-term fluctuations and long-term trends, resulting in problems such as high false alarm rates like high-frequency false alarms and low-frequency missed alarms. Secondly, the existing technology uses simple counting and frequency limiting, without dynamically adjusting in combination with the business cycle and alarm level, easily leading to the suppression of valid alarms during peak business periods and the flooding of invalid alarms during off-peak periods, which is not conducive to the long-term operation and maintenance management of cloud data center monitoring. Summary of the Invention

[0004] To solve the above technical problems, a monitoring and operation and maintenance management method and system for a cloud data center are provided. This technical solution solves the problems of lacking hierarchical management of cloud data monitoring data and being unable to separate the influences of short-term fluctuations, medium-term fluctuations and long-term trends, resulting in high false alarm rates such as high-frequency false alarms and low-frequency missed alarms as mentioned in the above background art. Secondly, the existing technology uses simple counting and frequency limiting, without dynamically adjusting in combination with the business cycle and alarm level, easily leading to the suppression of valid alarms during peak business periods and the flooding of invalid alarms during off-peak periods, which is not conducive to the long-term operation and maintenance management problem of cloud data center monitoring.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A monitoring and operation and maintenance management method for a cloud data center includes:

[0007] Obtain the historical data of cloud data monitoring metrics based on the monitoring historical data of the cloud data center, and establish a cloud data monitoring database;

[0008] Based on wavelet transform, extract the data in the cloud data monitoring database through high, medium, and low frequency layers to obtain the short-term fluctuations, medium-term fluctuations, and long-term trends of each cloud data monitoring threshold;

[0009] According to the cloud data monitoring database, establish a hardware performance index evaluation model to obtain the impact of hardware performance degradation on the cloud data monitoring threshold;

[0010] According to the cloud data monitoring database, based on the attention mechanism, obtain the correlation between each cloud data monitoring metric, and establish an adaptive adjustment and alarm model for the cloud data monitoring threshold;

[0011] According to the cloud data monitoring threshold alarm type, determine the dynamic response time of each cloud data monitoring metric, and set the cloud data monitoring threshold alarm level;

[0012] According to the cloud data monitoring threshold alarm level, establish a cloud data monitoring threshold alarm frequency model to dynamically adjust the number of cloud data monitoring threshold alarms per unit time.

[0013] Preferably, the extracting the data in the cloud data monitoring database through high, medium, and low frequency layers based on wavelet transform to obtain the short-term fluctuations, medium-term fluctuations, and long-term trends of each cloud data monitoring threshold specifically includes:

[0014] According to the multi-scale decomposition principle of wavelet transform, divide the historical data of the cloud data monitoring metric into a three-layer structure, where the three-layer structure includes: high frequency layer, medium frequency layer, and low frequency layer;

[0015] Extract the low-frequency, medium-frequency, and high-frequency component data in the historical data of the cloud data monitoring metric through the discrete wavelet transform formula;

[0016] According to the high-frequency component data, obtain the short-term fluctuation coefficient of the high-frequency layer of the cloud data monitoring threshold through the standard deviation formula;

[0017] According to the low-frequency component data and the linear regression equation, obtain the long-term trend increment of the low-frequency layer of the cloud data monitoring threshold;

[0018] According to the short-term fluctuation coefficient and the long-term trend increment, obtain the medium-term fluctuation coefficient of the medium-frequency layer of the cloud data monitoring threshold;

[0019] According to the short-term fluctuation coefficient, medium-term fluctuation coefficient, and long-term trend increment, obtain the dynamic threshold increment of the cloud data monitoring threshold based on the weighted formula.

[0020] Preferably, establishing a hardware performance index evaluation model based on the cloud data monitoring database and obtaining the impact of hardware performance degradation on the cloud data monitoring threshold specifically includes:

[0021] Extracting hardware performance index and environmental monitoring index data according to the cloud data monitoring database;

[0022] Based on the probability density function of the Weibull distribution and historical hardware performance index data, setting a fixed period to fit the values of the shape parameter and the scale parameter;

[0023] Establishing a hardware performance index evaluation model according to the scale parameter and obtaining the adjustment amount of the cloud data monitoring threshold due to hardware performance degradation;

[0024] Dynamically adjusting the value of the shape parameter in the probability density function of the Weibull distribution according to the environmental monitoring index data;

[0025] The expression of the hardware performance index evaluation model is:

[0026] ;

[0027] In the formula, is the adjustment amount of the cloud data monitoring threshold due to hardware performance degradation, is the reference value of each cloud data monitoring threshold, is the adjustment constant, where, , is the scale parameter value, is the change amount of the scale parameter, where, is the change rate of the scale parameter;

[0028] The expression for dynamically adjusting the value of the shape parameter in the probability density function of the Weibull distribution is:

[0029] ;

[0030] In the formula, is the dynamically adjusted value of the shape parameter, is the estimated value determined by the maximum likelihood estimation function, , are the weights of the normalized temperature value and the voltage deviation value respectively, is the normalized temperature value monitored by the hardware, is the voltage deviation value monitored by the hardware, is the jitter amplitude value of the hardware monitored voltage.

[0031] Preferably, determining the dynamic response time of each cloud data monitoring index according to the cloud data monitoring threshold alarm type and setting the cloud data monitoring threshold alarm level specifically includes:

[0032] Based on the security risk assessment of cloud data monitoring threshold alarms by the security protection end of the cloud data center, obtain the security risk assessment value of cloud data monitoring threshold alarms;

[0033] According to the security risk assessment value of cloud data monitoring threshold alarms and in combination with the types of cloud data monitoring threshold alarms, obtain the dynamic response time of each cloud data monitoring indicator;

[0034] Based on the dynamic response time of each cloud data monitoring indicator and use it as the dynamic response time reference value of the cloud data monitoring indicator;

[0035] According to the type of cloud data monitoring indicator, the incremental value exceeding the reference value, and the security risk assessment value, set the cloud data monitoring threshold alarm level based on the multi-level cloud data monitoring threshold alarm strategy.

[0036] Preferably, the establishment of a cloud data monitoring threshold alarm frequency model according to the cloud data monitoring threshold alarm level and the dynamic adjustment of the number of cloud data monitoring threshold alarms within a unit time specifically includes:

[0037] Set the initial number of alarms exceeding the dynamic response time of the cloud data monitoring indicator, denoted as ;

[0038] Establish a cloud data monitoring threshold alarm frequency model according to the cloud data monitoring threshold alarm level;

[0039] Through the cloud data monitoring threshold alarm frequency model, dynamically adjust the number of cloud data monitoring threshold alarms within a unit time;

[0040] The expression of the cloud data monitoring threshold alarm frequency model is:

[0041] ;

[0042] In the formula, is the number of cloud data monitoring threshold alarms within a unit time, is the cloud data monitoring threshold alarm level, is the gain coefficient, is the ceiling symbol.

[0043] Furthermore, this solution proposes a monitoring and operation and maintenance management system for a cloud data center, which is used to implement the monitoring and operation and maintenance management method of the cloud data center as described above, including:

[0044] A data acquisition module, which is used to obtain the historical data of cloud data monitoring indicators according to the monitoring historical data of the cloud data center and establish a cloud data monitoring database;

[0045] Adjustment and Alarm Module, which is used to extract data from the cloud data monitoring database through high, medium, and low frequency layers based on wavelet transform, and obtain the short-term fluctuations, medium-term fluctuations, and long-term trends of each cloud data monitoring threshold; establish a hardware performance index evaluation model according to the cloud data monitoring database, and obtain the impact of hardware performance degradation on the cloud data monitoring threshold; based on the attention mechanism, obtain the correlation between each cloud data monitoring index according to the cloud data monitoring database, and establish an adaptive adjustment and alarm model for the cloud data monitoring threshold.

[0046] Alarm Level and Frequency Module, which is used to determine the dynamic response time of each cloud data monitoring index according to the cloud data monitoring threshold alarm type, and set the cloud data monitoring threshold alarm level; establish a cloud data monitoring threshold alarm frequency model according to the cloud data monitoring threshold alarm level, and dynamically adjust the number of cloud data monitoring threshold alarms per unit time.

[0047] Preferably, the adjustment and alarm module specifically includes:

[0048] Dynamic Threshold Increment Unit, which is used to extract data from the cloud data monitoring database through high, medium, and low frequency layers based on wavelet transform, and obtain the short-term fluctuations, medium-term fluctuations, and long-term trends of each cloud data monitoring threshold.

[0049] Hardware Threshold Adjustment Unit, which is used to establish a hardware performance index evaluation model according to the cloud data monitoring database, and obtain the impact of hardware performance degradation on the cloud data monitoring threshold.

[0050] Adjustment and Alarm Unit, which is used to obtain the correlation between each cloud data monitoring index according to the cloud data monitoring database based on the attention mechanism, and establish an adaptive adjustment and alarm model for the cloud data monitoring threshold.

[0051] Preferably, the alarm level and frequency module specifically includes:

[0052] Alarm Level Unit, which is used to determine the dynamic response time of each cloud data monitoring index according to the cloud data monitoring threshold alarm type, and set the cloud data monitoring threshold alarm level.

[0053] Alarm Frequency Unit, which is used to establish a cloud data monitoring threshold alarm frequency model according to the cloud data monitoring threshold alarm level, and dynamically adjust the number of cloud data monitoring threshold alarms per unit time.

[0054] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0055] This solution proposes a monitoring, operation and maintenance management method for a cloud data center. Based on wavelet transform, data in the cloud data monitoring database is extracted through high, medium and low frequency layers to obtain the short-term fluctuations, medium-term fluctuations and long-term trends of each cloud data monitoring threshold. Based on a weighted formula, the dynamic threshold increment of the cloud data monitoring threshold is obtained. Secondly, according to the probability density function of the Weibull distribution, a hardware performance index evaluation model is established to obtain the impact of hardware performance degradation on the cloud data monitoring threshold, and the adjustment amount of the cloud data monitoring threshold due to hardware performance degradation is determined. Then, according to the attention mechanism formula, the correlation between each cloud data monitoring index is obtained, and based on the LSTM network algorithm, combined with the dynamic threshold increment of the cloud data monitoring threshold, the adjustment amount of the cloud data monitoring threshold due to hardware performance degradation and the correlation between each cloud data monitoring index, a cloud data monitoring threshold adaptive adjustment and alarm model is established to comprehensively evaluate the cloud data monitoring threshold alarm type and adaptively adjust the cloud data monitoring threshold. Finally, according to the cloud data monitoring threshold alarm type, the dynamic response time of each cloud data monitoring index is determined, the cloud data monitoring threshold alarm level is set, and a cloud data monitoring threshold alarm frequency model is established to dynamically adjust the number of cloud data monitoring threshold alarms per unit time, so as to improve the dynamic adjustment of the cloud data monitoring threshold alarm, effectively reduce the high false alarm rate caused by business complexity, improve the relevance and environmental adaptability of each cloud data monitoring threshold, and enhance the comprehensive quality of monitoring and operation and maintenance. Description of the Drawings

[0056] Figure 1 It is a flowchart of the monitoring, operation and maintenance management method for the cloud data center of the present invention;

[0057] Figure 2 It is a flowchart of the present invention for extracting data in the cloud data monitoring database through high, medium and low frequency layers based on wavelet transform to obtain the short-term fluctuations, medium-term fluctuations and long-term trends of each cloud data monitoring threshold;

[0058] Figure 3 It is a flowchart of the present invention for establishing a hardware performance index evaluation model based on the cloud data monitoring database to obtain the impact of hardware performance degradation on the cloud data monitoring threshold;

[0059] Figure 4 It is a flowchart of the present invention for determining the dynamic response time of each cloud data monitoring index according to the cloud data monitoring threshold alarm type and setting the cloud data monitoring threshold alarm level;

[0060] Figure 5 It is a flowchart of the present invention for establishing a cloud data monitoring threshold alarm frequency model according to the cloud data monitoring threshold alarm level and dynamically adjusting the number of cloud data monitoring threshold alarms per unit time. Detailed Embodiment

[0061] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments in the following description are only examples, and other obvious variations can be conceived by those skilled in the art.

[0062] Referring to Figure 1 As shown, a monitoring, operation and maintenance management method for a cloud data center includes:

[0063] According to the monitoring historical data of the cloud data center, obtain the historical data of cloud data monitoring metrics and establish a cloud data monitoring database;

[0064] Based on wavelet transform, extract the data in the cloud data monitoring database through high, medium and low frequency layers to obtain the short-term fluctuations, medium-term fluctuations and long-term trends of each cloud data monitoring threshold;

[0065] According to the cloud data monitoring database, establish a hardware performance metric evaluation model to obtain the impact of hardware performance degradation on the cloud data monitoring threshold;

[0066] According to the cloud data monitoring database, based on the attention mechanism, obtain the correlations between each cloud data monitoring metric and establish a cloud data monitoring threshold adaptive adjustment and alarm model;

[0067] According to the cloud data monitoring threshold alarm type, determine the dynamic response time of each cloud data monitoring metric and set the cloud data monitoring threshold alarm level;

[0068] According to the cloud data monitoring threshold alarm level, establish a cloud data monitoring threshold alarm frequency model to dynamically adjust the number of cloud data monitoring threshold alarms per unit time.

[0069] It can be explained that during the monitoring, operation and maintenance management of the cloud data center, the dynamic adjustment of the cloud data monitoring threshold needs to pay attention to the periodic and seasonal impacts, as well as the impact of hardware performance degradation. Secondly, simple counting and frequency limiting, without dynamically adjusting in combination with the business cycle and alarm level, are prone to the problem that valid alarms are suppressed during the business peak period and invalid alarms flood during the low peak period. Therefore, based on wavelet transform, this solution extracts data from the cloud data monitoring database through high, medium, and low frequency layers to obtain the short-term fluctuations, medium-term fluctuations, and long-term trends of each cloud data monitoring threshold. Based on the weighted formula, the dynamic threshold increment of the cloud data monitoring threshold is obtained. Secondly, according to the probability density function of the Weibull distribution, a hardware performance index evaluation model is established to obtain the impact of hardware performance degradation on the cloud data monitoring threshold, and the adjustment amount of the cloud data monitoring threshold due to hardware performance degradation is determined. Then, according to the attention mechanism formula, the correlation between each cloud data monitoring index is obtained, and based on the LSTM network algorithm, combined with the dynamic threshold increment of the cloud data monitoring threshold, the adjustment amount of the cloud data monitoring threshold due to hardware performance degradation, and the correlation between each cloud data monitoring index, a cloud data monitoring threshold adaptive adjustment and alarm model is established to comprehensively evaluate the cloud data monitoring threshold alarm type and adaptively adjust the cloud data monitoring threshold. Finally, according to the cloud data monitoring threshold alarm type, the dynamic response time of each cloud data monitoring index is determined, the cloud data monitoring threshold alarm level is set, and a cloud data monitoring threshold alarm frequency model is established to dynamically adjust the number of cloud data monitoring threshold alarms per unit time, thereby improving the dynamic adjustment of the cloud data monitoring threshold alarm, effectively reducing the high false alarm rate caused by business complexity, improving the correlation and environmental adaptability of each cloud data monitoring threshold, and enhancing the comprehensive quality of monitoring and operation and maintenance.

[0070] Specifically, obtaining the historical data of the cloud data monitoring metrics according to the historical data of the cloud data center monitoring, and establishing a cloud data monitoring database includes:

[0071] Obtaining the historical data of the cloud data monitoring metrics according to the historical data of the cloud data center monitoring, where the cloud data monitoring metrics include: metrics with threshold alarms, hardware performance metrics, and environmental monitoring metrics;

[0072] Establishing a cloud data monitoring database according to the historical data of the cloud data center monitoring, where the cloud data monitoring database includes: historical data of cloud data monitoring metrics, historical fault records, and threshold adjustment records;

[0073] Using the cloud data monitoring database, according to the time stamps of data sampling, aligning and synchronizing multi-source data through interpolation and data downsampling;

[0074] Performing normalization processing on the cloud data monitoring database, mapping the data in the cloud data monitoring database to the interval to eliminate the influence of dimension.

[0075] It can be explained that historical data needs to be preprocessed when dynamically adjusting the cloud data monitoring threshold, especially the alignment and synchronization of multi-source data. In this solution, historical data monitored by the cloud data center is used to obtain historical data of cloud data monitoring metrics, a cloud data monitoring database is established, and according to the timestamps of data sampling, multi-source data is aligned and synchronized through interpolation and data downsampling. At the same time, the cloud data monitoring database is normalized, and the data in the cloud data monitoring database is mapped to an interval to eliminate the influence of dimension. Among them, the metrics with threshold alarms refer to the metrics with preset critical values for triggering alarms or notifications, such as CPU usage, memory usage, network bandwidth, and disk usage, etc. The environmental monitoring metrics refer to the temperature and voltage of the monitored hardware.

[0076] Refer to Figure 2 As shown, based on wavelet transform, the data in the cloud data monitoring database is extracted through high, medium, and low frequency layers, and obtaining the short-term fluctuations, medium-term fluctuations, and long-term trends of each cloud data monitoring threshold specifically includes:

[0077] According to the multi-scale decomposition principle of wavelet transform, the historical data of cloud data monitoring metrics is divided into a three-layer structure, where the three-layer structure includes: high frequency layer, medium frequency layer, and low frequency layer;

[0078] Through the discrete wavelet transform formula, the low-frequency, medium-frequency, and high-frequency component data in the historical data of cloud data monitoring metrics are extracted;

[0079] According to the high-frequency component data, through the standard deviation formula, the short-term fluctuation coefficient of the high-frequency layer of cloud data monitoring threshold is obtained;

[0080] According to the low-frequency component data and the linear regression equation, the long-term trend increment of the low-frequency layer of cloud data monitoring threshold is obtained;

[0081] According to the short-term fluctuation coefficient and the long-term trend increment, the medium-term fluctuation coefficient of the medium-frequency layer of cloud data monitoring threshold is obtained;

[0082] According to the short-term fluctuation coefficient, medium-term fluctuation coefficient, and long-term trend increment, based on the weighted formula, the dynamic threshold increment of cloud data monitoring threshold is obtained.

[0083] It can be explained that, during the monitoring, operation and maintenance management of the cloud data center, the dynamic adjustment of the cloud data monitoring threshold needs to consider periodic and seasonal impacts. By obtaining the short-term fluctuations, medium-term fluctuations and long-term trends of each cloud data monitoring threshold, the cloud data monitoring threshold can be effectively adjusted dynamically. Therefore, in this solution, through discrete wavelet transform, the historical data of the cloud data monitoring indicators is divided into a three-layer structure. By analyzing the data frequency components of each layer, the dynamic threshold increment of the cloud data monitoring threshold can be effectively obtained;

[0084] Among them, the first layer is the high-frequency layer, which is used to capture the data cycle fluctuations at the minute and hour levels. The second layer is the medium-frequency layer, which is used to capture the data cycle fluctuations at the day and week levels. The second layer is the low-frequency layer, which is used to capture the long-term trends at the month and quarter levels;

[0085] First layer decomposition:

[0086] ;

[0087] Second layer decomposition:

[0088] ;

[0089] Third layer decomposition:

[0090] ;

[0091] In the formula, is the low-frequency output of the first layer, is the low-frequency output of the second layer, is the low-frequency output of the third layer. Among them, the low-frequency output of each layer is the low-frequency output after low-pass filtering and downsampling of the data of the previous layer, is the lowest-frequency output, is the high-frequency output of the first layer, is the medium-frequency output of the second layer, is the medium-low-frequency output of the third layer, is the historical data of the cloud data monitoring indicators, is the number of data groups of the historical data of the cloud data monitoring indicators, is the low-pass filter, which captures the low-frequency components of the signal, is the high-pass filter, which captures the high-frequency components of the signal, is the filter displacement index, which is used to control the sliding window position of the filter;

[0092] The expression of the medium-term fluctuation coefficient of the medium-frequency layer of the cloud data monitoring threshold is:

[0093] ;

[0094] In the formula, is the medium-term fluctuation coefficient of the intermediate frequency layer of the cloud data monitoring threshold, is the short-term fluctuation coefficient of the high-frequency layer of the cloud data monitoring threshold, is the long-term trend increment of the low-frequency layer of the cloud data monitoring threshold, is the weight of the short-term fluctuation coefficient of the high-frequency layer of the cloud data monitoring threshold;

[0095] The dynamic threshold increment expression of the cloud data monitoring threshold is:

[0096] ;

[0097] In the formula, is the dynamic threshold increment of the cloud data monitoring threshold, is the equilibrium constant term, used to limit the dimension of the value, , , are the weights of the short-term fluctuation coefficient, medium-term fluctuation coefficient and long-term trend increment respectively, is the error constant term.

[0098] Referring to Figure 3 as shown, establishing a hardware performance index evaluation model based on the cloud data monitoring database and obtaining the impact of hardware performance degradation on the cloud data monitoring threshold specifically includes:

[0099] Extracting hardware performance index and environmental monitoring index data according to the cloud data monitoring database;

[0100] Based on the probability density function of the Weibull distribution and the historical data of the hardware performance index, setting a fixed period to fit the values of the shape parameter and the scale parameter;

[0101] Establishing a hardware performance index evaluation model according to the scale parameter and obtaining the adjustment amount of the cloud data monitoring threshold caused by the hardware performance degradation;

[0102] Dynamically adjusting the value of the shape parameter in the probability density function of the Weibull distribution according to the environmental monitoring index data;

[0103] The expression of the hardware performance index evaluation model is:

[0104] ;

[0105] In the formula, is the adjustment amount of the cloud data monitoring threshold caused by the hardware performance degradation, is the reference value of each cloud data monitoring threshold, is the adjustment constant, where, , is the scale parameter value, is the change amount of the scale parameter, where, is the rate of change of scale parameter;

[0106] The numerical expression of the shape parameter in the probability density function of the dynamically adjusted Weibull distribution is:

[0107] ;

[0108] In the formula, Dynamically adjust the value of the shape parameter, is the estimated value determined by the maximum likelihood estimation function, 、 are the weights of the normalized temperature value and voltage deviation value, is the normalized temperature value of hardware monitoring, is the voltage deviation value monitored by the hardware, The jitter amplitude value of the hardware monitoring voltage.

[0109] It can be explained that the hardware performance of the cloud data center's monitoring is also an important factor affecting the dynamic adjustment of the cloud data monitoring threshold. As the hardware performance declines, the stable cloud data monitoring threshold may change. Therefore, when dynamically adjusting the cloud data monitoring threshold, the impact of hardware performance needs to be considered. This scheme uses the probability density function of the Weibull distribution to set a fixed period to fit the shape parameter and scale parameter values, and establishes a hardware performance index evaluation model based on the scale parameter to obtain the adjustment amount of the cloud data monitoring threshold due to the reduction in hardware performance. The probability density function of the Weibull distribution is used to describe the distribution law of continuous random variables, such as hardware failure time and performance indicators. This scheme effectively feeds back the loss of the hardware performance of the cloud data center's monitoring through the probability density function of the Weibull distribution, thereby obtaining the adjustment amount of the cloud data monitoring threshold due to the reduction in hardware performance.

[0110] The method of obtaining the correlation between cloud data monitoring indicators based on the cloud data monitoring database and the attention mechanism, and establishing a cloud data monitoring threshold adaptive adjustment and alarm model specifically includes:

[0111] Extract historical data of each cloud data monitoring indicator based on the cloud data monitoring database;

[0112] According to the attention mechanism formula, the correlation between the monitoring indicators of each cloud data is obtained;

[0113] Based on the LSTM network algorithm, a cloud data monitoring threshold adaptive adjustment and alarm model is established to comprehensively evaluate the cloud data monitoring threshold alarm type and adaptively adjust the cloud data monitoring threshold.

[0114] It can be explained that by obtaining the dynamic threshold increment of the cloud data monitoring threshold and the adjustment amount of the cloud data monitoring threshold due to the reduction of hardware performance, the cloud data monitoring threshold can be effectively adjusted adaptively. And through the attention mechanism formula, the correlation between various cloud data monitoring indicators can be obtained. Based on the correlation between various cloud data monitoring indicators, the ability to identify the warning types of the cloud data monitoring threshold can be improved. Therefore, based on the LSTM network algorithm, a double-layer LSTM network mode is set. Among them, for the first-layer LSTM network mode, the input is the cloud data monitoring indicator data, and the output is the adaptive adjustment value of the cloud data monitoring threshold. For the second-layer LSTM network mode, the input is the adaptive adjustment value of the cloud data monitoring threshold and the correlation value between various cloud data monitoring indicators, and the output is the label value of the warning type of the data monitoring threshold. Combining the dynamic threshold increment formula of the cloud data monitoring threshold and the hardware performance index evaluation model, an adaptive adjustment and warning model for the cloud data monitoring threshold is established to comprehensively evaluate the warning type of the cloud data monitoring threshold and adaptively adjust the cloud data monitoring threshold.

[0115] Referring to Figure 4 As shown, determining the dynamic response time of each cloud data monitoring indicator according to the warning type of the cloud data monitoring threshold and setting the warning level of the cloud data monitoring threshold specifically includes:

[0116] Based on the security risk assessment of the cloud data monitoring threshold warning by the security protection end of the cloud data center, obtain the security risk assessment value of the cloud data monitoring threshold warning;

[0117] Based on the security risk assessment value of the cloud data monitoring threshold warning and combined with the warning type of the cloud data monitoring threshold, obtain the dynamic response time of each cloud data monitoring indicator;

[0118] Based on the dynamic response time of each cloud data monitoring indicator and use it as the dynamic response time reference value of the cloud data monitoring indicator;

[0119] Based on the cloud data monitoring indicator type, the increment value exceeding the reference value, and the security risk assessment value, set the warning level of the cloud data monitoring threshold based on the multi-level cloud data monitoring threshold warning strategy.

[0120] It can be explained that when monitoring and managing a cloud data center, setting the response time of cloud data monitoring metrics and the warning level of cloud data monitoring thresholds is a very important step. Among them, setting an appropriate response time value for cloud data monitoring metrics can effectively and timely obtain and respond to key warning information, optimize resource allocation, avoid false alarms, and at the same time, through quantifying risk priorities, achieve efficient resource allocation and rapid fault isolation. This solution obtains the dynamic response time and warning level of cloud data monitoring thresholds by establishing dynamic response time expressions for each cloud data monitoring metric and a formula for the warning level of cloud data monitoring thresholds. Among them, the multi-level cloud data monitoring threshold warning strategy refers to a dynamic hierarchical warning mechanism based on the importance weight of metrics, deviation quantization values, and security risk assessment values. Its core logic is: when the impact weight of the monitoring metric is low, the increment of the actual value exceeding the benchmark threshold is small and the security risk assessment value is small, a low-level warning is triggered; when the impact weight of the monitoring metric is high, the deviation increment is significant and the security risk assessment value is large, a high-level warning is triggered. Among them, according to the impact weight of different monitoring metrics, the increment of the actual value exceeding the benchmark threshold, and the security risk assessment value, different warning evaluation values for cloud data monitoring thresholds are obtained through weighting. Through the warning evaluation value of cloud data monitoring thresholds, grade intervals are set by the expert scoring method, and the warning evaluation values of cloud data monitoring thresholds within the same grade interval are divided into the same grade;

[0121] The dynamic response time expression for each cloud data monitoring metric is:

[0122] ;

[0123] In the formula, is the dynamic response time of the th cloud data monitoring metric, is the minimum response time threshold to prevent the response time from being too short, is the adjustment coefficient to control the sensitivity of the response time to the security risk assessment value, is the security risk assessment value of the th cloud data monitoring metric, is the th label value of the cloud data monitoring metric;

[0124] The expression for the warning level of cloud data monitoring thresholds is:

[0125] ;

[0126] In the formula, is the warning evaluation value of cloud data monitoring thresholds, 、 、 They are the weights of the cloud data monitoring indicator type label value, the incremental value exceeding the baseline value, and the security risk assessment value. 、 、 They are the cloud data monitoring indicator type label value, the incremental value exceeding the baseline value, and the security risk assessment value;

[0127] The incremental value exceeding the benchmark value refers to the positive deviation between the actual dynamic response time of the cloud data monitoring indicator and the dynamic response time benchmark value.

[0128] Reference Figure 5 As shown, the cloud data monitoring threshold alarm frequency model is established according to the cloud data monitoring threshold alarm level, and the number of cloud data monitoring threshold alarms per unit time is dynamically adjusted, specifically including:

[0129] Set the initial alarm times for exceeding the dynamic response time of the cloud data monitoring indicator, recorded as ;

[0130] According to the cloud data monitoring threshold alarm level, establish a cloud data monitoring threshold alarm frequency model;

[0131] Dynamically adjust the number of cloud data monitoring threshold alarms per unit time through the cloud data monitoring threshold alarm frequency model;

[0132] The cloud data monitoring threshold alarm frequency model expression is:

[0133] ;

[0134] In the formula, The number of cloud data monitoring threshold alarms per unit time. Cloud data monitoring threshold alarm level, is the gain coefficient, The symbol for rounding up.

[0135] It can be explained that the traditional fixed-frequency alarm strategy cannot adapt to the dynamics and business complexity of the cloud environment. By simply counting and limiting the frequency without dynamically adjusting in combination with the business cycle and alarm level, it is easy to cause the problem that valid alarms are suppressed during the business peak period and invalid alarms flood during the low peak period, resulting in the problem of "alarms that should be reported are not reported, and alarms that should not be reported are reported wildly". Therefore, in this solution, a cloud data monitoring threshold alarm frequency model is established according to the cloud data monitoring threshold alarm level to dynamically adjust the number of cloud data monitoring threshold alarms within a unit time. Among them, the cloud data monitoring threshold alarm level reflects the comprehensive evaluation of the cloud data monitoring index type, the incremental value exceeding the reference value, and the security risk assessment value. Therefore, the severity of the cloud data monitoring threshold alarm can be effectively reflected through the cloud data monitoring threshold alarm level, so as to achieve the precise adaptation of the alarm frequency to the business cycle and the alarm severity, be able to more timely and accurately feedback the abnormal situation of cloud data, and dynamically suppress low-risk repeated alarms according to the alarm level to ensure that operation and maintenance personnel focus on high-priority issues, guarantee the stable operation of the system and improve the operation and maintenance efficiency.

[0136] Furthermore, based on the same inventive concept as the above-mentioned monitoring and operation and maintenance management method for the cloud data center, this solution proposes a monitoring and operation and maintenance management system for the cloud data center, including:

[0137] A data acquisition module, which is used to obtain the historical data of cloud data monitoring indicators according to the monitoring historical data of the cloud data center and establish a cloud data monitoring database;

[0138] An adjustment and alarm module, which is used to extract data in the cloud data monitoring database through high, medium, and low frequency layers based on wavelet transform to obtain the short-term fluctuations, medium-term fluctuations, and long-term trends of each cloud data monitoring threshold; establish a hardware performance index evaluation model according to the cloud data monitoring database to obtain the impact of hardware performance degradation on the cloud data monitoring threshold; based on the attention mechanism, obtain the correlation between each cloud data monitoring indicator according to the cloud data monitoring database, and establish a cloud data monitoring threshold adaptive adjustment and alarm model;

[0139] An alarm level and frequency module, which is used to determine the dynamic response time of each cloud data monitoring indicator according to the cloud data monitoring threshold alarm type and set the cloud data monitoring threshold alarm level; establish a cloud data monitoring threshold alarm frequency model according to the cloud data monitoring threshold alarm level to dynamically adjust the number of cloud data monitoring threshold alarms within a unit time;

[0140] The adjustment and alarm module specifically includes:

[0141] Dynamic threshold increment unit, which is used to extract data in the cloud data monitoring database through high, medium, and low frequency layers based on wavelet transform, and obtain the short-term fluctuations, medium-term fluctuations, and long-term trends of each cloud data monitoring threshold;

[0142] Hardware threshold adjustment unit, which is used to establish a hardware performance index evaluation model according to the cloud data monitoring database, and obtain the impact of hardware performance degradation on the cloud data monitoring threshold;

[0143] Adjustment and alarm unit, which is used to obtain the correlation between each cloud data monitoring index according to the cloud data monitoring database based on the attention mechanism, and establish an adaptive adjustment and alarm model for the cloud data monitoring threshold;

[0144] The alarm level and frequency module specifically includes:

[0145] Alarm level unit, which is used to determine the dynamic response time of each cloud data monitoring index according to the cloud data monitoring threshold alarm type, and set the cloud data monitoring threshold alarm level;

[0146] Alarm frequency unit, which is used to establish a cloud data monitoring threshold alarm frequency model according to the cloud data monitoring threshold alarm level, and dynamically adjust the number of cloud data monitoring threshold alarms per unit time.

[0147] In summary, the advantages of the present invention are as follows: effectively reducing the high false alarm rate caused by business complexity, improving the relevance and environmental adaptability of each cloud data monitoring threshold, and enhancing the comprehensive quality of monitoring and operation and maintenance.

[0148] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art of this industry should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification is only the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A monitoring, operation and maintenance management method for a cloud data center, characterized in that Including: Obtain the historical data of cloud data monitoring metrics based on the historical monitoring data of the cloud data center, and establish a cloud data monitoring database. Based on wavelet transform, extract the data in the cloud data monitoring database through high, medium, and low frequency layers, and obtain the short-term fluctuations, medium-term fluctuations, and long-term trends of each cloud data monitoring threshold. According to the cloud data monitoring database, establish a hardware performance index evaluation model, and obtain the impact of hardware performance degradation on the cloud data monitoring threshold. According to the cloud data monitoring database, based on the attention mechanism, obtain the correlation between each cloud data monitoring metric, and establish an adaptive adjustment and alarm model for the cloud data monitoring threshold. Based on the adaptive adjustment and alarm model for the cloud data monitoring threshold, determine the alarm type of the cloud data monitoring threshold. According to the alarm type of the cloud data monitoring threshold, determine the dynamic response time of each cloud data monitoring metric, and set the alarm level of the cloud data monitoring threshold. According to the alarm level of the cloud data monitoring threshold, establish an alarm frequency model for the cloud data monitoring threshold, and dynamically adjust the number of cloud data monitoring threshold alarms per unit time.

2. The monitoring, operation and maintenance management method of a cloud data center according to claim 1, wherein, The step of obtaining the historical data of cloud data monitoring metrics based on the historical data of the cloud data center monitoring and establishing a cloud data monitoring database specifically includes: Obtain the historical data of cloud data monitoring metrics according to the historical data of the cloud data center monitoring. Among them, the cloud data monitoring metrics include: metrics with threshold alarms, hardware performance metrics, and environmental monitoring metrics. Establish a cloud data monitoring database according to the historical data of the cloud data center monitoring. Among them, the cloud data monitoring database includes: historical data of cloud data monitoring metrics, historical fault records, and threshold adjustment records. Using the cloud data monitoring database, according to the time stamps of data sampling, align and synchronize multi-source data through interpolation and data downsampling methods. Normalize the cloud data monitoring database and map the data in the cloud data monitoring database to interval to eliminate the influence of dimension.

3. A cloud data center monitoring, operation and maintenance management method according to claim 2, characterized in that: The step of extracting the data in the cloud data monitoring database through high, medium, and low frequency layers based on wavelet transform to obtain the short-term fluctuations, medium-term fluctuations, and long-term trends of each cloud data monitoring threshold specifically includes: According to the multi-scale decomposition principle of wavelet transform, divide the historical data of cloud data monitoring metrics into three-layer structures. Among them, the three-layer structures include: high frequency layer, medium frequency layer, and low frequency layer. Extract the low-frequency, medium-frequency, and high-frequency component data in the historical data of cloud data monitoring metrics through the discrete wavelet transform formula. According to the high-frequency component data, obtain the short-term fluctuation coefficient of the high-frequency layer of the cloud data monitoring threshold through the standard deviation formula. According to the low-frequency component data and the linear regression equation, obtain the long-term trend increment of the low-frequency layer of the cloud data monitoring threshold. According to the short-term fluctuation coefficient and the long-term trend increment, obtain the medium-term fluctuation coefficient of the medium-frequency layer of the cloud data monitoring threshold. According to the short-term fluctuation coefficient, medium-term fluctuation coefficient, and long-term trend increment, obtain the dynamic threshold increment of the cloud data monitoring threshold based on the weighted formula.

4. A cloud data center monitoring, operation and maintenance management method according to claim 3, characterized in that: The step of establishing a hardware performance index evaluation model according to the cloud data monitoring database and obtaining the impact of hardware performance degradation on the cloud data monitoring threshold specifically includes: Extract the hardware performance metrics and environmental monitoring metrics data according to the cloud data monitoring database. According to the probability density function of the Weibull distribution, based on the historical data of hardware performance indicators, set a fixed period to fit the values of the shape parameter and the scale parameter; According to the scale parameter, establish a hardware performance indicator evaluation model to obtain the adjustment amount of the cloud data monitoring threshold due to the reduction of hardware performance; According to the environmental monitoring indicator data, dynamically adjust the value of the shape parameter in the probability density function of the Weibull distribution; The expression of the hardware performance indicator evaluation model is: Where, Reduce the amount of adjustment to cloud data monitoring thresholds for hardware performance, is the baseline value of each cloud data monitoring threshold, is the adjustment constant, where , is the scale parameter value, is the scale parameter change, where is the rate of change of scale parameter; The expression for dynamically adjusting the value of the shape parameter in the probability density function of the Weibull distribution is: Where, Dynamically adjust the value of the shape parameter, is the estimated value determined by the maximum likelihood estimation function, 、 are the weights of the normalized temperature value and voltage deviation value, is the normalized temperature value of hardware monitoring, is the voltage deviation value monitored by the hardware, The jitter amplitude value of the hardware monitoring voltage.

5. A monitoring, operation and maintenance management method for a cloud data center according to claim 4, characterized in that, The method of obtaining the correlation between cloud data monitoring indicators based on the attention mechanism according to the cloud data monitoring database and establishing an adaptive adjustment and alarm model for the cloud data monitoring threshold specifically includes: Extract the historical data of each cloud data monitoring indicator according to the cloud data monitoring database; Obtain the correlation between cloud data monitoring indicators according to the attention mechanism formula; Based on the LSTM network algorithm, establish an adaptive adjustment and alarm model for the cloud data monitoring threshold, comprehensively judge the alarm type of the cloud data monitoring threshold, and adaptively adjust the cloud data monitoring threshold.

6. A cloud data center monitoring, operation and maintenance management method according to claim 5, characterized in that: The method of determining the dynamic response time of each cloud data monitoring indicator according to the alarm type of the cloud data monitoring threshold and setting the alarm level of the cloud data monitoring threshold specifically includes: Obtain the security risk assessment value of the cloud data monitoring threshold alarm according to the security risk assessment of the cloud data monitoring threshold alarm by the security protection end of the cloud data center; According to the security risk assessment value of the cloud data monitoring threshold alarm, combined with the alarm type of the cloud data monitoring threshold, obtain the dynamic response time of each cloud data monitoring indicator; Based on the dynamic response time of each cloud data monitoring indicator and use it as the dynamic response time reference value of the cloud data monitoring indicator; According to the cloud data monitoring indicator type, the incremental value exceeding the reference value, and the security risk assessment value, set the alarm level of the cloud data monitoring threshold based on the multi-level cloud data monitoring threshold alarm strategy.

7. A monitoring, operation and maintenance management method for a cloud data center according to claim 6, characterized in that, The method of establishing an alarm frequency model for the cloud data monitoring threshold according to the alarm level of the cloud data monitoring threshold and dynamically adjusting the number of cloud data monitoring threshold alarms per unit time specifically includes: Set the initial alarm times for exceeding the dynamic response time of the cloud data monitoring indicator, recorded as ; Establish an alarm frequency model for the cloud data monitoring threshold according to the alarm level of the cloud data monitoring threshold; Dynamically adjust the number of cloud data monitoring threshold alarms per unit time through the alarm frequency model of the cloud data monitoring threshold; The expression of the alarm frequency model of the cloud data monitoring threshold is: Where, The number of cloud data monitoring threshold alarms per unit time. Cloud data monitoring threshold alarm level, is the gain coefficient, The symbol for rounding up.

8. A monitoring, operation and maintenance management system for a cloud data center, characterized in that, Used to implement the monitoring and operation and maintenance management method of the cloud data center as described in any one of claims 1-7, including: A data acquisition module, which is used to obtain the historical data of cloud data monitoring indicators according to the monitoring historical data of the cloud data center and establish a cloud data monitoring database; Adjustment and Alarm Module, which is used to extract data from the cloud data monitoring database through high, medium, and low frequency layers based on wavelet transform, and obtain the short-term fluctuations, medium-term fluctuations, and long-term trends of each cloud data monitoring threshold; establish a hardware performance index evaluation model according to the cloud data monitoring database to obtain the impact of hardware performance degradation on the cloud data monitoring threshold; based on the attention mechanism, obtain the correlation between each cloud data monitoring index according to the cloud data monitoring database, and establish an adaptive adjustment and alarm model for the cloud data monitoring threshold. Alarm Level and Frequency Module, which is used to determine the dynamic response time of each cloud data monitoring index according to the alarm type of the cloud data monitoring threshold, and set the alarm level of the cloud data monitoring threshold; establish an alarm frequency model for the cloud data monitoring threshold according to the alarm level of the cloud data monitoring threshold, and dynamically adjust the number of alarm times of the cloud data monitoring threshold per unit time.

9. A cloud data center monitoring, operation and maintenance management system according to claim 8, characterized in that: The adjustment and alarm module specifically includes: Dynamic Threshold Increment Unit, which is used to extract data from the cloud data monitoring database through high, medium, and low frequency layers based on wavelet transform, and obtain the short-term fluctuations, medium-term fluctuations, and long-term trends of each cloud data monitoring threshold. Hardware Threshold Adjustment Unit, which is used to establish a hardware performance index evaluation model according to the cloud data monitoring database to obtain the impact of hardware performance degradation on the cloud data monitoring threshold. Adjustment and Alarm Unit, which is used to obtain the correlation between each cloud data monitoring index according to the cloud data monitoring database based on the attention mechanism, and establish an adaptive adjustment and alarm model for the cloud data monitoring threshold.

10. A monitoring, operation and maintenance management system for a cloud data center according to claim 9, characterized in that, The alarm level and frequency module specifically includes: Alarm Level Unit, which is used to determine the dynamic response time of each cloud data monitoring index according to the alarm type of the cloud data monitoring threshold, and set the alarm level of the cloud data monitoring threshold. Alarm Frequency Unit, which is used to establish an alarm frequency model for the cloud data monitoring threshold according to the alarm level of the cloud data monitoring threshold, and dynamically adjust the number of alarm times of the cloud data monitoring threshold per unit time.

Images