System health degree evaluation method and device and electronic equipment

Abstract

The embodiment of the invention relates to the technical field of operation and maintenance monitoring, in particular to a system health degree evaluation method and device and electronic equipment. The method comprises the following steps: acquiring infrastructure performance data, application performance data and business index data of a target system as original data; performing standardization processing on the original data, and performing data arrangement according to the dimension of technical index-business index to generate standardized data; calling a business influence quantification model to calculate an influence coefficient of each technical index on each business index in the standardized data, and carrying out statistics to generate an influence matrix; according to the influence matrix, assigning a score weight of each technical index; and calculating single-index scores based on real-time technical index data in the standardized data, and performing weighted calculation on the single-index scores according to the score weights to generate a system health degree score. According to the method, the actual influence degree of the technical abnormality on the core business can be accurately quantified, and intelligent operation and maintenance transformation from pure technical monitoring to business value guarantee is realized.

Description

Technical Field

[0001] This application relates to the field of operation and maintenance monitoring technology, and in particular to a method, apparatus and electronic device for assessing system health. Background Technology

[0002] Against the backdrop of rapid development in the digital economy, online financial services, with their high-frequency transactions, real-time services, and diversified scenarios, have become a core support for economic activities. The stability and reliability of their system operation directly affect transaction success rates, user experience, and core corporate profits. Therefore, system health assessment has become a crucial link in the field of operations and maintenance monitoring. Currently, operations and maintenance work is transforming from traditional technical fault repair to business value assurance. The core demand is to accurately map the actual impact of system anomalies on core businesses through health scoring, providing clear priority guidance for operations and maintenance decisions, and avoiding resource misallocation or critical business losses caused by the disconnect between technology and business.

[0003] In existing technologies, system health assessment mainly relies on three types of schemes: First, rule-based scoring based on static thresholds, which presets fixed threshold ranges for technical indicators such as CPU utilization and memory utilization, assigns scores according to the range in which the indicators fall, and calculates a comprehensive score by weighting; second, scorecard models based on fixed weights, where experts pre-set the weights of each indicator, obtain a health score through linear weighting, and the weights remain unchanged during system operation; and third, machine learning models, which rely on historical technical monitoring data to train models for anomaly detection and scoring.

[0004] The aforementioned existing technologies have significant drawbacks: First, the technical perspective is severely disconnected from business value. The scoring results only reflect the operational status of technical components and cannot quantify the impact of system anomalies on core businesses such as transaction success rate and revenue, making it difficult for the operations and maintenance team to determine the business severity of the failure. Second, the strategies are rigid and resource allocation is inefficient. Fixed thresholds or fixed weights cannot adapt to the differentiated needs of peak and off-peak business periods, which may underestimate the impact of failures on key business-related indicators or over-alarm non-core indicator anomalies. Third, the intelligence level is insufficient and it is prone to triggering alarm storms. Machine learning models cannot converge alarms and prioritize multiple indicator anomalies based on business impact, making it difficult to support efficient operations and maintenance decisions.

[0005] Therefore, how to make the system health score accurately reflect the actual impact of technical anomalies on core business is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] The purpose of this application is to provide at least one method, device, and electronic device for assessing system health, which can accurately quantify the actual impact of technical anomalies on core business, and realize the transformation from pure technical monitoring to intelligent operation and maintenance to business value assurance.

[0007] To address the aforementioned technical problems, at least one embodiment of this application provides a method for assessing system health, comprising: Collect multi-source performance data of the target system as raw data; the multi-source performance data includes: infrastructure performance data, application performance data, and business indicator data. After standardizing the raw data, the data is organized according to the dimension of technical indicators and business indicators to generate standardized data; the standardization process includes: format standardization and time dimension alignment; The business impact quantification model is invoked to calculate the impact coefficient of each technical indicator in the standardized data on each business indicator, and an impact matrix is ​​generated statistically. Based on the influence matrix, assign scoring weights to each of the technical indicators; Based on the real-time technical indicator data in the standardized data, the scores of each individual indicator are calculated, and the scores of each individual indicator are weighted according to the scoring weight. The weighted result is used as the system health score.

[0008] In one embodiment, the method for generating the business impact quantification model includes: Machine learning algorithms are used to extract historical standardized data from the data lake and analyze the correlation between fluctuations in technical indicators and changes in business indicators. Based on the aforementioned relationship, the calculation logic for the changes in technical indicators and business indicators is defined, and an initial model is constructed. The standardized data is stored in real time in the data lake; Periodically extract real-time incremental data from the data lake, and update the model parameters in the initial model based on the real-time incremental data. Use the parameter-optimized initial model as the business impact quantification model.

[0009] In one embodiment, after standardizing the raw data, the data is organized according to the dimension of technical indicators - business indicators, including: The original data is subjected to invalid data removal and noise data filtering, and the generated data is used as preprocessed data. The preprocessed data is converted according to a preset standardized format and sorted by time to generate time series data; The technical indicator data and business indicator data within the same time window in the time series data are correlated and stored according to the correlation dimension of technical indicator-business indicator, which serves as the standardized data.

[0010] In one embodiment, the acquisition of multi-source performance data of the target system includes: The system collects server CPU utilization, memory utilization, network throughput, database query time, and middleware service status at the infrastructure layer and technical component layer as infrastructure performance data. The response time, error rate, and call frequency of key interfaces in the application layer are collected as application performance data. The transaction success rate, order volume, payment time, and user activity of the business system are collected as the business indicator data.

[0011] In one embodiment, after weighting the scores of each individual indicator according to the scoring weights, the method further includes: The weighted calculation result will be used as the preliminary health score; Obtain the business indicator data from the standardized data; Calculate the direction and degree of deviation between the business indicator data and the corresponding normal threshold; The preliminary health score is adjusted based on the direction and degree of deviation. The statement that the weighted calculation result is used as the system health score specifically means that the adjusted score is used as the system health score.

[0012] In one embodiment, adjusting the preliminary health score based on the direction of deviation and the degree of deviation includes: If the deviation direction shows that the business indicator data is lower than the corresponding normal threshold, the preliminary health score will be adjusted downward according to the degree of deviation. If the deviation direction shows that the business indicator data is higher than the corresponding normal threshold, the preliminary health score will be adjusted upwards according to the degree of deviation.

[0013] In one embodiment, after adjusting the preliminary health score based on the direction of deviation and the degree of deviation, the method further includes: Based on the aforementioned influence matrix, the core technical indicators that cause the change in the initial health score are identified in reverse. Extract the influence coefficients and influence directions corresponding to the core technical indicators; the influence direction is determined based on the positive or negative value of the influence coefficients. Estimate the change in business indicators when the core technical indicators are abnormal, and obtain the estimated value of the change in business indicators; Collect the names, impact coefficients, impact directions, and estimated changes in the business indicators corresponding to the core technical indicators, and generate an impact factor labeling report.

[0014] In one embodiment, based on the influence matrix, the core technical indicators that cause the change in the preliminary health score are identified in reverse, including: Technical indicators whose absolute values ​​of influence coefficients in the influence matrix are greater than the core indicator threshold are selected as preliminary indicators. The preliminary selection indicators are sorted in descending order of the absolute value of their influence coefficients, and the top N indicators are selected; where N is a preset positive integer. Determine the chain reaction paths between the technical indicators, and determine all technical indicators associated with the first N indicators based on the chain reaction paths; The first N indicators and their corresponding related technical indicators are taken as the core technical indicators.

[0015] At least one embodiment of this application also provides a system health assessment device, comprising: A multi-source acquisition unit is used to collect multi-source performance data of the target system as raw data; the multi-source performance data includes: infrastructure performance data, application performance data, and business indicator data. The data processing unit is used to standardize the raw data and then organize the data according to the dimension of technical indicators and business indicators to generate standardized data; the standardization process includes: format standardization and time dimension alignment. The technology and business impact assessment unit is used to call the business impact quantification model to calculate the impact coefficient of each technical indicator on each business indicator in the standardized data, and to generate an impact matrix. The weight allocation unit is used to allocate the scoring weights of each of the technical indicators according to the influence matrix. The scoring calculation unit is used to calculate the individual indicator scores based on the real-time technical indicator data in the standardized data, and to perform weighted calculation on each individual indicator score according to the scoring weight, and to use the weighted calculation result as the system health score.

[0016] At least one embodiment of this application also provides an electronic device, including: at least one processor; and a memory communicatively connected to said at least one processor; The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the system health assessment method.

[0017] The system health assessment method provided in this application collects three core data types from the target system: infrastructure performance data, application performance data, and business indicator data. This covers the entire chain from underlying support to application delivery to business value, avoiding the bias caused by single technical data. It eliminates heterogeneous differences in timestamps, numerical units, and field naming of multi-source raw data through format standardization, establishes temporal consistency between technical and business indicators through time dimension alignment, and then organizes the data into structured and standardized data according to the technical and business indicator dimensions. Based on this, it calls a business impact quantification model to calculate the impact coefficient of each technical indicator on each business indicator and generates an impact matrix. This transforms the relationship between technology and business from vague qualitative to precise quantitative, solving the problem of technical... The core pain point is the inability to measure the impact on business. Subsequently, scoring weights are assigned based on the impact matrix, avoiding subjective experience bias and enabling dynamic adaptation of weights as business scenarios change and data is updated, ensuring that the weights always match the actual business impact of technical indicators. Finally, individual indicator scores are calculated based on real-time technical indicator data in standardized data and then weighted and summed. This not only accurately captures the real-time operating status of each technical component through individual indicator quantification, but also allows the status of technical indicators with high business impact to dominate the comprehensive score through weighted calculation. This achieves a precise mapping from local technical status to overall business health, taking into account both the comprehensiveness and business orientation of the assessment. The output unified health score can help the operation and maintenance team quickly judge the system health status, clarify the priority of fault handling, and significantly reduce the cost of operation and maintenance decision-making. Attached Figure Description

[0018] One or more embodiments are illustrated by way of example with reference to the accompanying drawings, and these illustrative descriptions do not constitute a limitation on the embodiments.

[0019] Figure 1 This is a flowchart illustrating a method for assessing system health, as provided in one embodiment of this application. Figure 2 This is a schematic diagram of the module division of a system health assessment device provided in one embodiment of this application; Figure 3 This is a schematic diagram of the module division of another system health assessment device provided in one embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in one embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the various embodiments of this application will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details have been provided in the various embodiments of this application to help readers better understand this application. However, the technical solutions claimed in this application can be implemented even without these technical details and various changes and modifications based on the following embodiments. The division of the various embodiments below is for the convenience of description and should not constitute any limitation on the specific implementation of this application. The various embodiments can be combined with and referenced by each other without contradiction.

[0021] The following is a detailed description of the implementation details of the system health assessment method in this embodiment. The following content is only for the convenience of understanding the implementation details and is not necessary for implementing this solution.

[0022] Example 1: The specific process of the system health assessment method in this embodiment can be as follows: Figure 1 As shown, it includes: Step 101: Collect multi-source performance data of the target system as raw data.

[0023] For target systems that need to be assessed for health, standardized data collection methods are used to comprehensively gather multi-dimensional performance data that reflects the system's operational status and business effectiveness, which will then serve as the raw data for all subsequent health assessments and analyses.

[0024] Multi-source performance data includes three core layers: infrastructure performance data, application performance data, and business metrics data. Infrastructure performance data refers to fundamental performance data reflecting the operational status of the target system's underlying hardware and core technology components; it is the hardware and component support data that ensures the system's normal operation. Examples include server CPU utilization and memory utilization.

[0025] Application performance data refers to key performance data that reflects the running quality and service capabilities of applications in a target system, and is directly related to the service delivery effect at the application layer. Examples include interface error rate and interface call frequency.

[0026] Business performance metrics refer to core data that directly measure the effectiveness, health status, and value output of the core business activities carried by the target system, and are a direct reflection of the realization of business value. Examples include transaction success rate and transaction failure rate.

[0027] Existing technologies only collect single-dimensional technical data, which can only reflect the operating status of underlying hardware or components and cannot reflect the service quality of applications, let alone the realization of business value. In contrast, the three types of data collected in this step correspond to the three core dimensions of business value output in the basic service delivery process of system operation, covering the entire chain from underlying support to application delivery to business value. This makes the originally isolated technical data and business data form a correlated dataset and a complete evaluation data chain. From a data perspective, this ensures that the evaluation can cover the entire process of system operation and the final end of business value, avoiding evaluation bias caused by single-dimensional data.

[0028] It should be noted that this embodiment does not limit the specific data items collected in infrastructure performance data, application performance data, and business indicator data, and can be set according to the needs of the actual application scenario. Specifically, one data item configuration is as follows: Collect server CPU utilization, memory utilization, network throughput, database query time, and middleware service status at the infrastructure layer and technical component layer as infrastructure performance data; collect response time, error rate, and call frequency of key interfaces at the application layer as application performance data; collect transaction success rate, order flow, payment time, and user activity of the business system as business indicator data. This data item configuration systematically covers the core indicators of the entire chain from infrastructure to application layer to business system. Each selected indicator is a key node related to the technical status and business value, which can accurately support the business impact quantification model to explore the inherent causal relationship and significantly reduce the interference cost of model calculation. This embodiment only uses the above data item configuration as an example for introduction; the execution methods under other configurations can refer to the introduction of this embodiment and will not be repeated here.

[0029] Step 102: After standardizing the raw data, organize the data according to the dimensions of technical indicators and business indicators to generate standardized data.

[0030] Since the raw data collected in step 101 comes from different systems (infrastructure monitoring system, application performance management system, business system) and there is no unified standard for format, unit, and naming rules, standardization processing is first completed through a series of standardized operations aimed at eliminating data heterogeneity and establishing data correlation. The data is then integrated according to the correlation logic of technical indicators → corresponding business indicators, clearly distinguishing between the technical indicator dataset (including infrastructure performance indicators and application performance indicators) and the business indicator dataset, and finally generating standardized data.

[0031] Standardization processes include, but are not limited to, format standardization and time dimension alignment.

[0032] To address the heterogeneous nature of multi-source raw data, standardized conversion is performed according to preset unified standards. Timestamps are unified to the same format (such as UTC format), numerical units are unified to preset standard units (such as time-related indicators being unified to milliseconds, and traffic-related indicators being unified to MB), and field names are standardized according to the indicator type-core meaning rule (such as database query time being uniformly named db_query_latency), etc., to ensure that data from different sources and of different types maintains consistency in format.

[0033] Based on millisecond-level timestamps, establish fixed time windows (such as the time interval corresponding to the synchronization period of data collected in step 101), and bind technical indicator data (infrastructure and application performance related indicators) and business indicator data within the same time window one by one to ensure that each piece of technical indicator data can correspond to the business indicator data of the same period, and eliminate the correlation deviation caused by the difference in data collection time.

[0034] Step 103: Call the business impact quantification model to calculate the impact coefficient of each technical indicator on each business indicator in the standardized data, and generate the impact matrix.

[0035] By invoking a pre-defined business impact quantification model, the correlation strength and direction of influence between each technical indicator and each business indicator are quantitatively calculated based on the standardized data generated in step 102. For each technical indicator, its influence coefficient on each business indicator is calculated. This coefficient is a continuous value, and its absolute value represents the influence strength, while positive and negative values ​​represent the influence direction. Positive correlations correspond to indicators with the same trend, while negative correlations correspond to opposite trends. Then, all technical indicators, business indicators, and their corresponding influence coefficients are statistically integrated according to a fixed dimension, ultimately forming a structured three-dimensional influence matrix of technical indicators, business indicators, and influence coefficients. The matrix rows represent all technical indicator items, the columns represent all business indicator items, and the matrix cells are filled with the corresponding influence coefficients, forming a structured data set and realizing the transformation of the technical-business relationship from qualitative correlation to quantitative mapping.

[0036] By employing a full calculation logic for each technical indicator across all business indicators, the omissions caused by focusing only on some core correlations in traditional assessments are avoided. Through the business impact quantification model, the correlation between technical indicators and business indicators is transformed from a qualitative description into quantitative coefficients, solving the problem of ambiguous impact in traditional assessments and providing accurate quantitative decision-making basis for subsequent stages.

[0037] The business impact quantification model refers to a pre-constructed mathematical model that describes the inherent relationship between technical indicators and business indicators. Based on the input structured data, it can quantitatively output the impact parameters of technical indicators on business indicators. This embodiment does not limit the specific model structure type used; for example, correlation analysis models, linear / nonlinear regression models, random forest models, gradient boosting models, etc., as long as they can achieve the quantitative calculation of the impact relationship between technical indicators and business indicators, can be used as the business impact quantification model in this step.

[0038] Step 104: Assign scoring weights to each technical indicator based on the influence matrix.

[0039] The influence matrix is ​​retrieved, and the influence coefficients of all business indicators corresponding to each technical indicator are extracted. The absolute value of the influence coefficient is used as the core criterion, following the core logic that the higher the influence intensity, the higher the weight. Simultaneously, the weights of technical indicators with high correlation to core business indicators can be fine-tuned based on business scenario priorities. This embodiment does not limit the specific configuration rules for weight calculation. For example, the basic weights can be allocated based on the normalized calculation results of the absolute value of the influence coefficient; the range of weight fine-tuning for core business-related indicators can be configured; the triggering conditions for weight adjustment during peak and off-peak business periods can be configured; the weight tilting strategy for technical indicators corresponding to different business lines can be configured; or the influence coefficient threshold can be configured to divide weight level ranges, etc. Any configuration rule that can quantify the influence intensity of technical indicators on business indicators through influence coefficients and optimize weight allocation to meet business scenario needs falls within the scope of this step.

[0040] In existing technologies, scoring weights are mostly preset manually or fixed, failing to adapt to changes in business scenarios and the actual business impact of technical indicators. This new approach uses an influence matrix, where the influence coefficient is a quantified result of the correlation between technical and business indicators, ensuring that weights accurately correspond to the actual business impact of technical indicators. Simultaneously, as standardized data is continuously updated, the influence matrix is ​​synchronously adjusted through closed-loop optimization, and weights are redistributed based on the updated influence matrix. This achieves dynamic linkage between data updates, matrix updates, and weight updates, naturally adapting to various dynamic changes such as peak business periods, adjustments to the responsibilities of technical components, or iterations in business models.

[0041] Step 105: Calculate the individual indicator scores based on the real-time technical indicator data in the standardized data, and perform weighted calculation on the individual indicator scores according to the scoring weights. The weighted calculation result is used as the system health score.

[0042] Based on the real-time updated technical indicator data in the standardized data, the real-time operating status of each technical indicator is first quantitatively scored to obtain a single indicator score, ensuring that the real-time status of each technical indicator can be converted into a calculable quantitative score; then, the scoring weights of each technical indicator allocated in step 104 are called, and the score of each single indicator is multiplied by its corresponding weight. Finally, all the product results are summarized and summed, and the summed result is used as the comprehensive system health score of the target system, which can transform the scattered multi-dimensional technical status into a unified comprehensive score.

[0043] Based on the above introduction, the system health assessment method provided in this embodiment collects three types of core data from the target system: infrastructure performance data, application performance data, and business indicator data. This covers the entire chain from underlying support to application delivery to business value, avoiding the bias caused by single technical data. It eliminates heterogeneous differences in timestamps, numerical units, and field naming of multi-source raw data through format standardization, establishes temporal consistency between technical and business indicators through time dimension alignment, and then organizes the data into structured and standardized data according to the technical and business indicator dimensions. Based on this, it calls the business impact quantification model to calculate the impact coefficient of each technical indicator on each business indicator and generates an impact matrix. This transforms the relationship between technology and business from vague qualitative to precise quantitative, solving the problems in traditional assessments. The core pain point is the inability to measure the impact of technology on business. Subsequently, scoring weights are assigned based on an impact matrix, avoiding subjective experience bias and enabling dynamic adaptation of weights as business scenarios change and data is updated, ensuring that weights always match the actual business impact of technical indicators. Finally, individual indicator scores are calculated based on real-time technical indicator data in standardized data and then weighted and summed. This approach not only accurately captures the real-time operational status of each technical component through individual indicator quantification but also allows the status of high-business-impact technical indicators to dominate the comprehensive score through weighted calculation. This achieves a precise mapping from local technical status to overall business health, balancing the comprehensiveness and business orientation of the assessment. The output unified health score helps operations teams quickly determine system health status, clarify fault handling priorities, and significantly reduce operations decision-making costs.

[0044] Example 2: The above embodiments do not limit the generation method of the business impact quantification model. In order to ensure that the model can continuously and accurately capture the technology-business relationship and adapt to the iteration of business form and environmental fluctuations, this embodiment proposes a generation method. By mining the correlation pattern of historical data and continuously integrating the dynamic iteration mechanism of real-time incremental data, it can ensure that the model always fits the actual business scenario and stably outputs accurate technology-business impact coefficients.

[0045] Specifically, the generation of a business impact quantification model can be performed according to the following steps: Step 106: Use machine learning algorithms to extract historical standardized data from the data lake and analyze the correlation between fluctuations in technical indicators and changes in business indicators.

[0046] Based on historical standardized data (including historical technical indicators, business indicators, and time-series correlation information) stored in the data lake, this study uses pre-defined machine learning algorithms (such as correlation analysis and regression analysis, which are not limited in this embodiment) to perform in-depth data mining. The focus is on identifying the inherent correlation patterns between the fluctuation characteristics of technical indicators (such as fluctuation amplitude, duration, and trend) and the changing characteristics of business indicators (such as growth / decline magnitude and abnormal fluctuation nodes), including linear correlation, lagged correlation, nonlinear correlation, and chain reaction paths. The core patterns of interaction between technology and business are extracted from historical data, providing data support and logical basis for subsequent model construction, ensuring that the model can capture key correlations based on existing historical experience.

[0047] Step 107: Based on the correlation, define the calculation logic for the changes in technical indicators and business indicators, and build the initial model.

[0048] Based on the technical-business correlations uncovered in step 106, the abstract correlation patterns are transformed into computable mathematical logic. This clarifies the quantitative mapping rules between changes in technical indicators (e.g., an increase of X units in database query time) and corresponding changes in business indicators (e.g., a decrease of Y units in transaction success rate), including core logic such as calculation factors, correlation functions, and threshold determination. Then, based on this computational logic, an initial business impact quantification model is built, capable of inputting technical indicator data and outputting quantitative results of business impact. This achieves the transformation from qualitative description of correlation patterns to the implementation of a quantitative computational model.

[0049] Step 108: Store the real-time standardized data in the data lake.

[0050] The real-time standardized data generated in step 102 is continuously and systematically incorporated into a unified data lake, forming a complete time-series dataset with historical standardized data. This provides fresh data material for continuous model optimization. As business scenarios iterate and technical architecture is adjusted, real-time data will contain new correlation patterns and changing characteristics. Storing it in the data lake ensures that subsequent model updates can capture these dynamic changes and avoids a decrease in model adaptability due to data lag.

[0051] Step 109: Periodically extract real-time incremental data from the data lake, and update the model parameters in the initial model based on the real-time incremental data. Use the parameter-optimized initial model as the business impact quantification model.

[0052] New, real-time standardized data (i.e., real-time incremental data) is extracted from the data lake according to a preset period (such as hourly or daily, adapting to the frequency of business data updates) and used as training data for model iteration. Through online learning algorithms, the internal parameters of the initial model (such as correlation weights, impact factor thresholds, function coefficients, etc.) are dynamically adjusted using this incremental data to correct the deviation between the model and the current business scenario. After the parameters are optimized, the iterated model becomes the final callable business impact quantification model.

[0053] Through a closed-loop mechanism of continuous data input and continuous model optimization, the model can adapt to dynamic factors such as changes in business models, seasonal fluctuations, and adjustments to the responsibilities of technical components in real time, and always maintain the accuracy of capturing the relationship between technology and business, providing a reliable quantitative basis for the subsequent generation of the impact matrix.

[0054] Based on the above introduction, the model generation method provided in this embodiment extracts historical standardized data from the data lake by calling machine learning algorithms, and deeply analyzes the multi-dimensional correlation between fluctuations in technical indicators and changes in business indicators, avoiding the limitations of traditional models that rely on single data or superficial correlations. In addition, this method continuously stores real-time standardized data in the data lake, providing fresh data materials for model iteration. Combined with the mechanism of periodically extracting incremental data to update model parameters, it breaks the defect of traditional models that are rigid and unchanging after being built once. It can timely capture dynamic changes such as business form iteration, seasonal fluctuations, and technical architecture adjustments, enabling the model to continuously correct deviations, accumulate business impact knowledge, always fit the actual business scenario, and continuously improve the accuracy and reliability of impact coefficient calculation.

[0055] Example 3: In complex operation and maintenance scenarios such as online financial business, multi-source raw data often contains invalid data (such as null data from failed collection) and noisy data (such as abnormal fluctuation data caused by sudden interference). In order to avoid the deviation between technical indicators and business indicators caused by data quality defects, this embodiment proposes a standardized processing and data cleaning process, which can generate standardized data with high purity, uniform format and accurate correlation, thereby avoiding the interference of data impurities and heterogeneity problems on subsequent analysis from the source.

[0056] Specifically, in step 102, after standardizing the raw data, the data is organized according to the dimensions of technical indicators and business indicators. This can be done by following these steps: Step 21: Remove invalid data and filter noisy data from the original data, and use the generated data as preprocessed data.

[0057] For the multi-source raw data collected in step 101, a data quality purification operation is first performed. Invalid data (such as null values ​​from failed collections, malformed data, and invalid records exceeding reasonable value ranges) is identified and removed using preset filtering rules. Then, algorithms such as smoothing and threshold filtering are used to filter noisy data (such as non-business-related abnormal fluctuations caused by sudden equipment interference or instantaneous network fluctuations). Finally, a dataset reflecting the true operating status of the system after removing impurities is obtained—this is the preprocessed data. Its core purpose is to improve the purity of the raw data, preventing invalid and noisy data from interfering with subsequent standardization transformation and correlation analysis, thus laying a high-quality foundation for the entire data processing process.

[0058] Step 22: Convert the preprocessed data according to the preset standardized format and sort it by time to generate time series data.

[0059] The preprocessed data is converted according to a pre-defined unified standard, including a unified timestamp format (e.g., all converted to UTC format), a unified numerical unit (e.g., time-related indicators are unified to milliseconds, and traffic-related indicators are unified to MB), and standardized field naming rules (e.g., unified naming according to indicator type and core meaning), to eliminate heterogeneous differences between multi-source data; then, based on the data collection time, all converted data are sorted in ascending or descending order to form time series data arranged continuously in time dimension.

[0060] Step 23: Associate the technical indicator data and business indicator data within the same time window in the time series data, and store them according to the association dimension of technical indicator-business indicator as standardized data.

[0061] Based on the ordered time series data generated in step 22, fixed time windows are first divided (such as 1 minute and 5 minutes, to adapt to the data collection frequency and business analysis needs). Then, the technical indicator data (including infrastructure performance indicators and application performance indicators) and business indicator data within the same time window are bound one by one to clarify the technical operation status within a certain time period and its corresponding business performance in the same period. Finally, the bound data is classified, archived, and stored according to the correlation dimension between the technical indicator item and the corresponding business indicator item, forming standardized data with a clear structure and clear correlation. By establishing the time series correspondence between technical indicators and business indicators, it is ensured that the subsequent business impact quantification model can accurately explore the causal relationship between the two.

[0062] Based on the above introduction, this embodiment accurately removes invalid information such as null values ​​and format errors, as well as abnormal fluctuation noise caused by sudden interference, from the original data through invalid data removal and noise data filtering. This improves data purity from the source, avoids interference from impurity data in subsequent correlation analysis, and ensures that the data can truly reflect the actual state of system technical operation and business development, laying a high-quality data foundation for subsequent processing. Secondly, by converting preprocessed data according to a preset standardized format and sorting it by time, the heterogeneous differences in format, unit, and naming rules of multi-source data are effectively eliminated. At the same time, an orderly time series is constructed, which not only solves the pain points of chaotic and difficult-to-coordinate traditional multi-source data formats, but also provides clear time series support for the time alignment of technical indicators and business indicators. Thirdly, technical indicators and business indicators within the same time window are further processed... By accurately associating and classifying data according to the association dimension, the temporal correspondence between the two is strengthened, avoiding misjudgments caused by time misalignment. This allows standardized data to form a structured form with a one-to-one correspondence between technology and business, directly meeting the subsequent business impact quantification model's need for associated data and significantly improving the efficiency and accuracy of subsequent model calculations. Fourth, the entire process forms a refined closed loop of data purification, format standardization, temporal association, and classified storage. This not only improves the standardization and reliability of data processing but also provides solid data support for subsequent impact coefficient calculation, weight allocation, and health scoring through high-quality, highly correlated standardized data output. This further ensures the scientificity and accuracy of the entire system's health assessment method and effectively avoids the assessment bias problem caused by the extensive process of traditional data processing.

[0063] Example 4: In scenarios such as online finance, where business continuity and value assurance are extremely critical, there may be a disconnect between normal technical indicators and abnormal core business indicators. For example, some technical components may be running smoothly, but the transaction success rate may decline significantly. To address the pain point of traditional scoring being out of sync with actual business performance, leading to misjudgments of business risks by the operations team, this embodiment proposes a health score calibration method based on the actual performance of business indicators. By acquiring business indicator data from standardized data, the deviation from normal thresholds is quantified, and the initial health score is adjusted accordingly. This ensures that the final health score aligns with both the technical operating status and the realization of business value, truly reflecting the system's actual support capability for the business.

[0064] Specifically, after weighting the scores of each individual indicator according to the scoring weights in step 105, the following steps can be further performed: Step 110: Use the weighted calculation result as the preliminary health score.

[0065] Following the previous calculation of individual indicator scores based on real-time technical data, and the weighted summation according to the scoring weights, this preliminary health score is defined as the initial health score. This preliminary score provides a comprehensive quantitative evaluation of the operational status of each component of the system from a technical perspective, fully preserving the weight-oriented characteristics of the technical indicators, and providing a basic benchmark for subsequent score optimization incorporating business dimensions.

[0066] Step 111: Obtain business indicator data from the standardized data.

[0067] Extract business metrics data that have been formatted and time-aligned, such as transaction success rate, order flow, and payment time, from the standardized data generated in step 102.

[0068] Step 112: Calculate the direction and degree of deviation between the business indicator data and the corresponding normal threshold.

[0069] For each extracted business indicator data, its preset normal operating threshold is first retrieved, such as the normal threshold for transaction success rate being 99.9%. Then, a quantitative algorithm is used to calculate the difference between the current business indicator data and the normal threshold.

[0070] Among them, deviation direction refers to whether the business indicator data is higher or lower than the normal threshold. For example, a transaction success rate of 98% is lower than the threshold, which is a negative deviation and represents business abnormality; a rate higher than the threshold is a positive deviation and represents business performance better than expected. Deviation degree refers to the quantitative value of the difference. For example, a transaction success rate that is 1.9% lower than the threshold, or an order volume that is 500 transactions / hour lower than the threshold, ultimately forming a quantitative description of the actual operating status of the business.

[0071] Step 113: Adjust the initial health score based on the direction and degree of deviation.

[0072] Based on the deviation results of the business indicators calculated in step 112, targeted scoring adjustment logic is executed. By examining the actual performance in the business dimension, the limitations of purely technical dimension scoring are corrected, achieving a dual quantitative fusion of technical status and business performance.

[0073] In this embodiment, the specific scoring adjustment logic is not limited. For example, it can be divided into three levels: mild, moderate, and severe, corresponding to score adjustment ranges of 5%, 15%, and 30%, respectively. Specifically, it can be executed according to the following logic: if the deviation direction shows that the business indicator data is lower than the corresponding normal threshold, the initial health score is adjusted downward according to the degree of deviation; if the deviation direction shows that the business indicator data is higher than the corresponding normal threshold, the initial health score is adjusted upward according to the degree of deviation.

[0074] If business metrics deviate negatively (business anomaly), the initial health score will be adjusted downwards according to the degree of deviation. The greater the deviation, the greater the downward adjustment. For example, if the transaction success rate drops significantly, the score should be significantly reduced even if the technical indicators are normal. If business metrics deviate positively (business performance is excellent), the initial score can be slightly adjusted upwards according to the degree of deviation to reflect the system's excess support for business value. If business metrics do not deviate significantly (within the normal threshold range), the initial health score can remain unchanged.

[0075] This scoring adjustment logic ensures that the initial health score is accurately calibrated based on actual business performance. It highlights abnormal business risks by lowering the score due to negative deviations and reflects the system's excess support value by raising the score due to positive deviations. It also avoids scoring distortion caused by a single business indicator or extreme deviations, ensuring that the final health score closely reflects the realized business value and provides a more realistic and reliable reference for operational decisions. It should be noted that this embodiment only uses the above scoring adjustment logic as an example, but it is not limited to this. Corresponding configurations can be made according to the needs of actual application scenarios, which will not be elaborated further here.

[0076] In step 105, the weighted calculation result is used as the system health score, which is then adjusted to: the adjusted score is used as the system health score.

[0077] Based on the above introduction, the scoring optimization method provided in this embodiment breaks through the limitations of purely technical indicator weighted scoring. By introducing actual operational data of business indicators for secondary calibration, it solves the disconnect problem in traditional scoring where technical performance is normal but business anomalies have occurred. The calibration process is based on business indicators in standardized data, quantifying the direction and degree of deviation from normal thresholds to provide clear quantitative support for scoring adjustments, ensuring the scientific and accurate nature of the adjustment logic. Simultaneously, this method employs a dual-process design of preliminary scoring based on technical weights + dynamic calibration of abnormal business states. This ensures that the final health score fully retains the weighted characteristics of technical indicators while accurately mapping the actual realization of business value, significantly improving the reliability and business relevance of the scoring. The calibrated score more realistically reflects the system's actual support capability for business operations, helping the operations and maintenance team quickly identify potential risks at the business level, clarify the business priorities for fault handling, avoid misallocation of operations and maintenance resources, and provide a more practical basis for intelligent operations and maintenance decisions.

[0078] Example 5: Traditional scoring systems only output results, failing to identify which core technical indicators caused the score fluctuations or quantify the specific impact of these anomalies on business metrics. This forces operations teams to blindly troubleshoot among massive amounts of technical indicators, resulting in low fault location efficiency and insufficient business risk prediction. To further address these issues, this embodiment proposes, after obtaining the system health score, to further establish a traceability chain from score changes to core technical indicators to business impact prediction, fulfilling the practical need for operations to quickly locate faults and accurately assess risks.

[0079] Based on Example 4, after adjusting the preliminary health score according to the direction and degree of deviation in step 113, the following steps can be further performed: Step 114: Based on the influence matrix, reverse-locate the core technical indicators that lead to changes in the initial health score.

[0080] Based on the generated technology-business impact matrix, a reverse tracing logic from result to cause is initiated. The impact matrix has quantified and stored the intensity (impact coefficient) of the impact of each technical indicator on each business indicator, while the change in the initial health score is essentially a comprehensive reflection of abnormal technical indicators or fluctuations in business indicators.

[0081] By correlating the magnitude of score changes with the corresponding fluctuation range of business indicators, technical indicators with large absolute values ​​of influence coefficients and strong correlation with score changes are selected from the influence matrix. This allows for the precise identification of key technical nodes, i.e. core technical indicators, that play a dominant role in score fluctuations, avoiding blindly searching through a massive number of technical indicators and enabling targeted location of the root cause of the fault.

[0082] The specific implementation method for locating core technical indicators is not limited in this embodiment. For example, technical indicators with the top N absolute values ​​of their impact coefficients can be directly selected as core indicators, or corresponding technical indicators can be screened according to business modules, or a comprehensive judgment can be made by combining the frequency of indicator anomalies with the impact coefficient. To solve the problems of traditional positioning methods that easily overlook the chain effects between technical indicators and the incomplete screening of core nodes, and to ensure accurate identification of all technical nodes that play a key role in changes in health scores, this embodiment further proposes a positioning method, which specifically includes the following steps: Step 1141: Select technical indicators in the influence matrix whose absolute value of the influence coefficient is greater than the core indicator threshold as preliminary indicators.

[0083] Based on the influence matrix, and by setting preset core indicator thresholds, technical indicators whose influence on business indicators reaches the threshold standard are selected and defined as preliminary indicators, thus initially identifying technical nodes with significant business impact potential. The specific numerical values ​​of the core indicator thresholds are not limited in this embodiment.

[0084] Step 1142: Sort the preliminary indicators in descending order of the absolute value of the influence coefficient, and select the top N indicators.

[0085] The initial indicators are sorted in descending order of their impact intensity (absolute value of the impact coefficient). Based on a preset positive integer N, the top N indicators are selected to further focus on the core candidate indicators with the most prominent impact.

[0086] Step 1143: Determine the chain reaction path between the technical indicators, and determine all technical indicators associated with the first N indicators based on the chain reaction path.

[0087] By exploring the chain reaction relationships between technical indicators, a complete influence path can be formed, and then all technical indicators that have a transmission relationship with the first N indicators can be identified, so as to avoid missing key nodes in the chain reaction.

[0088] Step 1144: Take the first N indicators and their corresponding related technical indicators as the core technical indicators.

[0089] By integrating the top N high-impact indicators and their related indicators, the scope of the core technical indicators is finally clarified, ensuring that the core technical indicators cover both directly high-impact nodes and key related nodes in the chain of impacts, providing a comprehensive basis for tracing the root causes of subsequent scoring changes.

[0090] This method initially identifies high-impact technical indicators by screening core indicator thresholds to avoid interference from irrelevant indicators; it selects the top N indicators by ranking them by impact coefficient to quickly focus on key core nodes; it mines related indicators by combining the chain influence paths between technical indicators to make up for the shortcomings of traditional positioning that miss the chain root causes; finally, it integrates high-impact indicators and related indicators to ensure the relevance of core technical indicators and achieve comprehensive coverage of the influence chain, providing a scientific and reliable basis for accurately tracing the root causes of score changes and efficiently troubleshooting faults.

[0091] It should be noted that this embodiment only uses the determination method of the above core technical indicators as an example for detailed introduction. Other determination methods can refer to the introduction of this embodiment, and will not be repeated here.

[0092] Step 115: Extract the influence coefficients and influence directions corresponding to the core technical indicators.

[0093] For the identified core technical indicators, two types of key quantitative information are extracted from the influence matrix and business influence knowledge base. The first is the influence coefficient, whose absolute value directly reflects the strength of the influence of the core technical indicator on the business indicator. The second is the direction of influence, which clarifies the correlation logic based on the positive or negative value of the influence coefficient. For example, a positive coefficient may indicate that an increase in database query time will lead to a decrease in transaction success rate, while a negative coefficient may indicate that an improvement in the frequency of interface calls will drive a reduction in payment time.

[0094] Step 116: Estimate the change in business indicators when core technical indicators are abnormal, and obtain the estimated value of business indicator changes.

[0095] By combining the actual degree of anomaly of the core technical indicators (such as the value of deviation from the normal operating threshold and the fluctuation range) with the impact coefficient extracted in step 115, the specific change range of business indicators that may be caused by the anomaly of the technical indicator is estimated through the preset quantitative calculation logic (such as the degree of anomaly × the impact coefficient), i.e. the estimated value of the change in business indicators.

[0096] For example, if the core technical indicator, the payment interface response time, increases by 50ms compared to the normal threshold, the corresponding impact coefficient on the transaction success rate is -0.03 (negative impact). It can be estimated that the transaction success rate will decrease by 1.5%. This estimate directly quantifies the actual impact of technical anomalies on business value.

[0097] Step 117: Collect the names, impact coefficients, impact directions, and estimated changes in business indicators corresponding to the core technical indicators, and generate an impact factor labeling report.

[0098] The key information scattered in the previous steps (names of core technical indicators, impact coefficients, impact directions, and estimated changes in business indicators) is systematically integrated and organized into an impact factor labeling report according to a preset structured format.

[0099] The report presents the technical indicators that caused the score fluctuations to the operations and maintenance team in an intuitive and standardized manner, as well as the intensity and direction of the impact of these indicators on the business and the specific losses that the business may suffer. It provides comprehensive and accurate decision support for the operations and maintenance team to quickly determine the priority of faults and formulate targeted handling plans.

[0100] Based on the above introduction, the method provided in this embodiment relies on the generated technology-business impact matrix to reverse locate core technical indicators, accurately pinpoint the key technical nodes that cause score changes, and significantly shorten the fault location time. In addition, by extracting the impact coefficient and direction of the core technical indicators and estimating the estimated value of business indicator changes, the abstract score fluctuations are transformed into a concrete quantitative correlation between technical anomalies and business impacts. This allows the operations and maintenance team to clearly understand the degree and direction of the impact of technical faults on the business, solving the problem that traditional operations and maintenance cannot predict business risks. This method integrates the names of core technical indicators, impact coefficients, impact directions, and estimated values ​​of business changes to generate an impact factor labeling report, providing the operations and maintenance team with a structured and visualized decision-making basis, helping to quickly formulate targeted handling plans, avoid resource misallocation, and significantly improve the accuracy and efficiency of operations and maintenance decisions.

[0101] Example 6: This embodiment relates to a system health assessment device. A schematic diagram of the system health assessment device in this embodiment is shown below. Figure 2 As shown, it includes: a multi-source acquisition unit 201, a data processing unit 202, a technology and business impact assessment unit 203, a weight allocation unit 204, and a scoring calculation unit 205.

[0102] Among them, the multi-source acquisition unit 201 is used to collect multi-source performance data of the target system as raw data; the multi-source performance data includes: infrastructure performance data, application performance data and business indicator data. The data processing unit 202 is used to standardize the raw data and then organize the data according to the dimensions of technical indicators and business indicators to generate standardized data. The standardization process includes: format standardization and time dimension alignment. The technology and business impact assessment unit 203 is used to call the business impact quantification model to calculate the impact coefficient of each technical indicator on each business indicator in the standardized data and generate an impact matrix. The weight allocation unit 204 is used to allocate the scoring weights of each technical indicator according to the influence matrix. The scoring calculation unit 205 is used to calculate the score of each single indicator based on the real-time technical indicator data in the standardized data, and to perform weighted calculation on the scores of each single indicator according to the scoring weight, and use the weighted calculation result as the system health score.

[0103] In the system health assessment device provided in this embodiment, the multi-source acquisition unit comprehensively covers the core performance data of infrastructure, application layer, and business system; secondly, the data processing unit performs refined processing on the raw data through format standardization, time dimension alignment, and technical-business dimension organization, effectively removing data impurities, unifying data specifications, and strengthening time-series correlation; the technical business impact assessment unit calculates the impact coefficient and generates the impact matrix with the help of the business impact quantification model, realizing the accurate quantification of the correlation between technical indicators and business indicators, breaking through the limitation of traditional devices that can only assess technical status and cannot quantify business impact; the weight allocation unit dynamically allocates scoring weights based on the impact matrix, replacing the traditional fixed weight mode, so that the scoring strategy fits the real-time business needs, and the final output health score can not only comprehensively reflect the technical operation status of the system, but also accurately map the actual impact on business value, which can effectively support the operation and maintenance team to quickly judge the fault priority and optimize resource allocation.

[0104] Additionally, it should be noted that all modules involved in this embodiment are logical modules. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units.

[0105] To enhance understanding, this embodiment introduces another unit configuration for a system health assessment device, comprising three core units: a multi-source data acquisition and business correlation unit, a business impact dynamic modeling and analysis unit, and an adaptive health scoring and output unit. (See [link to documentation]). Figure 3 As shown.

[0106] Among them, the multi-source data acquisition and business association unit comprehensively covers the core data of the infrastructure layer, application layer and business system through an extensible access framework. It collects key indicators such as server CPU utilization, interface response time and transaction success rate. After standardized processing such as cleaning, noise reduction, format unification and time dimension alignment, the heterogeneous data is integrated into a unified central data lake, providing high-quality and integrated data support for subsequent analysis.

[0107] As the core of the device's intelligent decision-making, the Business Impact Dynamic Modeling and Analysis Unit uses machine learning technology to mine the linear and non-linear correlations and chain impact paths between technical indicators and business indicators based on historical and real-time data in the data lake. It constructs an initial business impact quantification model and dynamically optimizes it through continuous online learning. It accurately calculates the impact coefficients of each technical indicator on the business indicator, forms a technology-business impact matrix, and deposits it into the business impact knowledge base, thus completing the transformation from data to knowledge.

[0108] The adaptive health score and output unit serves as the value realization terminal of the device. It dynamically allocates the scoring weights of technical indicators based on the quantitative relationships in the knowledge base. After obtaining the initial score by weighted aggregation of the scores of individual indicators, it is calibrated in combination with the actual performance of business indicators to avoid the disconnect between technology and business. Finally, the health score and core influencing factors are visualized and output through channels such as operation and maintenance monitoring dashboards and intelligent alarms. At the same time, operation and maintenance handling feedback is recorded to provide data support for continuous model optimization and realize a closed loop from analysis to action.

[0109] This embodiment uses the unit configuration of the above two devices as an example for introduction. Due to different configuration requirements, other unit configurations can also be performed, and all can refer to the introduction of this embodiment, which will not be repeated here.

[0110] Furthermore, in order to highlight the innovative aspects of this application, no units that are not closely related to solving the technical problems proposed in this application are introduced in this embodiment, but this does not mean that there are no other units in this embodiment.

[0111] Example 7: Another embodiment of this application relates to an electronic device, such as... Figure 4As shown, it includes: at least one processor 301; and a memory 302 communicatively connected to at least one processor 301; wherein the memory 302 stores instructions executable by at least one processor 301, the instructions being executed by at least one processor 301 to enable at least one processor 301 to perform the steps of the system health assessment method in the above embodiments.

[0112] The memory and processor are connected via a bus, which can include any number of interconnecting buses and bridges, connecting various circuits of one or more processors and memories. The bus can also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and will not be described further herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by the processor is transmitted over the wireless medium via an antenna, which further receives data and transmits it to the processor.

[0113] The processor manages the bus and general processing, and also provides various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory is used to store data used by the processor during operation.

[0114] Those skilled in the art will understand that the above embodiments are specific embodiments for implementing this application, and in practical applications, various changes can be made to them in form and detail without departing from the spirit and scope of this application.

Claims

1. A method for assessing system health, characterized in that, include: Collect multi-source performance data of the target system as raw data; The multi-source performance data includes: infrastructure performance data, application performance data, and business indicator data; After standardizing the raw data, the data is organized according to the dimension of technical indicators and business indicators to generate standardized data; the standardization process includes: format standardization and time dimension alignment; The business impact quantification model is invoked to calculate the impact coefficient of each technical indicator in the standardized data on each business indicator, and an impact matrix is ​​generated statistically. Based on the influence matrix, assign scoring weights to each of the technical indicators; Based on the real-time technical indicator data in the standardized data, the scores of each individual indicator are calculated, and the scores of each individual indicator are weighted according to the scoring weight. The weighted result is used as the system health score.

2. The system health assessment method according to claim 1, characterized in that, The method for generating the business impact quantification model includes: Machine learning algorithms are used to extract historical standardized data from the data lake and analyze the correlation between fluctuations in technical indicators and changes in business indicators. Based on the aforementioned relationship, the calculation logic for the changes in technical indicators and business indicators is defined, and an initial model is constructed. The standardized data is stored in real time in the data lake; Periodically extract real-time incremental data from the data lake, and update the model parameters in the initial model based on the real-time incremental data. Use the parameter-optimized initial model as the business impact quantification model.

3. The system health assessment method according to claim 1, characterized in that, After standardizing the raw data, the data is organized according to the dimension of technical indicators - business indicators, including: The original data is subjected to invalid data removal and noise data filtering, and the generated data is used as preprocessed data. The preprocessed data is converted according to a preset standardized format and sorted by time to generate time series data; The technical indicator data and business indicator data within the same time window in the time series data are correlated and stored according to the correlation dimension of technical indicator-business indicator, which serves as the standardized data.

4. The system health assessment method according to claim 1, characterized in that, The multi-source performance data of the target system being collected includes: The system collects server CPU utilization, memory utilization, network throughput, database query time, and middleware service status at the infrastructure layer and technical component layer as infrastructure performance data. The response time, error rate, and call frequency of key interfaces in the application layer are collected as application performance data. The transaction success rate, order volume, payment time, and user activity of the business system are collected as the business indicator data.

5. The method for assessing system health according to any one of claims 1 to 4, characterized in that, After weighting the scores of each individual indicator according to the aforementioned scoring weights, the method further includes: The weighted calculation result will be used as the preliminary health score; Obtain the business indicator data from the standardized data; Calculate the direction and degree of deviation between the business indicator data and the corresponding normal threshold; The preliminary health score is adjusted based on the direction and degree of deviation. The statement that the weighted calculation result is used as the system health score specifically means that the adjusted score is used as the system health score.

6. The system health assessment method according to claim 5, characterized in that, Adjusting the preliminary health score based on the direction and degree of deviation includes: If the deviation direction shows that the business indicator data is lower than the corresponding normal threshold, the preliminary health score will be adjusted downward according to the degree of deviation. If the deviation direction shows that the business indicator data is higher than the corresponding normal threshold, the preliminary health score will be adjusted upwards according to the degree of deviation.

7. The system health assessment method according to claim 5, characterized in that, After adjusting the preliminary health score based on the direction and degree of deviation, the method further includes: Based on the aforementioned influence matrix, the core technical indicators that cause the change in the initial health score are identified in reverse. Extract the influence coefficients and influence directions corresponding to the core technical indicators; the influence direction is determined based on the positive or negative value of the influence coefficients. Estimate the change in business indicators when the core technical indicators are abnormal, and obtain the estimated value of the change in business indicators; Collect the names, impact coefficients, impact directions, and estimated changes in the business indicators corresponding to the core technical indicators, and generate an impact factor labeling report.

8. The system health assessment method according to claim 7, characterized in that, Based on the aforementioned influence matrix, the core technical indicators leading to the change in the initial health score are identified in reverse, including: Technical indicators whose absolute values ​​of influence coefficients in the influence matrix are greater than the core indicator threshold are selected as preliminary indicators. The preliminary selection indicators are sorted in descending order of the absolute value of their influence coefficients, and the top N indicators are selected; where N is a preset positive integer. Determine the chain reaction paths between the technical indicators, and determine all technical indicators associated with the first N indicators based on the chain reaction paths; The first N indicators and their corresponding related technical indicators are taken as the core technical indicators.

9. A system health assessment device, characterized in that, include: The multi-source acquisition unit is used to acquire multi-source performance data of the target system as raw data. The multi-source performance data includes: infrastructure performance data, application performance data, and business indicator data; The data processing unit is used to standardize the raw data and then organize the data according to the dimension of technical indicators and business indicators to generate standardized data; the standardization process includes: format standardization and time dimension alignment. The technology and business impact assessment unit is used to call the business impact quantification model to calculate the impact coefficient of each technical indicator on each business indicator in the standardized data, and to generate an impact matrix. The weight allocation unit is used to allocate the scoring weights of each of the technical indicators according to the influence matrix. The scoring calculation unit is used to calculate the individual indicator scores based on the real-time technical indicator data in the standardized data, and to perform weighted calculation on each individual indicator score according to the scoring weight, and use the weighted calculation result as the system health score.

10. An electronic device, characterized in that, include: At least one processor; And, a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the system health assessment method as described in any one of claims 1 to 8.

Images