IT resource health degree assessment method and system

Abstract

The invention relates to the technical field of IT resource health degree assessment, in particular to an IT resource health degree assessment method and system, and the method comprises the following steps: obtaining the number of requests of each business activity type; calculating the real-time proportion of each service activity type in the total amount of all service requests; when the real-time proportion deviates from the conventional service composition reference, identifying and quantifying a service composition abnormal event, and generating service composition abnormal event information; searching a matched adjustment strategy from a preset weight adjustment rule, wherein the adjustment strategy comprises a trigger condition, an affected index and a weight adjustment mode; calculating and applying a new weight of each business index in health degree evaluation; calculating an IT resource health score by using the new weight; and outputting the IT resource health score and the business composition abnormal event information. The accuracy and timeliness of evaluation are improved.

Description

Technical Field

[0001] This invention relates to the technical field of IT resource health assessment, and specifically to an IT resource health assessment method and system. Background Technology

[0002] In modern enterprise information technology (IT) operations, an IT resource health assessment system is typically deployed to ensure the stable operation of systems supporting core business functions. These systems aim to provide a comprehensive view of the system's health status by comprehensively analyzing various IT resource and business metrics. However, with rapid changes in the market environment and frequent adjustments to business strategies, traditional health assessment methods often struggle to keep pace with these changes. When business priorities or user behavior patterns undergo drastic and unpredictable shifts, the assessment system may fail to update its internal rules in a timely manner, leading to biased judgments about the system's true health status and even delays in responding to critical issues, ultimately causing business losses. Summary of the Invention

[0003] The purpose of this invention is to address the aforementioned shortcomings by proposing a method and system for assessing the health of IT resources.

[0004] The present invention adopts the following technical solution: An IT resource health assessment method, comprising the following steps: At the IT system entry point, business activity characteristics are extracted from incoming business requests, and the business activity type is identified according to preset rules to obtain the number of requests for each business activity type. Calculate the real-time percentage of each business activity type in the total number of business requests based on the number of requests for each business activity type. The real-time percentage is compared with the preset benchmark for regular business composition. When the real-time percentage deviates from the benchmark for regular business composition, abnormal business composition events are identified and quantified, and abnormal business composition event information is generated. Based on the abnormal events in the business, the matching adjustment strategy is searched from the preset weight adjustment rules. The adjustment strategy includes the triggering conditions, the affected indicators, and the weight adjustment method. Based on the matching adjustment strategy, calculate and apply new weights for each business indicator in the health assessment; The new weights are used to calculate the IT resource health score; Output IT resource health scores and information on abnormal events in business operations.

[0005] Through this technical solution, this application can dynamically adjust the weights of various business indicators in the IT resource health assessment, enabling it to adapt to rapid changes in business composition and user behavior patterns, thereby improving the accuracy and timeliness of the assessment.

[0006] This application also discloses an IT resource health assessment system applied to the aforementioned IT resource health assessment method. The system includes: The parsing module extracts business activity characteristics from incoming business requests at the IT system entry point, identifies the business activity type according to preset rules, and obtains the number of requests for each business activity type. The percentage calculation module calculates the real-time percentage of each business activity type in the total number of business requests based on the number of requests for each business activity type. The processing module compares the real-time percentage with the preset benchmark for regular business composition. When the real-time percentage deviates from the benchmark for regular business composition, it identifies and quantifies abnormal business composition events and generates abnormal business composition event information. The matching module searches for matching adjustment strategies from preset weight adjustment rules based on the abnormal events constituted by the business. The adjustment strategy includes the triggering conditions, the affected indicators, and the weight adjustment method. The update module calculates and applies new weights to various business indicators in the health assessment based on the matching adjustment strategy. The weighting calculation module uses new weights to calculate the IT resource health score; The output module outputs IT resource health scores and information on abnormal events in business operations.

[0007] This application provides a system for implementing an IT resource health assessment method. Through modular design, it can efficiently execute the aforementioned assessment method, thereby providing strong technical support for the stable operation of IT systems.

[0008] The IT resource health assessment in this application can more accurately reflect the true health status of the system, promptly identify and alert on critical issues, avoid business losses caused by assessment bias, and significantly improve the response speed and efficiency of IT operations.

[0009] To further understand the features and technical content of the present invention, please refer to the following detailed description and drawings of the present invention. However, the drawings provided are for reference and illustration only and are not intended to limit the present invention. Attached Figure Description

[0010] Figure 1 This is a flowchart of an IT resource health assessment method according to the present invention; Figure 2This is a schematic diagram of the structure of an IT resource health assessment system according to the present invention. Detailed Implementation

[0011] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can understand the advantages and effects of the present invention from the content disclosed in this specification. The present invention can be implemented or applied through other different specific embodiments, and various details in this specification can also be modified and changed based on different viewpoints and applications without departing from the spirit of the present invention. Furthermore, the accompanying drawings of the present invention are for simple illustrative purposes only and are not depictions of actual dimensions; this is stated in advance. The following embodiments will further describe the relevant technical content of the present invention in detail, but the disclosed content is not intended to limit the scope of protection of the present invention.

[0012] This embodiment provides a method and system for assessing the health of IT resources, combined with... Figure 1 and Figure 2 As shown.

[0013] refer to Figure 1 An IT resource health assessment method, comprising the following steps: At the IT system entry point, business activity characteristics are extracted from incoming business requests, and the business activity type is identified according to preset rules to obtain the number of requests for each business activity type. Calculate the real-time percentage of each business activity type in the total number of business requests based on the number of requests for each business activity type. The real-time percentage is compared with the preset benchmark for regular business composition. When the real-time percentage deviates from the benchmark for regular business composition, abnormal business composition events are identified and quantified, and abnormal business composition event information is generated. Based on the abnormal events in the business, the matching adjustment strategy is searched from the preset weight adjustment rules. The adjustment strategy includes the triggering conditions, the affected indicators, and the weight adjustment method. Based on the matching adjustment strategy, calculate and apply new weights for each business indicator in the health assessment; The new weights are used to calculate the IT resource health score; Output IT resource health scores and information on abnormal events in business operations.

[0014] The IT system entry point refers to the first point of contact for a business request entering the IT system, such as a load balancer, API gateway, or message queue. Business activity characteristics refer to information in the business request that can distinguish different types of business activities, such as URL path, request parameters, user ID, and session ID. Preset rules refer to the set of rules used to identify business activity types; these rules can be defined based on regular expressions, pattern matching, or machine learning models. The normal business composition baseline refers to the distribution of each business activity type within the total number of business requests under normal business operation conditions; this baseline can be set through historical data analysis or expert experience. Weight adjustment rules refer to the set of strategies that dynamically adjust the weights of various business indicators in the health assessment based on the type and severity of abnormal events in the business composition. These strategies can be predefined or adaptively learned through machine learning models.

[0015] First, at the IT system entry point, it's necessary to extract business activity characteristics from incoming business requests and identify the business activity type according to preset rules, thus obtaining the request count for each business activity type. For example, a probe deployed on an API gateway can capture all incoming HTTP requests in real time. For each request, its URL path, HTTP method, and User-Agent header can be extracted as business activity characteristics. Preset rules can be defined as follows: if the URL path contains " / user / register", it's identified as a "user registration" business activity type; if the URL path contains " / product / view", it's identified as a "product browsing" business activity type. The system will count the request count for each business activity type in real time. Another implementation method is to listen for business events in a message queue, parse the business type identifier from the event message body, and use this to identify the business activity type and count the request count.

[0016] Secondly, based on the number of requests for each business activity type, calculate the real-time percentage of each business activity type in the total number of business requests. For example, if within a time window, the number of "user registration" requests is 1000, the number of "product browsing" requests is 9000, and the total number of all business requests is 10000, then the real-time percentage of "user registration" is 10%, and the real-time percentage of "product browsing" is 90%. This calculation can be performed periodically, for example, once per minute, to reflect real-time changes in the business composition.

[0017] Next, the real-time percentage is compared with a preset benchmark for regular business composition. When the real-time percentage deviates from the benchmark, anomaly events in business composition are identified and quantified, generating anomaly event information. For example, in the benchmark, the percentage of "user registration" is typically 5%, while the current real-time percentage reaches 10%, indicating a deviation in the "user registration" business composition. The deviation can be quantified by calculating the absolute or relative difference between the real-time percentage and the benchmark percentage. For example, if the benchmark percentage of "user registration" is 5% and the real-time percentage is 10%, the deviation quantification result could be a relative increase of 5 percentage points or 100%. The generated anomaly event information can include the type of anomaly (e.g., "abnormally high increase in user registration percentage"), the degree of deviation, and the time of occurrence.

[0018] Then, based on the abnormal business event, the system searches for a matching adjustment strategy from the preset weight adjustment rules. The adjustment strategy includes the triggering conditions, affected metrics, and weight adjustment method. For example, if an abnormally high user registration rate is detected, the system will search for weight adjustment rules. A matching adjustment strategy might be: trigger this strategy when the "user registration" business rate exceeds 50% of the normal benchmark; the affected metrics are "user registration success rate" and "user registration response time"; the weight adjustment method is to increase the weight of "user registration success rate" by 20% and the weight of "user registration response time" by 15%.

[0019] Subsequently, based on the matching adjustment strategy, new weights for each business indicator in the health assessment are calculated and applied. For example, if the original weight of "user registration success rate" was 0.1, according to the above adjustment strategy, its new weight will become 0.1*(1+0.2)=0.12. These new weights will be applied to subsequent health score calculations.

[0020] Next, the IT resource health score is calculated using the new weights. For example, the IT resource health score can be obtained by multiplying the current values ​​of each business indicator by their corresponding new weights and then summing them. If the current value of "user registration success rate" is 90%, the current value of "user registration response time" is 200ms, and other indicators also have corresponding current values ​​and new weights, then the final IT resource health score will be the sum of all these weighted values.

[0021] Finally, the system outputs the IT resource health score and information on abnormal business events. The output can be displayed as a dashboard or sent via an alert system. For example, the system could display that the current IT resource health score is 85, and simultaneously show "Abnormal Business Event: User registration percentage is abnormally high, deviation rate 100%".

[0022] The above IT resource health assessment method also includes the following steps: Get a list of currently active marketing campaigns, preset business objective tags, and core business objectives set by business departments; The business objective tags for active marketing campaigns are derived based on the current list of active marketing campaigns and preset business objective tags. Based on real-time percentages, active marketing campaign business target tags, and core business objectives set by business departments, comprehensive business intents are inferred, and priority scores are assigned to each inferred comprehensive business intent. Establish a resource consumption profile for each type of business activity on shared basic services, and collect details of the internal operation of shared basic services; Based on the resource consumption characteristics profile of shared basic services for each type of business activity, the resource consumption characteristics of the business activities are derived. Based on the priority score of comprehensive business intent, real-time proportion, resource consumption characteristics of business activities, and internal operation details of shared basic services, calculate the business impact value corresponding to each type of business activity; The degree of deviation of abnormal events in business composition is determined by the difference between the real-time percentage and the baseline of regular business composition. The deviation degree and business impact value of the integrated business constitute anomaly events are combined to obtain the integrated anomaly event quantification result. The integrated anomaly event quantification result is used to further quantify the business constitute anomaly events and generate business constitute anomaly event information.

[0023] Specifically, when acquiring the list of currently active marketing activities, preset business objective tags, and core business objectives set by business departments, the list of currently active marketing activities refers to ongoing or planned marketing or operational activities that may have a significant impact on IT system traffic and business operations. Preset business objective tags can be understood as predefined categories or identifiers of business objectives associated with different types of business activities, such as "user growth," "order conversion," and "brand exposure." Core business objectives set by business departments refer to the business metrics or strategic goals that the company prioritizes and focuses on during a specific period, such as "increasing order volume by 50% during the Double Eleven promotion." Acquiring this information aims to provide foundational data for subsequent business intent inference.

[0024] Specifically, the business objective tags for active marketing campaigns are derived from the current list of active campaigns and preset business objective tags. This involves matching or associating active marketing campaigns with preset business objective tags to clarify the business objectives of each active marketing campaign. For example, a "new user registration rewards points" marketing campaign might be tagged with the business objective tags of "user growth" and "increased user activity."

[0025] In practical applications, comprehensive business intent is inferred based on real-time percentage, active marketing campaign business objective tags, and core business objectives set by business departments. A priority score is then assigned to each inferred comprehensive business intent. Comprehensive business intent can be understood as the overall business goal or purpose undertaken by the current IT system, comprised of multiple business activities. The inference process can utilize machine learning models or rule-based expert systems, combining real-time business traffic composition (real-time percentage), the business objectives of the current marketing campaign, and the core business objectives set by the enterprise to identify the primary driving intent behind the current business traffic. For example, if the real-time percentage shows a surge in registration and login traffic, and there is a "new user registration points" marketing campaign, while the core business objective is "user growth," then the current comprehensive business intent can be inferred to be "high-priority user growth." Priority scores are used to quantify the importance of different business intents and are typically set by business departments based on strategic priorities.

[0026] Furthermore, a resource consumption characteristic profile for each type of business activity on shared basic services is established, and internal operational details of shared basic services are collected. The resource consumption characteristic profile records the usage patterns and quantitative indicators of resources such as CPU, memory, network bandwidth, and IOPS for various shared basic services (e.g., databases, caches, message queues, microservices) under normal operating conditions for a specific business activity type. This profile can be constructed through historical data analysis, benchmark testing, or performance monitoring data. Internal operational details of shared basic services include real-time resource utilization, internal queue length, response time, error rate, and concurrent connections. The collection of these details aims to reflect the real-time health status and capacity of shared basic services.

[0027] Therefore, based on the resource consumption characteristics profile of each type of business activity for shared basic services, the resource consumption characteristics of each business activity are derived. Specifically, this means predicting or calculating the expected resource consumption of a particular business activity for a specific shared basic service based on the established profile and the actual request volume and type of the current business activity.

[0028] Based on this, the business impact value for each type of business activity is calculated according to the priority score of the comprehensive business intent, its real-time proportion, the resource consumption characteristics of the business activity, and the internal operational details of the shared basic services. The business impact value is a quantitative indicator used to measure the potential negative impact on overall business objectives and IT system stability when a certain type of business activity experiences anomalies or performance degradation. Its calculation comprehensively considers the priority of the business intent carried by the activity, its proportion in the overall traffic, its resource consumption on the shared basic services, and the current operational status of the shared basic services. For example, a core business activity under a high-priority business intent, even if its real-time proportion is not high, will still have a higher business impact value if it significantly consumes resources from a shared basic service that is already under high load.

[0029] Simultaneously, based on the difference between the real-time percentage and the baseline of normal business composition, the degree of deviation of abnormal events in business composition is determined. The degree of deviation can be calculated using various statistical methods, such as absolute difference, relative difference, and standard deviation multiple, aiming to quantify the magnitude of the difference between the current business composition and the normal baseline.

[0030] Ultimately, the deviation degree and business impact value of anomaly events are combined to obtain the quantified anomaly event result. The fusion process can employ weighted averaging, product, or other composite functions to comprehensively consider both deviation degree and business impact value. For example, an anomaly event with a high deviation degree but a low business impact value may have a lower quantified result than an anomaly event with a moderate deviation degree but an extremely high business impact value. The quantified anomaly event result is used to further quantify business-related anomaly events and generate business-related anomaly event information, thereby providing more insightful anomaly event reports.

[0031] This application's solution effectively addresses the insufficient quantification depth of abnormal business composition events in the basic solution by introducing business intent inference and business impact value calculation. Specifically, by acquiring a list of active marketing activities, preset business goal tags, and core business objectives, the system can more accurately understand the true intent behind current business traffic and assign priorities to these intents. This allows the assessment of deviations in business composition to move beyond simply focusing on changes in traffic share; it enables the differentiation between normal fluctuations consistent with high-priority business objectives and abnormal situations that may indicate risks.

[0032] Furthermore, by establishing resource consumption characteristic profiles for shared infrastructure services based on business activity types and collecting internal operational details of these services, this solution can deeply analyze the actual dependence and potential contention of business activities on underlying IT resources. When a deviation occurs in the business composition, the system can not only identify the deviation itself but also calculate the potential business impact by combining the priority of the business intent, its real-time proportion, the resource consumption characteristics of the business activity, and the real-time operational status of the shared infrastructure services. For example, an abnormal increase in traffic from a low-priority business activity, if it significantly puts pressure on critical shared services, will have its business impact amplified, thus receiving due attention in the health assessment.

[0033] Ultimately, by integrating the deviation of business-related anomalies with the calculated business impact values, the quantitative results of anomalies are made more comprehensive and accurate. This integration mechanism ensures that IT resource health assessments not only identify "what" the anomalies occurred, but also assess "why" the anomalies occurred and "how significant" the impact was, thus providing more instructive input for subsequent weight adjustments and health score calculations.

[0034] In some preferred embodiments, it is assumed that an e-commerce platform needs to assess the health of its IT system during the "Double Eleven" promotion.

[0035] First, the system retrieves a list of currently active marketing activities, such as "flash sales," "discount promotions," and "new user registration bonuses." Simultaneously, preset business objective tags may include "order conversion," "user growth," and "GMV increase." The core business objective set by the business department is clearly defined as "ensuring the smooth operation of flash sales and increasing overall order volume."

[0036] Next, based on the list of active marketing activities and preset business target tags, the system derives the business target tags for "Flash Sale" as "Order Conversion" and "GMV Increase," and for "New User Registration Bonus" as "User Growth."

[0037] Then, the system combines real-time traffic share, active marketing campaign business objective tags, and core business objectives to infer the current overall business intent. For example, if traffic surges during a flash sale and the core business objective is to ensure the success of the flash sale, the overall business intent is inferred to be "high-priority flash sale order conversion," and a higher priority score is assigned to it.

[0038] Meanwhile, the system has established resource consumption profiles for each type of business activity (such as "product browsing," "adding to cart," "order payment," and "user registration") on shared basic services (such as "product database," "order service," "payment gateway," and "user service"). For example, the "order payment" activity consumes a lot of CPU, memory, and network bandwidth from the "order service" and "payment gateway." The system also collects real-time internal operational details of these shared basic services, such as CPU utilization and queue depth.

[0039] When the system detects that the real-time proportion of "order placement and payment" business activities deviates from the baseline of regular business composition (for example, a sudden drop in order placement and payment traffic), and the CPU utilization of "order service" increases abnormally and the internal queue length increases, the system will identify and quantify abnormal business composition events.

[0040] At this point, the system will calculate the business impact value corresponding to the "order placement and payment" business activity based on the resource consumption characteristics of the "order service" and "payment gateway" for the "order placement and payment" business activity, combined with the priority score and real-time proportion of the comprehensive business intent of "high-priority flash sale order conversion," as well as the internal operating details of the "order service" and "payment gateway." Because "order placement and payment" is a key link of the high-priority business intent, and its dependent shared service experiences bottlenecks, its business impact value will be calculated as very high.

[0041] At the same time, the system will determine the degree of deviation of abnormal events in business composition based on the difference between the real-time proportion of "order payment" and the benchmark of regular business composition.

[0042] Finally, the system integrates the deviation degree of the business component anomaly event with the calculated business impact value. Even if the "order placement and payment" traffic deviation degree is not the highest, its extremely high business impact value will significantly enhance the quantification result of the integrated anomaly event, clearly indicating that this is a high-risk anomaly event requiring urgent attention, and generating business component anomaly event information containing this detailed information. In this way, IT resource health assessment can more accurately reflect the issues that have the greatest impact on core business, guiding operations personnel to prioritize their handling.

[0043] The steps for calculating the business impact value corresponding to each type of business activity further include: Identify business activities related to multiple currently active high-priority integrated business intentions; Predict the resource requirements of each identified business activity for the shared basic service under the current business traffic composition, and derive the predicted total resource requirements of multiple business activities for the same shared basic service. Monitor the real-time resource utilization and internal queuing status of shared basic services; When the total resource demand of multiple predicted business activities on the same shared basic service exceeds the available resource capacity of the shared basic service, or when the internal queuing waiting situation of the shared basic service exceeds the preset waiting threshold, or when the real-time resource utilization rate of the shared basic service exceeds the preset utilization rate threshold, it is determined that resource contention has occurred in the shared basic service. Based on the shared basic services where resource contention has occurred, the expected performance impact coefficient of each business activity on the shared basic services where resource contention has occurred will be dynamically adjusted. The adjustment will be based on the priority of each business activity to the shared resources and the degree of dependence of each business activity on the shared resources. The expected performance impact coefficients and indicators after integration and adjustment are used to calculate the business impact value corresponding to each type of business activity. Specifically, the indicators are the actual performance indicators of the business activities on shared resources.

[0044] Specifically, when calculating the business impact value corresponding to each type of business activity, it is first necessary to identify the business activities related to the multiple currently active high-priority integrated business intentions. These high-priority integrated business intentions can be inferred through analysis of marketing activities, business objectives, and core business objectives, representing the key business directions that the current business system needs to prioritize. The business activities related to these high-priority integrated business intentions refer to the IT operations or processes that directly support or serve these high-priority business intentions. For example, in an e-commerce system, a high-priority integrated business intention might be "ensuring the order payment success rate during the Double Eleven promotion," then related business activities might include "user order placement," "payment processing," and "inventory deduction," etc.

[0045] Furthermore, it is necessary to predict the resource requirements of each identified business activity on the shared infrastructure service under the current business traffic composition, and to derive the predicted total resource requirements of multiple business activities on the same shared infrastructure service. This prediction can be based on historical data, machine learning models, or expert experience, aiming to estimate how much CPU, memory, network bandwidth, or I / O resources each business activity will consume from the shared infrastructure service (e.g., database, message queue, caching service, etc.) under the current business load and composition. By aggregating the predicted requirements of all relevant business activities on a specific shared infrastructure service, the total load that shared infrastructure service may face can be obtained.

[0046] Meanwhile, the system continuously monitors the real-time resource utilization and internal queuing status of shared basic services. Real-time resource utilization refers to the proportion of CPU, memory, and other resources currently consumed by the shared basic services relative to the total available resources. Internal queuing status reflects metrics such as the queue length and average waiting time for requests processed within the shared basic services; these metrics directly indicate bottlenecks in service processing capacity.

[0047] Resource contention in a shared infrastructure service is considered to occur when any of the following conditions are met: First, the total resource demand of multiple predicted business activities on the same shared infrastructure service exceeds the available resource capacity of the shared infrastructure service, indicating that theoretically, the service will not be able to meet the needs of all business activities; second, the internal queuing situation of the shared infrastructure service exceeds a preset waiting threshold, indicating that the service has experienced significant processing delays; third, the real-time resource utilization rate of the shared infrastructure service exceeds a preset utilization rate threshold, which usually means that the service is under high load and there is a risk of performance degradation. These thresholds can be set based on the service's SLA (Service Level Agreement), historical performance baseline, and business risk appetite.

[0048] Once resource contention is detected in a shared infrastructure service, the system dynamically adjusts the expected performance impact coefficient of each business activity on the shared infrastructure service experiencing contention. The expected performance impact coefficient is a quantitative indicator representing the potential performance impact on a specific business activity when contention occurs. The adjustment is primarily based on the priority of each business activity to the shared resource and the degree of dependence of each business activity on the shared resource. The priority of a business activity to the shared resource can be determined based on the priority score of its overall business intent; higher-priority business activities should have their performance impact minimized during resource contention. The degree of dependence of a business activity on the shared resource measures the strength of its reliance on a specific shared resource to complete its function; the higher the dependence, the greater the potential impact of resource contention. By comprehensively considering these factors, the performance impact of resource contention can be allocated more rationally.

[0049] Finally, the adjusted expected performance impact coefficient is integrated with the indicators to calculate the business impact value for each type of business activity. The indicators specifically refer to the actual performance metrics of the business activity on shared resources, such as response time, throughput, and error rate. The integration process can employ weighted averaging, product, or other mathematical models, using the expected performance impact coefficient as a correction factor to adjust the actual performance indicators, thereby obtaining a more accurate business impact value that reflects the actual impact of resource contention on the performance of business activities.

[0050] Specifically, firstly, by identifying high-priority business intentions and their related business activities, the system ensures focus on critical business processes. Secondly, by predicting the resource demands of business activities on shared infrastructure services and combining this with monitoring real-time resource utilization and internal queuing, the system can promptly and accurately determine whether resource contention is occurring in shared infrastructure services. Once resource contention is detected, the system no longer simply evaluates based on conventional performance indicators, but dynamically adjusts the expected performance impact coefficient according to the priority and dependence of business activities on shared resources. This adjustment mechanism ensures that the calculation of business impact values ​​fully reflects the differentiated impact of resource contention on the performance of different business activities, thereby avoiding evaluation blind spots caused by resource contention and making the IT resource health assessment results closer to the actual business operation status.

[0051] In some preferred embodiments, it is assumed that an e-commerce platform faces massive traffic during the "Double Eleven" promotion. "User order placement" and "payment processing" are identified as key business activities under high-priority overall business intent, both relying on a shared database service. The system first predicts the total resource requirements of the database service for "user order placement" and "payment processing" under the current traffic. Simultaneously, the system continuously monitors the database service's CPU utilization, I / O throughput, and average wait time of the internal transaction queue.

[0052] For example, when the predicted total resource demand exceeds the available database capacity, or when the database CPU utilization consistently exceeds 90% (a preset utilization threshold), and the average wait time of the internal transaction queue exceeds 500 milliseconds (a preset wait threshold), the system determines that resource contention has occurred in the database service. In this case, the system dynamically adjusts the expected performance impact coefficients of "user order placement" and "payment processing" based on their priority to database resources (e.g., "payment processing" has a higher priority than "user order placement") and their dependence on the database (e.g., both are highly dependent on the database). For example, the expected performance impact coefficient for "payment processing" might be adjusted to 1.5, while for "user order placement" it might be adjusted to 1.2. This means that under resource contention, the performance degradation of "payment processing" is considered to have a greater impact on the business.

[0053] The system then integrates these adjusted expected performance impact coefficients with the actual collected database performance metrics (such as response time and transaction success rate). For example, if the actual response time for "payment processing" increases from 100 milliseconds under normal circumstances to 500 milliseconds, the business impact value will be calculated as a higher value after integrating the adjusted coefficients, reflecting the severe impact of resource contention on its critical business operations. In this way, even if the overall health score of the database service may be high, the corresponding business impact value will be significantly increased due to the impact of resource contention on high-priority business activities. This more accurately reflects the potential risks of IT resources to core business operations, prompting the operations team to prioritize and resolve database performance bottlenecks.

[0054] This application further proposes steps for calculating the business impact value corresponding to each type of business activity, which further include: Monitor the update timestamp of the adjusted expected performance impact coefficient and the collection timestamp of the indicator; When the adjusted expected performance impact coefficient is updated, the latest values ​​of all related metrics are collected. From the collected metrics, extract the latest metric value within the current time window whose collection timestamp is closest to the adjusted expected performance impact coefficient. If the indicator is not updated within the current time window, use the latest value of the corresponding indicator in the previous valid time window; By integrating the adjusted expected performance impact coefficient with the extracted indicator values, the business impact value corresponding to each type of business activity is calculated.

[0055] Specifically, monitoring the update timestamps of the adjusted expected performance impact coefficients and the collection timestamps of the metrics refers to the system continuously tracking and recording the most recent update time of each adjusted expected performance impact coefficient, as well as the collection time of the actual performance metrics of each related business activity on shared resources. These timestamps are used to ensure that the coefficient and metric data used in calculating business impact values ​​are consistent and up-to-date in time.

[0056] Furthermore, when the adjusted expected performance impact coefficient is updated, the system automatically triggers the collection of the latest values ​​for all metrics associated with that coefficient. This is designed to ensure that the latest performance metric data is obtained immediately once the impact coefficient changes, reflecting the latest business context and resource status.

[0057] Specifically, from the collected indicators, the system extracts the latest indicator values ​​within the current time window whose collection timestamp is closest to the update timestamp of the adjusted expected performance impact coefficient. This can be understood as the system prioritizing indicator values ​​within the current analysis time window whose collection timestamps are closest to the update timestamp of the adjusted expected performance impact coefficient after acquiring the latest indicator data. This maximizes the synchronization between the coefficients and indicators over time, improving calculation accuracy.

[0058] In practical applications, when an indicator is not updated within the current time window, the system will use the latest value of the corresponding indicator in the previous valid time window. This is a data backtracking mechanism used to handle situations where indicator data is missing or not updated within a specific time window due to various reasons (such as collection delays, system failures, etc.), ensuring the continuity and integrity of the calculation process and avoiding interruptions in evaluation due to data gaps.

[0059] Therefore, by integrating the adjusted expected performance impact coefficient with the extracted indicator values, the business impact value corresponding to each type of business activity is calculated. This step combines the coefficients and indicators that have undergone time synchronization and data integrity processing to obtain a more accurate and real-time business impact value.

[0060] In some preferred embodiments, it is assumed that the adjusted expected performance impact coefficient of a certain shared infrastructure service (e.g., a database service) was updated at 10:00:00 AM, indicating that the impact of this service on a certain high-priority business activity has changed. According to the scheme of this application, the system will immediately trigger the collection of the latest values ​​of performance indicators (e.g., database response time, transaction volume, etc.) for all business activities associated with the database service. Assume that the latest database response time indicator value is collected at 10:00:05, with a collection timestamp of 10:00:05. When calculating the business impact value of this business activity, the system will extract this indicator value that is the latest within the current time window and whose collection timestamp is closest to 10:00:00 (i.e., the response time collected at 10:00:05).

[0061] Furthermore, if a business activity lacks new data collection for a key performance indicator (such as the success rate of a specific API call) within the current time window (e.g., 10:00:00 to 10:01:00), the system will not interrupt the calculation. Instead, it will search for and use the latest value of the API call success rate collected in the previous valid time window (e.g., 09:59:00 to 10:00:00). In this way, even in the event of brief delays or intermittent data collection, the calculation of business impact values ​​can continue, using the latest and most valid data possible, thus ensuring the accuracy and continuity of the evaluation results.

[0062] This application further proposes that the steps for determining resource contention in shared basic services in the above-mentioned IT resource health assessment method include: Continuously obtain the elastic scaling status, dynamic load balancing information, and currently configured resource quotas of shared basic services; Based on the elastic scaling status of shared basic services, dynamic load balancing information, and currently configured resource quotas, the dynamic available resource capacity of shared basic services is calculated in real time. By combining historical traffic patterns, current business context, and the impact of external events on resource demand, the uncertainty of the total resource demand of multiple business activities on the same shared basic service is quantified, and the quantification results are obtained. Compare the quantification results with the dynamic available resource capacity of shared basic services; When the quantification result exceeds the dynamic available resource capacity of the shared basic service, it is determined that resource contention has occurred in the shared basic service.

[0063] Specifically, continuously acquiring the elastic scaling status, dynamic load balancing information, and currently configured resource quotas of shared infrastructure services aims to comprehensively and in real-time understand the current resource supply capacity and potential resource bottlenecks of these services. Elastic scaling status refers to the shared infrastructure service's ability to automatically adjust resource quantities based on load changes; for example, when the load increases, the service automatically scales up instances, and when the load decreases, the service automatically scales down instances. Dynamic load balancing information refers to how the load balancer distributes incoming requests to backend service instances, including the current number of connections, response time, and health check status, reflecting the real-time situation of traffic distribution. Currently configured resource quotas refer to the maximum resource limits allocated to the shared infrastructure service, such as CPU, memory, and network bandwidth; these quotas may be set by the cloud platform or system administrator.

[0064] Based on this, the dynamic available resource capacity of the shared basic service is calculated in real time according to the elastic scaling status, dynamic load balancing information, and currently configured resource quotas. The dynamic available resource capacity refers to the total amount of resources that the shared basic service can actually provide at any given moment, after considering factors such as elastic scaling, load balancing, and resource quotas. It is a dynamically changing value, not a static fixed value. This capacity can be obtained by multiplying the number of currently active instances by the effective resource capacity of each instance, combined with the actual distribution capacity of the load balancer and unused resource quotas. Its purpose is to ensure a more accurate assessment of the service's actual carrying capacity and reflect its true situation in a dynamic environment.

[0065] Furthermore, by combining historical traffic patterns, current business context, and the impact of external events on resource demand, the uncertainty of the total resource demand for the same shared basic service from multiple business activities is quantified, yielding a quantified result. Uncertainty quantification means that when predicting resource demand, in addition to providing a single predicted value, the possible error range or confidence interval of the predicted value is also considered. Historical traffic patterns refer to the periodic and trending data of business request volume and resource consumption over a past period; current business context refers to factors such as ongoing marketing campaigns, system maintenance, and sudden hot events, all of which affect short-term resource demand; external events refer to factors such as news hotspots, social media trends, and seasonal peaks, which may lead to unexpected increases in business traffic. The quantified result can be a prediction interval, such as a 95% confidence interval, or a probability distribution representing the likelihood that resource demand falls within a certain range. Its purpose is to improve the robustness of resource demand prediction and avoid misjudgments due to the inaccuracy of a single predicted value.

[0066] Subsequently, the quantification results are compared with the dynamic available resource capacity of the shared basic services. This comparison can be made by comparing the upper limit of the quantification results, such as the upper bound of the confidence interval, with the dynamic available resource capacity, or by comparing the probability distribution of the quantification results with the capacity threshold. The purpose is to more comprehensively assess the degree of matching between resource demand and supply, taking into account the uncertainty of demand.

[0067] Therefore, when the quantified result exceeds the dynamic available resource capacity of the shared basic service, resource contention is determined to have occurred in the shared basic service. This criterion means that when the resource demand, which takes into account uncertainty, such as its predicted upper limit, exceeds the dynamic available resource capacity that the shared basic service can actually provide at the current moment, resource contention is considered to have occurred. Its purpose is to ensure that potential resource contention risks can still be identified in a timely and accurate manner even when there are fluctuations and uncertainties in resource demand.

[0068] This application further proposes steps for determining resource contention in shared basic services, including: Continuously collect resource utilization data of shared basic services within multiple consecutive time windows, and calculate the average resource utilization rate and the rate of change of resource utilization rate within multiple consecutive time windows. Set dynamic thresholds, which are adjusted based on the historical load characteristics of shared basic services, the current business context, and the preset business risk level; When the average resource utilization rate within multiple consecutive time windows continuously exceeds the dynamic threshold, and the rate of change of resource utilization rate within multiple consecutive time windows indicates a continuous upward trend, it is determined that resource contention has occurred in the shared basic service. Monitor resource utilization spikes within a single time window. If a resource utilization spike exceeds a preset instantaneous threshold and its duration is shorter than a preset tolerance time, it is not considered a resource contention and is instead marked as a potential risk event. Increase the monitoring frequency and sensitivity of subsequent time windows.

[0069] Specifically, continuously collecting resource utilization data for shared basic services across multiple consecutive time windows refers to the system uninterruptedly collecting information on resource utilization of shared basic services at different time periods, such as CPU utilization, memory utilization, and network bandwidth utilization. By collecting data across multiple consecutive time windows, more comprehensive historical data can be obtained, providing a data foundation for subsequent trend analysis and anomaly detection. Calculating the average resource utilization rate across multiple consecutive time windows aims to smooth out instantaneous fluctuations and reflect the overall trend and baseline level of resource usage; while calculating the rate of change of resource utilization across multiple consecutive time windows is used to capture the growth or decline trend of resource usage and identify potential performance bottlenecks or resource exhaustion risks.

[0070] Furthermore, setting dynamic thresholds can be understood as flexibly adjusting the critical value for resource contention determination based on the actual operation of shared basic services and business needs. This dynamic threshold is not fixed but is adjusted in real time based on the historical load characteristics of the shared basic services (e.g., different load patterns during weekdays and weekends, peak and off-peak periods), the current business context (e.g., whether there are large-scale marketing campaigns or sudden traffic surges), and the preset business risk level (e.g., different tolerance levels for resource contention between core and non-core businesses). The purpose is to make resource contention determination more accurate and avoid false alarms or missed alarms caused by static thresholds.

[0071] Therefore, when the average resource utilization rate across multiple consecutive time windows consistently exceeds the dynamic threshold, and the rate of change of resource utilization rate across multiple consecutive time windows indicates a continuous upward trend, the system will determine that resource contention has occurred in the shared basic service. This dual judgment mechanism, which combines the absolute level of resource utilization rate (average exceeding the dynamic threshold) and the trend of change (continuous growth), can more accurately identify the true resource pressure and avoid misjudging short-term peaks or normal fluctuations as contention.

[0072] Furthermore, this application proposes a mechanism for monitoring resource utilization spikes within a single time window. Specifically, when a resource utilization spike within a single time window exceeds a preset instantaneous threshold, but the duration of the spike is shorter than a preset tolerance time, the system will not immediately classify it as resource contention. Such instantaneous spikes are usually caused by short-term bursts of traffic or occasional events, rather than continuous resource pressure. In this case, the spike will be marked as a potential risk event, rather than directly classified as contention, to avoid unnecessary alarms and resource adjustments. As a preventative measure, the system will increase the monitoring frequency and sensitivity of subsequent time windows to more closely monitor the operational status of the shared infrastructure service, enabling timely detection and handling of potential risks that evolve into persistent problems.

[0073] In some preferred embodiments, suppose that the resource utilization rate of a shared basic service (e.g., order processing service) of an e-commerce platform typically fluctuates between 40% and 60% during daily operations. On a Friday evening, due to a limited-time flash sale event held by the platform, the resource utilization rate of the order processing service rapidly surged to 90% in a short period of time, and then returned to normal levels after about 10 minutes.

[0074] If judged solely based on the instantaneous resource utilization exceeding a preset threshold (e.g., 80%), the service might be immediately identified as experiencing resource contention, triggering a series of emergency responses such as resource scaling or traffic limiting. However, since the peak duration is shorter than the preset tolerance time (e.g., 15 minutes), it may actually just be an acceptable instantaneous load spike, not a genuine resource contention requiring emergency intervention.

[0075] According to the solution proposed in this application, the system continuously collects resource utilization data for the order processing service across multiple consecutive time windows. During a flash sale, although a peak in resource utilization within a single time window may exceed a preset instantaneous threshold, the system will not immediately classify it as resource contention because its duration is shorter than the preset tolerance time. Instead, this peak will be marked as a potential risk event. Simultaneously, the system will increase the monitoring frequency and sensitivity of subsequent time windows, closely monitoring the resource utilization of the order processing service. If subsequent monitoring reveals that the average resource utilization consistently exceeds a dynamic threshold (which may be adjusted higher based on historical flash sale data and the current business risk level), and the rate of change in resource utilization indicates a continuous upward trend, then the system will determine that resource contention has occurred and trigger corresponding adjustment strategies. This approach avoids overreacting to instantaneous traffic fluctuations while ensuring timely identification and response to persistent resource pressure, thereby improving the accuracy and efficiency of IT resource health assessment.

[0076] This application further proposes steps for setting dynamic thresholds, including: Perform integrity checks and outlier identification on historical load characteristic data of shared basic services; When the historical load characteristic data of shared basic services is incomplete, the incomplete data is interpolated based on data from adjacent time points or data from similar shared services. When outliers are identified in the historical load characteristic data of shared basic services, the outliers are smoothed or removed based on the degree and duration of deviation. Periodic and trend analyses are performed on the processed historical load characteristics data of shared basic services to identify load patterns at different time granularities. Based on the identified load patterns, current business context, and preset business risk levels, adjust the baseline coefficient and risk weighting factor in the dynamic threshold calculation formula; By combining the baseline coefficient and risk weighting factor in the adjusted dynamic threshold calculation formula, a confidence interval or prediction error range is introduced in the dynamic threshold calculation process to obtain the dynamic threshold.

[0077] Specifically, performing integrity checks and outlier identification on historical load characteristic data of shared infrastructure services refers to conducting a preliminary quality assessment of various performance metrics data (such as CPU utilization, memory usage, network I / O, etc.) collected from shared infrastructure services over a past period. Integrity checks aim to identify missing data, interrupted recording, or failed data collection; outlier identification uses statistical methods or machine learning algorithms to detect data points that significantly deviate from the normal range, such as abnormally high or low values ​​caused by transient failures or misconfigurations.

[0078] When historical load characteristic data for shared basic services is incomplete, the incomplete data is imputed based on data from adjacent time points or data from similar shared services. This means that when data loss is detected, time series interpolation methods, such as linear interpolation and spline interpolation, can be used to fill in the missing data points using valid data adjacent to them. As an alternative, if the quality of adjacent data is poor or the range of missing data is large, historical load data from other similar shared services with similar business functions and architectural characteristics can be referenced, and the incomplete data can be supplemented through appropriate calibration and matching.

[0079] When outliers are identified in the historical load characteristic data of shared infrastructure services, they are smoothed or removed based on their degree of deviation and duration. Specifically, for outliers with small deviations and short durations, methods such as moving averages and exponential smoothing can be used to reduce their impact on the overall data trend. For outliers with large deviations and long durations, they may be considered data errors or special events and can be directly removed to avoid misleading load pattern analysis.

[0080] Periodic and trend analyses are performed on the processed historical load characteristic data of shared infrastructure services to identify load patterns at different time granularities. Periodic analysis aims to discover recurring patterns in the data, such as daily peaks, weekly periodic fluctuations, or monthly settlement peaks. Trend analysis focuses on the long-term evolution of the data, such as the continuous growth or decline in business volume. Through these analyses, typical load patterns exhibited by shared infrastructure services at different time scales (such as hours, days, weeks, and months) can be identified.

[0081] Based on the identified load patterns, current business context, and preset business risk levels, the baseline coefficient and risk weighting factor in the dynamic threshold calculation formula are adjusted. Specifically, the baseline coefficient can be adjusted according to the identified periodic load patterns, such as increasing the baseline during daily peak hours. The risk weighting factor can be dynamically adjusted according to the current business context (such as whether there are large-scale promotional activities, system upgrades, etc.) and preset business risk levels (such as high-risk businesses having low tolerance for resource contention), to make the threshold more sensitive or more lenient.

[0082] By combining the baseline coefficient and risk weighting factor in the adjusted dynamic threshold calculation formula, a confidence interval or prediction error range is introduced during the dynamic threshold calculation process to obtain the dynamic threshold. This means that the final dynamic threshold is not a fixed value, but a range that takes into account data uncertainty and prediction error. For example, an average value based on historical data and the prediction model can be calculated, and a confidence interval (such as a 95% confidence interval) can be added to this value, so that the threshold can better adapt to the volatility of the actual load and reduce false alarms or missed alarms.

[0083] In some preferred embodiments, suppose a shared database service has historical CPU utilization data missing for a certain period due to a monitoring system failure, and that a temporary batch data import task on a weekend causes an abnormally high CPU utilization value that lasts for several hours. First, the system performs an integrity check on this historical load characteristic data, discovering missing data and identifying the abnormally high value. For missing data, the system can perform linear imputation based on CPU utilization data before and after the failure. For abnormally high values, due to their large deviation and long duration, the system may mark them as special events and smooth them out, or reduce their weight during analysis to avoid excessively affecting the identification of normal load patterns. Next, the system performs periodic analysis on the processed data, identifying that the database service has significant CPU utilization peaks during weekdays and low values ​​at night and on weekends. Simultaneously, trend analysis may show that the overall load of the database service has been slowly increasing over the past few months. When setting dynamic thresholds, the system will increase the baseline coefficient during weekdays based on these identified load patterns to allow for higher normal loads. If the current business scenario involves a large-scale online promotional campaign, the system will further increase the risk weighting factor based on a preset business risk level (e.g., the database performance requirements are extremely high during the promotion, with a risk level of "high"). This makes the CPU utilization threshold more sensitive, allowing for timely detection of resource contention in its early stages. Ultimately, when calculating the dynamic threshold, the system combines the adjusted baseline coefficient and the risk weighting factor, introducing a confidence interval. For example, the dynamic threshold might be set as "predicted average CPU utilization + 1.5 times the standard deviation + 10% risk buffer," resulting in a dynamic threshold that can adapt to both daily fluctuations and special business scenarios, ensuring accurate judgment of database resource contention.

[0084] When the historical load characteristic data of shared basic services is incomplete, the steps to interpolate the incomplete data based on data from adjacent time points or data from similar shared services include: Perform quality assessment on data at adjacent time points to identify whether there are missing or abnormal data at adjacent time points; When the data quality is poor at adjacent time points, historical load characteristic data of similar shared services with similar business functions and architectural characteristics to the current shared basic service are extracted as a reference. Based on the current business activity composition, real-time traffic share, and dependence on shared resources of the shared basic services, business scenario matching analysis and load mode calibration are performed on the extracted historical load characteristic data of similar shared services to obtain calibrated historical load characteristic data of similar shared services. The historical load characteristic data of the calibrated shared services are weighted and fused with the existing valid historical data of the current shared basic services to generate interpolated data.

[0085] The quality assessment of data from adjacent time points aims to ensure the reliability of the data used for interpolation. Specifically, this involves checking the time series continuity, the reasonableness of the numerical range, and the presence of obvious jumps or zero values ​​in the data from adjacent time points. If missing or abnormal data is found, it indicates that the data from adjacent time points is of poor quality and should not be used directly for interpolation.

[0086] Furthermore, when data quality at adjacent time points is deemed poor, historical load characteristic data can be extracted from similar shared services with similar business functions and architectural characteristics to the current shared infrastructure service in order to obtain more reliable reference data. For example, if the current shared infrastructure service is an order processing service for an e-commerce platform, then historical load data from order processing services of another e-commerce platform or other regions within the same platform can be referenced. This selection of reference data aims to ensure the similarity of data in business logic and technical implementation, thereby improving the accuracy of interpolation.

[0087] Based on this, and considering the current shared basic service's business activity composition, real-time traffic share, and dependence on shared resources, historical load characteristic data of similar shared services are analyzed for business scenario matching and load pattern calibration. Specifically, this involves analyzing the current service's peak and off-peak periods, the impact of specific marketing activities, etc., and comparing this data with historical data of similar services. The time axis or numerical proportions of the similar service's data are then adjusted to better reflect the current service's actual operating context. For example, if the current service experiences a surge in traffic due to a special promotional activity during a certain period, while similar services do not, the data for similar services needs to be amplified or adjusted accordingly to calibrate their load patterns.

[0088] Finally, the calibrated historical load characteristic data of similar shared services is weighted and fused with the existing valid historical data of the current shared infrastructure service to generate imputation data. Weighted fusion can be performed in various ways; for example, different weights can be assigned based on factors such as data timeliness, completeness, and confidence level. Typically, the valid historical data of the current service itself is given higher weight, while the calibrated data from similar services serves as a supplement and correction. Through this weighted fusion, existing information can be utilized to the maximum extent to generate more accurate and complete historical load characteristic data that considers both its own characteristics and external experience.

[0089] The steps to identify load patterns at different time granularities include: Data aggregation is performed on the historical load characteristic data of the processed shared basic services that belong to low-frequency business activities or off-peak periods to increase data density; Based on the amount and noise level of the aggregated data, a smoothing filter is selected to preprocess the aggregated data to obtain the preprocessed data. For low-frequency business activity data that is sparse or has high noise levels in the preprocessed data, a pattern matching method based on event sequences is used to identify periodic patterns. For data from off-peak periods in the preprocessed data, piecewise linear regression or locally weighted regression methods are used to identify trend patterns.

[0090] The process involves aggregating data from historical load characteristics of shared basic services that pertain to low-frequency business activities or off-peak periods. This aims to address the potential sparsity of this data. Low-frequency business activities typically refer to operations that occur infrequently, resulting in sparse data distribution across the timeline. Off-peak period data may also appear less dense due to lower transaction volumes. Data aggregation, such as summing, averaging, or counting data within a specific time window, effectively increases data density, providing subsequent pattern recognition algorithms with sufficient information for analysis.

[0091] Furthermore, based on the amount and noise level of the aggregated data, a smoothing filter is selected to preprocess the aggregated data. The selection of the smoothing filter aims to reduce random noise in the data, making potential patterns clearer. For example, when the data volume is large and the noise exhibits a Gaussian distribution, a Gaussian filter can be used; when there are sharp outliers in the data, a median filter may be more suitable. The preprocessed data can better reveal its inherent periodicity or trend characteristics.

[0092] Specifically, for low-frequency business activity data that is sparse or has high noise levels in the preprocessed data, a pattern matching method based on event sequences is used to identify periodic patterns. This method focuses on analyzing the order and time intervals of events, rather than continuous numerical changes. For example, it can identify patterns in which specific low-frequency business activities recur at fixed times each week or month, even if the data volume for each activity may be small or fluctuate.

[0093] Furthermore, for preprocessed data belonging to off-peak periods, piecewise linear regression or locally weighted regression methods are used to identify trend patterns. Off-peak data typically exhibits a relatively gentle trend, but may show different slopes in different time periods. Piecewise linear regression divides the data into multiple intervals and fits a straight line within each interval, thus capturing the piecewise linear trend. Locally weighted regression, on the other hand, assigns different weights to local data points for fitting, more flexibly adapting to local changes in the data and identifying non-linear trends.

[0094] refer to Figure 2 This application proposes an IT resource health assessment system, applied to the aforementioned IT resource health assessment method. The system includes: The parsing module extracts business activity characteristics from incoming business requests at the IT system entry point, identifies the business activity type according to preset rules, and obtains the number of requests for each business activity type. The percentage calculation module calculates the real-time percentage of each business activity type in the total number of business requests based on the number of requests for each business activity type. The processing module compares the real-time percentage with the preset benchmark for regular business composition. When the real-time percentage deviates from the benchmark for regular business composition, it identifies and quantifies abnormal business composition events and generates abnormal business composition event information. The matching module searches for matching adjustment strategies from preset weight adjustment rules based on the abnormal events constituted by the business. The adjustment strategy includes the triggering conditions, the affected indicators, and the weight adjustment method. The update module calculates and applies new weights to various business indicators in the health assessment based on the matching adjustment strategy. The weighting calculation module uses new weights to calculate the IT resource health score; The output module outputs IT resource health scores and information on abnormal events in business operations.

[0095] Specifically, the parsing module is configured at the IT system entry point to monitor and analyze all incoming business requests in real time. This module can identify and extract key business activity characteristics from the requests, such as the request source, target service, operation type, and user identity. Subsequently, based on a pre-defined set of rules, these extracted characteristics are matched and classified to identify different business activity types. For example, user login, product browsing, and order submission can be identified as different business activity types. By counting the identified business activity types, the number of requests for each business activity type within a specific time window can be obtained.

[0096] The percentage calculation module is configured to receive the request counts for each business activity type output by the parsing module. Based on these counts, this module can calculate the real-time percentage of each business activity type in the total number of business requests. For example, if the total number of requests is 1000, and 300 are user login requests, then the real-time percentage of user login requests is 30%. Calculating the real-time percentage helps to dynamically reflect the composition of current business traffic.

[0097] In practical applications, the processing module is configured to compare the real-time percentage output by the percentage calculation module with a preset benchmark for normal business composition. This benchmark can be a normal business traffic pattern obtained through historical data analysis. When a significant deviation occurs between the real-time percentage and the benchmark, such as a sudden increase or decrease in the percentage of a certain type of business activity, the processing module identifies and quantifies this abnormal business composition event. The quantification process may include calculating the magnitude and duration of the deviation, and generating abnormal business composition event information containing this information.

[0098] Furthermore, the matching module is configured to receive abnormal event information related to business components generated by the processing module. This module searches for the most matching adjustment strategy from a pre-defined weight adjustment rule base based on the type, severity, and other characteristics of the abnormal event. Each adjustment strategy explicitly defines the conditions for triggering the strategy, the affected business metrics (such as response time, throughput, error rate, etc.), and the specific weight adjustment method (such as increasing, decreasing, or adjusting proportionally).

[0099] Furthermore, the update module is configured to calculate and apply new weights to various business metrics in the health assessment based on the adjustment strategy selected by the matching module. For example, if an anomaly indicates a surge in user login traffic, it may be necessary to increase the weight of IT resource performance metrics related to user login (such as the response time of authentication services) in the overall health assessment to more sensitively reflect the health status of this critical business.

[0100] The weight calculation module is configured to receive new weights from the update module. This module will use these new weights, combined with real-time data from various business metrics, to calculate the overall health score of the IT resources. The health score is typically a quantitative indicator used to visually represent the operational status of the IT system.

[0101] Finally, the output module is configured to output the calculated IT resource health score and the generated business component anomaly event information. This information can be displayed on the monitoring dashboard or notified to operations and maintenance personnel through the alarm system so that timely intervention measures can be taken.

[0102] The content disclosed above is only a preferred and feasible embodiment of the present invention, and is not intended to limit the scope of protection of the present invention. Therefore, all equivalent technical changes made based on the content of the present invention specification and drawings are included within the scope of protection of the present invention. Furthermore, the elements therein can be updated as technology develops.

Claims

1. A method for assessing the health of IT resources, characterized in that, The method comprises the following steps: extracting service activity features from the incoming service requests at the IT system entrance, and identifying the service activity types according to preset rules to obtain the number of service activity requests of each service activity type; According to the number of service activity requests of each service activity type, the real-time proportion of each service activity type in the total amount of all service requests is calculated; Compare the real-time proportion with the preset regular business composition benchmark. When the real-time proportion deviates from the regular business composition benchmark, identify and quantify the business composition abnormal event, and generate business composition abnormal event information; According to the business composition abnormal event, find the matching adjustment strategy from the preset weight adjustment rule. The adjustment strategy includes trigger conditions, affected indicators and weight adjustment methods; According to the matched adjustment strategy, the new weight of each business indicator in the health degree evaluation is calculated and applied; Using the new weight, the IT resource health score is calculated; Output the IT resource health score and the business composition abnormal event information.

2. The method of claim 1, wherein, The method further comprises the following steps: Obtain the current active marketing activity list, the preset business target label and the core business target set by the business department; According to the current active marketing activity list and the preset business target label, the active marketing activity business target label is obtained; According to the real-time proportion, the active marketing activity business target label and the core business target set by the business department, the comprehensive business intention is inferred, and a priority score is assigned to each inferred comprehensive business intention; Establish the resource consumption feature file of each business activity type to the shared basic service, and collect the internal operation details of the shared basic service; According to the resource consumption feature file of each business activity type to the shared basic service, the business activity resource consumption feature is obtained; According to the priority score of the comprehensive business intention, the real-time proportion, the business activity resource consumption feature and the internal operation details of the shared basic service, the business impact value corresponding to each business activity type is calculated; According to the difference between the real-time proportion and the regular business composition benchmark, the deviation degree of the business composition abnormal event is obtained; Fuse the deviation degree of the business composition abnormal event and the business impact value to obtain the fused abnormal event quantification result. The fused abnormal event quantification result is used to further quantify the business composition abnormal event and generate the business composition abnormal event information.

3. The method of claim 2, wherein, The step of calculating the business impact value corresponding to each business activity type further comprises: Identify the business activities related to the current active multiple high-priority comprehensive business intentions; Predict the resource demand amount of each identified business activity to the shared basic service under the current business traffic composition, and obtain the total resource demand amount of the predicted multiple business activities to the same shared basic service; Monitor the real-time resource utilization rate and internal queuing waiting condition of the shared basic service; When the total resource demand amount of the predicted multiple business activities to the same shared basic service exceeds the available resource capacity of the shared basic service, or when the internal queuing waiting condition of the shared basic service exceeds the preset waiting threshold, or when the real-time resource utilization rate of the shared basic service exceeds the preset utilization rate threshold, it is determined that resource contention occurs in the shared basic service; According to the shared basic service determined to have resource contention, dynamically adjusting an expected performance influence coefficient of each service activity on the shared basic service having resource contention, the adjustment being based on a priority of each service activity on the shared resource and a dependence degree of each service activity on the shared resource; Fusing the adjusted expected performance influence coefficient and an index to calculate a service influence value corresponding to each service activity type, wherein the index is specifically an actual performance index of the service activity on the shared resource.

4. The method of claim 3, wherein, The step of calculating the service influence value corresponding to each service activity type further includes: Monitoring an update timestamp of the adjusted expected performance influence coefficient and a collection timestamp of the index; When the adjusted expected performance influence coefficient is updated, triggering collection of the latest value of all associated indexes; From the collected indexes, extracting an index value that is the latest in a current time window and has a collection timestamp closest to the update timestamp of the adjusted expected performance influence coefficient; When the index is not updated in the current time window, using the latest value of the corresponding index in the last valid time window; Fusing the adjusted expected performance influence coefficient and the extracted index value to calculate the service influence value corresponding to each service activity type.

5. The method of claim 3, wherein, The step of determining that the shared basic service has resource contention further includes: Continuously obtaining an elastic scaling state, dynamic load balancing information and a currently configured resource quota of the shared basic service; According to the elastic scaling state, dynamic load balancing information and the currently configured resource quota of the shared basic service, calculating a dynamic available resource capacity of the shared basic service in real time; Combining the influence of historical traffic patterns, current business scenarios and external events on resource demand, quantifying the total resource demand of the predicted multiple service activities on the same shared basic service, and obtaining a quantification result; Comparing the quantification result with the dynamic available resource capacity of the shared basic service; When the quantification result exceeds the dynamic available resource capacity of the shared basic service, determining that the shared basic service has resource contention.

6. The method for assessing the health of IT resources of claim 5, wherein, The step of determining that the shared basic service has resource contention further includes: Continuously collecting resource usage rate data of the shared basic service in multiple continuous time windows, and calculating an average value of the resource usage rate in the multiple continuous time windows and a change rate of the resource usage rate in the multiple continuous time windows; Setting a dynamic threshold, which is adjusted according to historical load characteristics of the shared basic service, a current business scenario and a preset business risk level; When the average value of the resource usage rate in the multiple continuous time windows continuously exceeds the dynamic threshold, and the change rate of the resource usage rate in the multiple continuous time windows indicates a continuous growth trend, it is determined that the shared basic service has resource contention; Monitoring resource usage rate spikes in a single time window, when a resource usage rate spike in a single time window exceeds a preset instantaneous threshold, and the duration of the resource usage rate spike is shorter than a preset tolerance time, it is not determined as resource contention, and the resource usage rate spike is marked as a potential risk event; Increasing the monitoring frequency and sensitivity of the subsequent time window.

7. The method of claim 6, wherein, The step of setting the dynamic threshold includes: Performing integrity checking and outlier identification on historical load characteristic data of the shared basic service; When the historical load characteristic data of the shared infrastructure service is incomplete, the incomplete data is interpolated according to adjacent time point data or similar shared service data; When an outlier is identified in the historical load characteristic data of the shared infrastructure service, the outlier is smoothed or removed according to the degree of deviation and the duration of the outlier; Periodic analysis and trend analysis are performed on the processed historical load characteristic data of the shared infrastructure service to identify load patterns of different time granularities; According to the identified load patterns, the current business context, and the preset business risk level, the baseline coefficient and the risk weighting factor in the dynamic threshold calculation formula are adjusted; In the dynamic threshold calculation process, the baseline coefficient and the risk weighting factor in the adjusted dynamic threshold calculation formula are combined to introduce a confidence interval or a prediction error range to obtain the dynamic threshold.

8. The method for assessing the health of IT resources of claim 7, wherein, When the historical load characteristic data of the shared infrastructure service is incomplete, the incomplete data is interpolated according to adjacent time point data or similar shared service data, including: Quality assessment is performed on the adjacent time point data to identify whether there are missing or abnormal data in the adjacent time point data; When the quality of the adjacent time point data is poor, historical load characteristic data of similar shared services is extracted from similar shared services with similar business functions and architecture characteristics as the current shared infrastructure service as a reference; Based on the business activity composition, real-time traffic proportion, and dependence on shared resources of the current shared infrastructure service, business scene matching degree analysis and load pattern calibration are performed on the extracted historical load characteristic data of similar shared services to obtain calibrated historical load characteristic data of similar shared services; The calibrated historical load characteristic data of similar shared services and the existing effective historical data of the current shared infrastructure service are weighted and fused to generate interpolated data.

9. The method of claim 7, wherein, The steps of identifying load patterns of different time granularities include: Data aggregation is performed on the data in the processed historical load characteristic data of the shared infrastructure service that belongs to low-frequency business activities or non-peak periods to increase data density; According to the amount and noise level of the aggregated data, a smoothing filter is selected to pre-process the aggregated data to obtain pre-processed data; For the pre-processed data that belongs to low-frequency business activities with sparse data or high noise level, a pattern matching method based on event sequence is used to identify periodic patterns; For the pre-processed data that belongs to non-peak periods, a piecewise linear regression or local weighted regression method is used to identify trend patterns.

10. An IT resource health assessment system, applied to the IT resource health assessment method of claim 1, characterized in that, The system includes: The analysis module extracts business activity features from incoming business requests at the IT system entrance and identifies business activity types according to preset rules to obtain the number of requests for each business activity type; The proportion calculation module calculates the real-time proportion of each business activity type in the total amount of all business requests according to the number of requests for each business activity type; The processing module compares the real-time proportion with the preset regular business composition benchmark, and when the real-time proportion deviates from the regular business composition benchmark, identifies and quantifies business composition abnormal events to generate business composition abnormal event information; The matching module finds a matched adjustment strategy from preset weight adjustment rules according to the service composition abnormal event, and the adjustment strategy contains a trigger condition, an affected index, and a weight adjustment manner. The updating module calculates and applies new weights of each business index in the health degree evaluation according to the matched adjustment strategy. The weight calculation module calculates the IT resource health score using the new weights. The output module outputs the IT resource health score and the service composition abnormal event information.

Images