Performance optimization method for digital twin system and related device

Abstract

The method for performance optimization of a digital twin system and related equipment provided in the application comprises: when it is determined that the digital twin system is in a performance risk state based on a key performance indicator, determining a processing resource bottleneck type according to a resource load performance indicator; constructing a corresponding cause-effect traceability chain according to a service load performance indicator corresponding to the processing resource bottleneck type and a corresponding target resource load performance indicator; calculating a resource load performance indicator density of each space unit of the digital twin system according to the cause-effect traceability chain, determining a bottleneck space range, and evaluating a load contribution degree of each service object to the processing resource bottleneck in the bottleneck space range to locate a bottleneck service object; and generating a performance optimization strategy of the digital twin system based on a scene type of the digital twin system, in combination with the processing resource bottleneck type, the bottleneck space range and the bottleneck service object. Thus, the accuracy of performance problem positioning of the digital twin system and the pertinence of the optimization strategy are improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a performance optimization method and related equipment for digital twin systems. Background Technology

[0002] With the development of 3D real-time applications such as digital twins, games, and virtual simulation, the system needs to handle complex rendering calculations and business logic simultaneously during operation. The system performance directly affects its stability and interactive experience. Therefore, it is of great significance to effectively monitor and optimize the system's operating performance.

[0003] Currently, in applications such as digital twins, games, and virtual simulations, system performance is typically monitored using independent performance indicator display tools. These tools display common rendering performance parameters such as frame rate, number of draw calls, number of triangles, and GPU / CPU time. Existing solutions primarily rely on passively displaying performance data and on human experience to analyze and optimize various indicators. Their monitoring metrics are concentrated at the general graphics rendering level, and their diagnostic methods are rather static, making it difficult to predict performance trends. Furthermore, the optimization suggestions they provide are mostly global guidance, lacking fine-grained positioning for specific spatial locations or business objects.

[0004] Therefore, existing performance monitoring and optimization methods are still designed for general 3D applications and are not specifically designed for the operational performance of digital twin systems, making it difficult to meet the performance diagnosis and optimization needs of digital twin systems. Summary of the Invention

[0005] The purpose of this application is to address at least one of the aforementioned technical deficiencies, particularly the technical deficiencies in the prior art that make it difficult to meet the performance diagnosis and optimization needs of digital twin systems.

[0006] Firstly, this application provides a performance optimization method for digital twin systems, the method comprising:

[0007] When a digital twin system is determined to be in a performance risk state based on key performance indicators, the type of processing resource bottleneck is determined according to resource load performance indicators. Key performance indicators represent the overall operating performance state of the digital twin system, while resource load performance indicators represent the load characteristics of processing resources.

[0008] Based on the business load performance indicators corresponding to the processing resource bottleneck type and the corresponding target resource load performance indicators, a corresponding causal tracing chain is constructed. The business load performance indicators represent the business load characteristics that trigger fluctuations in processing resources, and the causal tracing chain represents the triggering path that leads to the processing resource bottleneck.

[0009] Based on the causal chain, the resource load performance index density of each spatial unit of the digital twin system is calculated to determine the bottleneck space range. Within the bottleneck space range, the load contribution of each business object to the processing resource bottleneck is evaluated to locate the bottleneck business object.

[0010] Based on the scenario type of the digital twin system, combined with the type of processing resource bottleneck, the scope of the bottleneck space, and the bottleneck business object, a performance optimization strategy for the digital twin system is generated.

[0011] In one embodiment, the process of determining whether a digital twin system is in a performance risk state based on key performance indicators includes:

[0012] Time-series monitoring of key performance indicators;

[0013] If, based on preset performance thresholds, it is determined that key performance indicators meet the instantaneous conditions for performance anomalies, or, based on time-series monitoring results, it is determined that key performance indicators meet the conditions for abnormal performance trends, then the digital twin system is determined to be in a performance risk state.

[0014] In one embodiment, the step of determining the type of resource bottleneck to be addressed based on resource load performance metrics includes:

[0015] Multiple resource load performance indicators are obtained, and the bottleneck type of the processing resource is determined based on the degree of deviation of each resource load performance indicator from the corresponding preset load threshold. Each resource load performance indicator corresponds to the load characteristics of different processing resources.

[0016] In one embodiment, the step of constructing a corresponding causal chain based on the business load performance index corresponding to the processing resource bottleneck type and the corresponding target resource load performance index includes:

[0017] Determine the business load performance metrics corresponding to the processing resource bottleneck type, and the corresponding target resource load performance metrics.

[0018] Analyze the temporal sequence and correlation between target resource load performance indicators and business load performance indicators;

[0019] Based on the chronological order and the correlation between indicators, identify the triggering correlations that cause changes in the performance indicators of the target resource load to occur due to changes in the performance indicators of the business load. The triggering correlations are used to describe the path that causes fluctuations in the processing resource load due to business load events.

[0020] Based on the triggering relationships, construct the causal tracing chain corresponding to the resource bottleneck type.

[0021] In one embodiment, the step of calculating the resource load performance index density of each spatial unit of the digital twin system and determining the bottleneck spatial range based on the causal chain includes:

[0022] By calling upon the spatial index information of the digital twin system, the digital twin system is divided into multiple spatial units;

[0023] Based on the causal chain, the target resource load performance index in each spatial unit is statistically analyzed, and the resource load performance index density in each spatial unit is calculated.

[0024] The calculated resource load performance index density is compared with a preset density threshold, and the spatial units that exceed the preset density threshold are identified as the bottleneck spatial range.

[0025] In one embodiment, the steps of assessing the load contribution of each business object to the processing resource bottleneck within the bottleneck space and locating the bottleneck business object include:

[0026] Within the bottleneck space, obtain the target resource load performance metrics for each business object;

[0027] Based on the target resource load performance index of each business object and its proportion of the total resource load within the bottleneck space, calculate the load contribution of each business object to the resource bottleneck.

[0028] Based on the order of load contribution from largest to smallest, select the top K business objects as bottleneck business objects, where K is a positive integer.

[0029] In one embodiment, the steps of generating a performance optimization strategy for the digital twin system based on the scenario type of the digital twin system, combined with the type of processing resource bottleneck, the scope of the bottleneck space, and the bottleneck business object, include:

[0030] Identify the scenario type of the digital twin system and call the corresponding optimization strategy library for the scenario type;

[0031] Based on the type of resource bottleneck, the scope of the bottleneck space, and the bottleneck business object, the performance optimization strategy of the digital twin system is generated using the optimization strategy library.

[0032] Secondly, this application provides a performance optimization device for digital twin systems, the device comprising:

[0033] The resource bottleneck type determination module is used to determine the type of resource bottleneck when the digital twin system is determined to be in a performance risk state based on key performance indicators. Key performance indicators represent the overall operating performance state of the digital twin system, and resource load performance indicators represent the load characteristics of the processing resources.

[0034] The causal tracing chain construction module is used to construct the corresponding causal tracing chain based on the business load performance indicators corresponding to the processing resource bottleneck type and the corresponding target resource load performance indicators. The business load performance indicators represent the business load characteristics that trigger fluctuations in processing resources, and the causal tracing chain represents the triggering path that causes the processing resource bottleneck.

[0035] The bottleneck business object location module is used to calculate the resource load performance index density of each spatial unit of the digital twin system based on the causal tracing chain, determine the bottleneck space range, and within the bottleneck space range, evaluate the load contribution of each business object to the processing of the resource bottleneck, and locate the bottleneck business object.

[0036] The performance optimization strategy generation module is used to generate performance optimization strategies for the digital twin system based on the scenario type of the digital twin system, combined with the type of processing resource bottleneck, the scope of the bottleneck space, and the bottleneck business object.

[0037] Thirdly, this application provides a storage medium storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of any of the performance optimization methods for digital twin systems described in the above embodiments.

[0038] Fourthly, this application provides a computer device, including: one or more processors, and a memory;

[0039] The memory stores computer-readable instructions that, when executed by one or more processors, perform the steps of any of the performance optimization methods for digital twin systems described in the above embodiments.

[0040] As can be seen from the above technical solutions, the embodiments of this application have the following advantages:

[0041] The performance optimization method for digital twin systems provided in this application effectively solves the technical problem of existing technologies that do not perform performance diagnosis and optimization for digital twin systems. First, it identifies the overall operational status based on key performance indicators. When the digital twin system is in a performance risk state, it determines the type of resource bottleneck by combining resource load performance indicators, thereby achieving accurate identification of performance problems. Then, based on the business load performance indicators corresponding to the resource bottleneck type and the target resource load performance indicators, a causal tracing chain is constructed to clarify the business triggering path that causes the bottleneck, achieving a fine mapping between business load and resource bottleneck. Further, the resource load performance indicator density of each spatial unit is calculated through the causal tracing chain to determine the bottleneck spatial range, and the contribution of each business object to the resource bottleneck is evaluated within this range, thereby achieving accurate location of the bottleneck business object. Finally, combined with the scenario type of the digital twin system, the resource bottleneck type, bottleneck spatial range, and bottleneck business object information are comprehensively processed to generate targeted performance optimization strategies. Therefore, this method can improve the accuracy of performance problem location and the targeting of optimization strategies for digital twin systems, effectively improving the overall operational performance of the digital twin system. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 A flowchart illustrating a performance optimization method for digital twin systems provided in an embodiment of this application;

[0044] Figure 2 A schematic diagram of the structure of a performance optimization device for digital twin systems provided in an embodiment of this application;

[0045] Figure 3 This is a schematic diagram of the internal structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0046] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0047] This application provides a performance optimization method for digital twin systems. The following embodiments illustrate this method using a computer device as an example. It is understood that the computer device can be any device with data processing capabilities, including but not limited to a single server, server cluster, personal laptop, desktop computer, etc. Figure 1 As shown, the method includes:

[0048] S101: When a digital twin system is determined to be in a performance risk state based on key performance indicators, the type of processing resource bottleneck is determined according to the resource load performance indicators. Key performance indicators represent the overall operating performance state of the digital twin system, and resource load performance indicators represent the load characteristics of processing resources.

[0049] Key performance indicators (KPIs) represent the overall operational performance status of the digital twin system, reflecting whether there are any abnormalities such as performance degradation, resource bottlenecks, or response delays during system operation. Resource load performance indicators represent the load characteristics of processing resources, specifically including data related to system resource consumption such as processor utilization, memory usage, storage access frequency, and network bandwidth usage. Processing resource bottleneck types indicate the categories of critical processing resources that are constrained in the digital twin system at a specific time, such as computing resource bottlenecks, storage resource bottlenecks, or network resource bottlenecks. A digital twin system in a performance risk state means that KPIs have reached preset thresholds, indicating potential performance problems that require timely analysis and optimization.

[0050] When a digital twin system is running, key performance indicators (KPIs) are continuously monitored to determine its overall performance status. These KPIs can include system response time, throughput, task completion rate, and the frequency of abnormal events. When monitoring results show that KPIs reach or exceed preset performance thresholds, the digital twin system is determined to be in a performance risk state, indicating that the system may have potential performance degradation, resource bottlenecks, or response latency issues, requiring timely analysis and handling.

[0051] After confirming that the digital twin system is in a performance risk state, the usage of various processing resources is analyzed based on resource load performance indicators. Resource load performance indicators reflect the real-time load characteristics of processing resources, including processor utilization, memory usage, storage read / write operations, and network bandwidth usage. By collecting these indicators in real time, the load levels of processing resources are statistically analyzed, such as calculating the average, peak, and fluctuation trends of resource utilization, to determine whether any resource is overloaded or approaching a bottleneck.

[0052] Subsequently, based on the analysis results, the type of resource bottleneck is determined, that is, the category of restricted key resources in the digital twin system is identified, such as computing resource bottlenecks, memory resource bottlenecks, storage resource bottlenecks, or network resource bottlenecks. In the process of determining the bottleneck type, real-time load data can be compared with historical operating data or preset target performance indicators to analyze whether the current load is abnormal or exceeds the expected range, thereby accurately locating the performance-constrained resources. By recording the bottleneck type and the corresponding resource load characteristics, a reliable data foundation can be provided for subsequent performance diagnosis and optimization.

[0053] By assessing the performance risk of a digital twin system based on key performance indicators during its operation, a unified perception and early warning of the system's overall performance can be achieved. This avoids blindly analyzing resource load before performance anomalies are observed, thus improving the targeting and efficiency of performance diagnosis. Furthermore, analyzing the load characteristics of processing resources using resource load performance indicators allows for the establishment of a correlation between overall performance risk and specific resource usage, accurately identifying the types of processing resource bottlenecks that trigger performance risks. This layered approach, refining from overall performance status to specific resource load, effectively avoids misjudgments caused by relying on a single resource indicator, enabling precise location of the causes of performance problems and providing a clear direction for subsequent performance optimization. This ultimately improves the accuracy and reliability of digital twin system performance diagnosis.

[0054] S102: Based on the business load performance indicators corresponding to the processing resource bottleneck type and the corresponding target resource load performance indicators, construct the corresponding causal tracing chain. The business load performance indicators represent the business load characteristics that trigger fluctuations in processing resources, and the causal tracing chain represents the triggering path that causes the processing resource bottleneck.

[0055] Among them, the business load performance index is a data indicator used to describe the load characteristics generated during business operations, reflecting the degree and manner in which business behavior affects processing resources. The target resource load performance index refers to the corresponding resource load performance index selected from multiple resource load performance indices based on the previously determined processing resource bottleneck type, used to characterize the load state of that processing resource. The causal chain is used to describe the correlation path in which changes in business load trigger fluctuations in processing resources step by step, ultimately forming a resource bottleneck, reflecting the causal relationship between business load characteristics and processing resource bottlenecks.

[0056] In this step, based on the identified bottleneck type of processing resources, the causes of abnormal processing resource load are further analyzed from the perspective of business operations, thereby establishing a causal relationship between changes in business load and processing resource bottlenecks. By introducing business load performance indicators and correlating them with the type of processing resource bottleneck, performance analysis can be extended from a single resource level to the business level, providing a foundation for subsequent accurate location of performance problems.

[0057] In practice, the first step is to select business load performance indicators that are directly related to the identified bottleneck type of processing resources. These selected indicators characterize the load features that affect this type of processing resource during business operations, such as the scale of business requests, processing complexity, or concurrency level. This ensures that the analysis accurately reflects the triggering effect of business behavior on fluctuations in processing resource load. This targeted indicator selection method avoids including irrelevant business loads in the analysis, ensuring the focus of causal analysis.

[0058] Subsequently, a correlation analysis was performed between the selected business load performance indicators and the corresponding target resource load performance indicators. By analyzing the relationship between changes in business load and deviations from the target resource load performance indicators, the temporal sequence and trend of the impact of business load changes on processing resource load can be identified. Based on this, the correlation between changes in business load characteristics, fluctuations in processing resource load, and the formation of resource bottlenecks was gradually sorted out and organized according to the order of triggering, thereby constructing a causal chain describing how business load triggers processing resource bottlenecks.

[0059] By identifying the types of processing resource bottlenecks and selecting corresponding business load performance indicators, and constructing a causal chain based on these indicators, a clear causal relationship can be established between abnormal processing resource load and specific business load characteristics. This avoids the problem of insufficient explanation of bottleneck causes relying solely on resource load data. By linking business load changes, processing resource fluctuations, and the bottleneck formation process step-by-step through the causal chain, the triggering path leading to processing resource bottlenecks can be clearly depicted, making the source of performance problems traceable and logically consistent. This helps improve the accuracy and interpretability of processing resource bottleneck cause analysis.

[0060] S103: Based on the causal chain, calculate the resource load performance index density of each spatial unit of the digital twin system, determine the bottleneck space range, and within the bottleneck space range, evaluate the load contribution of each business object to the processing resource bottleneck, and locate the bottleneck business object.

[0061] In this context, a spatial unit refers to the smallest analytical unit obtained by dividing the space structure or logical layout of a digital twin system, used to carry business objects and their corresponding resource load information. Resource load performance index density refers to the degree of aggregation of resource load performance indicators within a unit of spatial unit, reflecting the concentration of processing resource load within that unit. The bottleneck space range refers to the set of spatial units where the resource load performance index density is abnormally high, significantly impacting the formation of processing resource bottlenecks. A business object refers to an object entity in the digital twin system that performs specific business behaviors and consumes processing resources. Load contribution refers to the degree of influence of a business object within the bottleneck space range on the formation of processing resource bottlenecks.

[0062] After establishing the causal chain, the performance bottleneck is further mapped from the causal relationship level to specific spatial locations and business objects, giving the resource bottleneck a clear spatial and business orientation. To this end, based on the business load performance indicators and resource load performance indicators associated with the causal chain, the digital twin system is analyzed in a spatial dimension to identify the distribution of performance bottlenecks within the system's spatial structure.

[0063] In practical implementation, the digital twin system is divided into multiple spatial units according to its spatial partitioning rules, and the resource load performance indicators involved in the causal tracing chain are mapped to the corresponding spatial units. By statistically analyzing and calculating the resource load performance indicators within each spatial unit, the resource load performance indicator density of each spatial unit is obtained, and the density distribution between different spatial units is compared and analyzed. When the resource load performance indicator density of some spatial units is significantly higher than that of other spatial units, these spatial units can be identified as bottleneck spatial ranges, used to characterize the concentrated areas of resource bottlenecks in the spatial dimension.

[0064] After determining the bottleneck space, further analysis is conducted on the business objects within that space. By associating business objects with their corresponding resource load performance indicators, the load contribution of each business object to the bottleneck within the bottleneck space is assessed. The load contribution assessment can be calculated based on the magnitude and duration of resource load changes caused by the business object, as well as its correlation with the causal chain. By comparing the load contributions of different business objects, the business object with the greatest impact on the resource bottleneck can be identified, thereby locating the bottleneck business object.

[0065] It's important to note that in digital twin systems, processing resource bottlenecks are typically the result of gradual amplification of business load changes through system scheduling and resource contention, superimposed within a specific spatial range, rather than being directly generated by a single business object at a particular moment. Therefore, directly locating objects based solely on identifying processing resource bottlenecks lacks constraints on the bottleneck formation process, making it difficult to determine whether object load behavior truly contributed to bottleneck formation. This can easily lead to misclassifying occasional high-resource-consuming objects or irrelevant objects as bottlenecks. By first constructing a causal chain based on the type of processing resource bottleneck, the complete path from business load changes triggering processing resource fluctuations to the formation of a bottleneck can be clearly identified, thus excluding business loads without a direct causal relationship to the bottleneck. Based on this, spatial location is then used to determine the concentrated spatial range of the processing resource bottleneck within the digital twin system, introducing spatial constraints and narrowing the analysis scope. Ultimately, only by evaluating the load contribution of business objects within the determined triggering path and bottleneck spatial range can the business objects that substantially affect the processing resource bottleneck be accurately identified. This ensures that the bottleneck object location results have clear causal and spatial basis, improving the accuracy and reliability of the location.

[0066] By calculating the resource load performance index density of each spatial unit based on causal chain tracing, and evaluating the load contribution of each business object to the bottleneck within the determined bottleneck space, the bottleneck can be progressively refined from the causal relationship level to specific spaces and specific business objects, giving performance problems a clear spatial location and business origin. This hierarchical processing approach avoids misjudgments that occur when analyzing only overall load data or a single business object, improves the accuracy and interpretability of bottleneck location, and provides a clear and reliable basis for subsequent performance optimization, making the performance diagnosis of the digital twin system more refined and operable.

[0067] S104: Based on the scenario type of the digital twin system, combined with the type of processing resource bottleneck, the scope of the bottleneck space, and the bottleneck business object, generate a performance optimization strategy for the digital twin system.

[0068] The scenario type of a digital twin system refers to the business and operational characteristics of the system under a specific operating environment, used to describe the system's operating conditions, task distribution, and resource usage characteristics. Performance optimization strategies refer to methods proposed for adjusting resource allocation, optimizing business load, or improving system scheduling for specific scenario types and system operating states, in order to alleviate or eliminate processing resource bottlenecks and improve the overall operating performance of the digital twin system.

[0069] In this step, based on the current scenario type of the digital twin system and combined with the processing resource bottleneck types, bottleneck spatial ranges, and bottleneck business objects obtained from the preceding analysis, targeted performance optimization strategies are formulated to improve the overall operational performance and resource utilization efficiency of the digital twin system. The scenario type of the digital twin system reflects the task distribution, business model, and resource usage characteristics of the digital twin system in a specific business environment. These characteristics determine the degree of impact and optimization priority of resource bottlenecks on the performance of the digital twin system under different scenarios.

[0070] In practical implementation, the first step is to analyze the current business load pattern, task execution order, and resource distribution based on the scenario type of the digital twin system, identifying the key factors that have the greatest impact on system performance in that scenario. Then, by combining the bottleneck type, bottleneck space range, and bottleneck business object information with scenario characteristics, a comprehensive analysis of resource allocation, task scheduling, and business load handling is conducted to generate optimization strategies. For example, the task execution order can be adjusted during peak load periods of bottleneck resources, or the resource consumption of business objects can be balanced within the bottleneck space, thereby alleviating resource pressure and improving overall system performance.

[0071] By generating performance optimization strategies based on scenario types within the digital twin system, combined with the types of resource bottlenecks, their spatial scope, and the bottleneck business objects, precise optimization measures can be implemented for resource bottlenecks in different business environments. This approach closely links performance optimization with the actual operating scenario of the digital twin system, the type of bottleneck, and the characteristics of the bottleneck object. This makes the optimization strategies targeted and operable, effectively mitigating the impact of bottlenecks on overall system performance, improving resource utilization efficiency and system stability, while providing a reliable basis for subsequent performance adjustments and long-term operation.

[0072] The above embodiments effectively address the technical problem in existing technologies that do not perform performance diagnosis and optimization for digital twin systems. First, the overall operational status is identified based on key performance indicators. When the digital twin system is in a performance risk state, the resource load performance indicators are combined to determine the type of resource bottleneck, thereby achieving accurate identification of performance problems. Then, based on the business load performance indicators corresponding to the resource bottleneck type and the target resource load performance indicators, a causal tracing chain is constructed to clarify the business triggering path that causes the bottleneck, achieving a fine mapping between business load and resource bottleneck. Further, the resource load performance indicator density of each spatial unit is calculated through the causal tracing chain to determine the bottleneck spatial range, and the contribution of each business object to the resource bottleneck is evaluated within this range, thereby achieving accurate location of the bottleneck business object. Finally, combined with the scenario type of the digital twin system, the resource bottleneck type, bottleneck spatial range, and bottleneck business object information are comprehensively processed to generate targeted performance optimization strategies. Therefore, this method can improve the accuracy of performance problem location and the targeting of optimization strategies for digital twin systems, effectively improving the overall operational performance of the digital twin system.

[0073] In one embodiment, the process of determining whether a digital twin system is in a performance risk state based on key performance indicators includes:

[0074] Time-series monitoring of key performance indicators;

[0075] If, based on preset performance thresholds, it is determined that key performance indicators meet the instantaneous conditions for performance anomalies, or, based on time-series monitoring results, it is determined that key performance indicators meet the conditions for abnormal performance trends, then the digital twin system is determined to be in a performance risk state.

[0076] Among them, the performance threshold is a pre-set reference value for key performance indicators (KPIs) to determine whether the KPIs are within the normal range. The instantaneous performance anomaly condition refers to a KPI deviating from the preset performance threshold at a specific point in time or within a short period, indicating that the digital twin system is currently in a performance anomaly state. The performance anomaly trend condition refers to a KPI showing a continuous downward, upward, or abnormally fluctuating trend over a continuous period, reflecting a potential performance degradation or instability risk in the digital twin system.

[0077] In this embodiment, by monitoring key performance indicators over time, the operating status of the digital twin system can be monitored in real time. This allows for timely identification of performance risks when the digital twin system exhibits immediate or potential abnormal trends, providing a basis for subsequent performance analysis and optimization.

[0078] In practice, key performance indicators are first continuously collected and recorded to form complete time-series data, capturing the performance changes of the digital twin system at different points in time. Then, the monitoring results are compared with preset performance thresholds. However, the judgment logic is not simply based on whether the values ​​are above or below the thresholds as the sole criterion for anomalies. Instead, judgment conditions are set according to the characteristics of the key indicators and the operational goals of the digital twin system. For example, some indicators may be judged as abnormal even if they do not reach the threshold, while others are only considered abnormal when they exceed the threshold, thus flexibly adapting to different types of performance indicators and system requirements. The judgment that meets this condition is the instantaneous condition for performance anomaly.

[0079] Simultaneously, the time-series changes of key performance indicators (KPIs) are analyzed to determine whether there are trends of continuous increase, continuous decrease, or abnormal fluctuations. This trend determination is used to identify potential performance degradation or instability risks in the digital twin system. By performing statistical and trend analysis on time-series data, potential problems can be detected in advance even if the KPIs have not yet triggered threshold judgments. When the monitoring results show that the instantaneous conditions for abnormal performance or the conditions for abnormal performance change trends are met, the digital twin system is determined to be in a performance risk state, triggering subsequent resource bottleneck analysis, causal tracing, and performance optimization.

[0080] By monitoring key performance indicators (KPIs) over time and combining this with assessments of both instantaneous and trend-based performance anomalies, a comprehensive and dynamic evaluation of the digital twin system's operational status can be achieved. Instantaneous performance anomaly conditions identify instances where KPIs deviate from preset performance thresholds at a single point in time or over a short period, enabling rapid detection of immediate system performance issues. Trend-based performance anomaly conditions capture trends of continuous increases, decreases, or abnormal fluctuations in KPIs over a period of time, thereby identifying potential performance degradation or instability risks in advance. This approach, combining instantaneous anomalies with trend analysis, not only promptly detects current anomalies but also provides early warnings of potential future performance problems. This provides a reliable basis for subsequent resource bottleneck analysis, causal tracing, and performance optimization, ultimately improving the operational stability, reliability, and overall performance level of the digital twin system.

[0081] In one embodiment, the step of determining the type of resource bottleneck to be addressed based on resource load performance metrics includes:

[0082] Multiple resource load performance indicators are obtained, and the bottleneck type of the processing resource is determined based on the degree of deviation of each resource load performance indicator from the corresponding preset load threshold. Each resource load performance indicator corresponds to the load characteristics of different processing resources.

[0083] Among them, the preset load threshold is a reference value set for various resource load performance indicators, used to determine whether the resource load is within the normal range or abnormally high or low.

[0084] In this embodiment, by acquiring the load performance indicators of various processing resources in the digital twin system, potential performance bottlenecks in the system are identified, providing a basis for subsequent performance analysis and optimization. First, continuous, real-time performance monitoring is performed on various processing resources in the digital twin system, collecting data including processor utilization, memory usage, storage read / write load, network throughput, and load performance indicators of GPUs, caches, and other specific computing or storage resources. This collected data comprehensively reflects the usage and stress levels of different resources during system operation. To ensure the stability and comparability of the monitoring data, the collected data undergoes smoothing, normalization, or outlier filtering to eliminate the impact of instantaneous fluctuations and noise, allowing the load data of different resources to be analyzed under the same standard.

[0085] Subsequently, each resource load performance indicator is compared with its corresponding preset load threshold one by one. By calculating the degree of deviation from the threshold and the duration of the deviation, the degree of load anomaly of the resource is assessed. For resources with large deviations or long durations, they are identified as potential performance bottleneck candidates. Further analysis, combining historical operating data and trends, determines whether the resource load anomaly is an occasional fluctuation, a short-term fluctuation, or a long-term persistent bottleneck. Through this multi-dimensional analysis, occasional load fluctuations and key bottleneck resources that truly affect the overall performance of the digital twin system can be effectively distinguished.

[0086] After completing the deviation analysis and trend judgment of the load of each resource, resources with significantly abnormal loads are classified and marked to clarify the current bottleneck type of the digital twin system. During this process, the determination of bottleneck types can be further refined by combining the business relevance, priority, and impact weight of each resource on overall system performance. This ensures that the determination results not only reflect the load status of the resource itself but also its actual constraint on system performance. Ultimately, by accurately identifying the bottleneck type of processing resources, a reliable basis can be provided for subsequent causal tracing analysis, spatial scope positioning, and performance optimization strategy generation, thereby achieving targeted optimization and improving the overall system performance and operational stability.

[0087] By acquiring multiple resource load performance indicators and determining the bottleneck type of each indicator based on its deviation from the corresponding preset load threshold, the impact of different resources on the overall performance of a digital twin system can be accurately identified. Each resource load performance indicator reflects the usage intensity and load characteristics of different resources such as processors, memory, storage, and networks. By analyzing their deviation from preset load thresholds, it is possible to distinguish which resources have abnormally significant loads, thereby identifying the key resources limiting the performance of the digital twin system. This approach can simultaneously identify bottleneck resources with instantaneous high loads and long-term deviation trends, avoiding the omission of resources that substantially restrict the performance of the digital twin system. This provides a reliable basis for subsequent causal analysis, spatial location, and performance optimization, contributing to improved system stability, efficiency, and overall performance.

[0088] In one embodiment, the step of constructing a corresponding causal chain based on the business load performance index corresponding to the processing resource bottleneck type and the corresponding target resource load performance index includes:

[0089] Determine the business load performance metrics corresponding to the processing resource bottleneck type, and the corresponding target resource load performance metrics.

[0090] Analyze the temporal sequence and correlation between target resource load performance indicators and business load performance indicators;

[0091] Based on the chronological order and the correlation between indicators, identify the triggering correlations that cause changes in the performance indicators of the target resource load to occur due to changes in the performance indicators of the business load. The triggering correlations are used to describe the path that causes fluctuations in the processing resource load due to business load events.

[0092] Based on the triggering relationships, construct the causal tracing chain corresponding to the resource bottleneck type.

[0093] The temporal sequence refers to the order in which changes in business load performance indicators and target resource load performance indicators occur during the operation of the digital twin system. This determines whether a particular business load event precedes the fluctuation of the corresponding resource load in time, thus identifying the possible causal direction. The correlation between indicators refers to the degree of correlation between the numerical values ​​or trends of business load performance indicators and target resource load performance indicators. This quantifies the strength and direction of the impact of business load changes on resource load fluctuations, thereby assisting in identifying triggering correlations. The triggering correlation refers to the path through which changes in business load performance indicators lead to changes in target resource load performance indicators, used to characterize the causal relationship between business events and resource fluctuations.

[0094] In this embodiment, business load events that cause processing resource bottlenecks are identified, and the causal relationship between these business load events and resource load fluctuations is clarified to provide a reliable basis for subsequent performance optimization. First, the identified processing resource bottleneck types are matched with their corresponding business load performance indicators and target resource load performance indicators. This matching process clarifies which business load events or operational behaviors may affect each type of processing resource bottleneck, and identifies which resource load indicators require focused attention, thus laying the foundation for subsequent analysis.

[0095] Subsequently, a detailed analysis was conducted on the historical and real-time time-series data of business load performance indicators and target resource load performance indicators. First, through temporal sequence analysis, it was determined whether business load events preceded changes in the corresponding resource load, thus providing a preliminary assessment of potential triggering paths. Then, combined with indicator correlation analysis, the intensity and direction of the impact of business load changes on target resource load fluctuations were quantified. By comprehensively considering both temporal sequence and indicator correlation, the key business load events that truly triggered abnormal resource loads can be identified, while occasional or irrelevant load fluctuations can be excluded.

[0096] After identifying key triggering relationships, a causal chain is constructed based on the chronological order and impact of each event to correspond to the type of resource bottleneck. This chain sequentially links each business load event with its resulting resource load fluctuations, forming a clear chain structure that visually demonstrates how changes in business load gradually propagate to resource bottlenecks. This causal chain not only reveals the root cause of resource bottlenecks but also provides a clear direction for subsequent bottleneck location and performance optimization strategy development.

[0097] For example, in a digital twin manufacturing system, if a server's CPU utilization is consistently high, analyzing the business load performance indicators corresponding to the CPU bottleneck can reveal that high-frequency data acquisition tasks and batch computing jobs are likely key factors. Time-series analysis shows that the data acquisition tasks are initiated before the CPU utilization increases, and correlation analysis indicates a high degree of correlation between the two. Therefore, it can be determined that data acquisition tasks trigger CPU load fluctuations. Based on these triggering relationships, a causal chain can be constructed, linking data acquisition tasks to the CPU bottleneck, thus providing a clear basis for adjusting task scheduling and optimizing resource allocation.

[0098] It's important to note that in digital twin systems, fluctuations in processing resource load are typically caused by the combined effects of multiple business load events. Simply observing resource load anomalies cannot determine the root cause. To construct an accurate causal chain, it's necessary to first identify the temporal sequence, as causal relationships have a clear temporal direction. Only when a business load event precedes a resource load fluctuation can it be preliminarily identified as a potential triggering factor. Secondly, it's crucial to analyze the correlation between metrics. By quantifying the correlation between business load performance metrics and target resource load performance metrics, we can identify the business events that truly impact resource load fluctuations, excluding occasional or irrelevant load changes. Based on the temporal sequence and metric correlations, we can further identify triggering relationships, clarifying which changes in business load events directly caused resource load fluctuations. This triggering relationship reveals the causal path between events and resource load. Only after clarifying the triggering relationships can we sequentially connect business load events according to their temporal order and impact relationships, thereby constructing a clear and accurate causal chain. This visually demonstrates how business load gradually leads to processing resource bottlenecks, enabling precise tracing and targeted optimization of system performance issues.

[0099] By identifying the business load performance indicators and target resource load performance indicators corresponding to the types of resource bottlenecks, and analyzing the temporal sequence and correlation between them, the triggering relationships of resource load fluctuations caused by changes in business load can be accurately identified, thus constructing a clear causal chain. This approach can precisely depict the actual impact path of business load events on resource bottlenecks, avoiding mislocation caused by relying solely on resource anomalies, and distinguishing between occasional fluctuations and critical events that truly affect system performance. By establishing a causal chain, it is possible to intuitively reveal how changes in business load gradually lead to resource bottlenecks, providing a reliable basis for subsequent bottleneck localization, performance optimization, and resource scheduling, thereby improving the controllability, stability, and optimization effectiveness of the overall performance of the digital twin system.

[0100] In one embodiment, the step of calculating the resource load performance index density of each spatial unit of the digital twin system and determining the bottleneck spatial range based on the causal chain includes:

[0101] By calling upon the spatial index information of the digital twin system, the digital twin system is divided into multiple spatial units;

[0102] Based on the causal chain, the target resource load performance index in each spatial unit is statistically analyzed, and the resource load performance index density in each spatial unit is calculated.

[0103] The calculated resource load performance index density is compared with a preset density threshold, and the spatial units that exceed the preset density threshold are identified as the bottleneck spatial range.

[0104] Spatial index information refers to the coordinates, boundaries, or topological relationships of each spatial unit in a digital twin system, which is used to divide and locate the physical or logical space of the digital twin system.

[0105] In this embodiment, the resource load in the digital twin system is analyzed in detail along the spatial dimension to identify key areas where resource load is abnormally concentrated and may form performance bottlenecks. First, the spatial index information of the digital twin system is invoked to spatially partition the system, decomposing the entire digital twin system into multiple analyzable spatial units. Each spatial unit can correspond to a physical region, logical region, or set of business objects in the digital twin system. This partitioning enables regional location and monitoring of resource load, providing a basic data structure for subsequent statistical analysis.

[0106] After completing the spatial unit division, the key to conducting spatial dimension analysis of the digital twin system using causal attribution chains lies in mapping the causal relationship between business load events and resource load fluctuations to each spatial unit. First, the causal attribution chain clarifies the source of each resource bottleneck, identifying which business load events triggered load fluctuations in specific processing resources and recording the trigger paths between these events and changes in resource metrics. This allows for the attribution of resource loads affected by the causal chain to specific spatial units when statistically analyzing resource load within a spatial unit, thereby achieving precise location of areas with concentrated resource loads.

[0107] In practical implementation, each business load event recorded in the causal tracing chain and its corresponding target resource load index can be traversed. The resource consumption or load changes caused by each event in its spatial unit can be statistically analyzed, and these load changes can be accumulated or weighted to obtain the resource load performance index density of each spatial unit. In this way, resource load statistics not only reflect the spatial load distribution but also incorporate information from the business triggering path, thereby more accurately identifying which spatial units' loads are bottlenecks caused by specific business events.

[0108] Subsequently, the calculated resource load density of spatial units is compared with a preset threshold, and spatial units exceeding the threshold are selected as bottleneck spatial ranges. Utilizing a causal chain ensures that these identified bottleneck spaces are triggered by actual business load events, rather than occasional random load fluctuations, thereby improving the accuracy and reliability of bottleneck location and providing a targeted spatial range for subsequent business object analysis and performance optimization. Through this method, resource bottleneck problems can be precisely decomposed from the overall system level to specific spatial units, enabling controllable monitoring and optimization of high-load areas in the digital twin system.

[0109] By accessing the spatial index information of the digital twin system, the system is divided into multiple spatial units. Combined with a causal chain, the performance indicators of target resource loads within each spatial unit are statistically analyzed and their density is calculated. This allows for the accurate identification of critical areas where resource loads are abnormally concentrated, potentially forming performance bottlenecks. Comparing the calculated density with a preset threshold allows for the selection of spatial units that truly have a significant impact on the overall performance of the digital twin system, thus pinpointing bottlenecks from the overall system level to a specific spatial range. This approach avoids misjudgments caused by occasional fluctuations or abnormal single indicators. Furthermore, by combining spatial distribution with business trigger paths, it clearly reveals concentrated areas of resource load and their causes, providing a reliable basis for subsequent location and performance optimization of bottleneck business objects, thereby improving the accuracy and controllability of digital twin system performance management.

[0110] In one embodiment, the steps of assessing the load contribution of each business object to the processing resource bottleneck within the bottleneck space and locating the bottleneck business object include:

[0111] Within the bottleneck space, obtain the target resource load performance metrics for each business object;

[0112] Based on the target resource load performance index of each business object and its proportion of the total resource load within the bottleneck space, calculate the load contribution of each business object to the resource bottleneck.

[0113] Based on the order of load contribution from largest to smallest, select the top K business objects as bottleneck business objects, where K is a positive integer.

[0114] In this embodiment, within the identified bottleneck space, the contribution of specific business objects to the formation of resource bottlenecks is further analyzed, thereby achieving precise location of performance bottlenecks. First, the target resource load performance indicators corresponding to each business object are obtained from the identified bottleneck space. These indicators are collected, organized, and archived to form resource load data for each business object within that space. This step includes not only reading real-time monitoring data but also summarizing historical load data to ensure a comprehensive and accurate description of the resource consumption of each business object. Subsequently, based on the target resource load performance indicators of each business object, its proportion in the total resource load within the bottleneck space is calculated, thus obtaining the load contribution of each business object to handling the resource bottleneck. During the calculation process, the resource load indicators of each business object can be compared with the total load of all business objects within that space, and data from different time points can be weighted or accumulated to reflect the contribution of business objects to resource bottlenecks at different operational stages.

[0115] After calculating the load contribution, all business objects are sorted from highest to lowest contribution to identify which objects occupy the majority of the resource load within the bottleneck space. Based on the sorting results, the top K business objects with the highest contribution are selected as bottleneck business objects, where K is a positive integer. These objects are marked and recorded to provide operational targets for subsequent performance optimization, load adjustment, or resource scheduling. In practice, the sorting and selection process can be refined by considering the running status, type, and priority of the business objects to ensure that the selected bottleneck business objects cover the main load and have optimization value.

[0116] Throughout the process, time-series analysis can be performed on the load contribution changes of each business object to identify potential short-term high-load events or long-term stable load trends. This allows the identification of bottleneck business objects to consider not only static proportions but also dynamic load changes, thus providing a more accurate and comprehensive analytical basis for subsequent performance optimization and enabling refined management and controllable intervention of resource bottlenecks in the digital twin system.

[0117] By acquiring the target resource load performance indicators for each business object within the bottleneck space, and calculating the load contribution of each business object to the bottleneck based on these indicators and their proportion in the total resource load within the space, the resource bottleneck problem can be precisely decomposed from the overall load of the space to specific business objects, thereby quantifying the actual impact of each business object in the bottleneck formation. By sorting the load contributions of each business object from largest to smallest and selecting the top K most contributing business objects as the bottleneck business objects, the main sources of load can be identified, providing key targets for subsequent performance optimization and resource scheduling. This approach ensures that optimization measures focus on the business objects that have the greatest impact on system performance, making resource allocation and performance tuning more targeted and efficient, thereby improving the overall operating performance and resource utilization efficiency of the digital twin system.

[0118] In one embodiment, the steps of generating a performance optimization strategy for the digital twin system based on the scenario type of the digital twin system, combined with the type of processing resource bottleneck, the scope of the bottleneck space, and the bottleneck business object, include:

[0119] Identify the scenario type of the digital twin system and call the corresponding optimization strategy library for the scenario type;

[0120] Based on the type of resource bottleneck, the scope of the bottleneck space, and the bottleneck business object, the performance optimization strategy of the digital twin system is generated using the optimization strategy library.

[0121] The optimization strategy library refers to a collection of pre-stored performance optimization schemes designed for different scenario types and resource bottleneck types. Each optimization strategy includes specific resource scheduling rules, load balancing methods, or business object optimization measures.

[0122] In this embodiment, the scenario type of the digital twin system is first identified to clarify the current business environment and application conditions in which the digital twin system operates. This can be achieved by analyzing the operating parameters of the digital twin system, the activity patterns of business objects, and the usage characteristics of spatial units. The identification results can classify the digital twin system into a predefined scenario type, such as a high-concurrency computing scenario, a data-intensive processing scenario, or a mixed load scenario. Subsequently, based on the identified scenario type, the corresponding optimization strategy library is invoked to retrieve a set of performance optimization strategies applicable to that scenario. Each strategy stored in the optimization strategy library is associated with a scenario type and covers scheduling schemes, load balancing methods, and optimization measures for key business objects for different types of processing resource bottlenecks.

[0123] Next, based on the identified bottleneck types, bottleneck spatial ranges, and bottleneck business objects, the acquired set of strategies is filtered and matched to generate performance optimization strategies suitable for the current system state. During this process, strategies that can alleviate major bottlenecks are prioritized based on the type and severity of the bottleneck resources. Simultaneously, the application scope and priority of optimization schemes are adjusted according to the resource distribution within the bottleneck spatial range, ensuring that optimization measures are effective in critical areas. For bottleneck business objects, specific business scheduling or load adjustment strategies can be selected based on their load contribution and operational characteristics to further reduce resource pressure and improve overall performance.

[0124] After generating optimization strategies, these strategies can be transformed into specific executable instructions and applied to resource scheduling and business object management within the digital twin system. This includes adjusting resource allocation ratios, migrating tasks to less loaded space units, and adjusting the execution order or concurrency of business objects, thereby achieving an overall improvement in system performance. During execution, the effects of the strategies can be monitored and adjusted in real time, dynamically optimizing the application of the strategies to adapt to changes in system operating states and ensuring that the digital twin system maintains efficient and stable operation in different scenarios.

[0125] By identifying the scenario types of the digital twin system and invoking the corresponding optimization strategy library, the most suitable performance optimization scheme can be selected for different business environments and operating conditions, ensuring a high degree of match between optimization measures and the current state of the digital twin system. By combining the type of processing resource bottlenecks, the scope of the bottleneck space, and the bottleneck business objects to generate performance optimization strategies, the optimization focus can be precisely located on the resources and business objects that have the greatest impact on the performance of the digital twin system, making resource scheduling and business execution adjustments targeted and prioritized. This approach ensures that the performance optimization strategy covers the entire system while focusing on key bottlenecks, thereby achieving rational utilization of digital twin system resources, improved operational performance, and efficient business processing. Simultaneously, it can dynamically adapt to operational changes under different scenarios, improving the stability and reliability of the digital twin system.

[0126] To facilitate understanding of the scheme in this application, specific examples are provided below.

[0127] This example solution uses a three-layer architecture—an enhanced data acquisition layer, a deep analysis and decision-making layer, and a precise execution output layer—as its core to build a performance optimization closed loop for the digital twin system. The functions and processes of each layer are closely connected, and the logic is clear.

[0128] The enhanced data acquisition layer, based on the collection of general rendering metrics, integrates core performance metrics of the digital twin scenario. General rendering metrics include FPS, absolute FPS, GPU Frame Time, Triangle Count (Tris), and Render Calls (DC). Core performance metrics include twin data throughput, physics simulation time, network coordination overhead, and object-level performance contribution. Twin data throughput is used to monitor the update frequency and data volume of real-time data sources such as IoT and business systems; physics simulation time measures the CPU time consumed by a single step of the physics engine simulation; network coordination overhead measures network synchronization latency and state synchronization packet rate in multi-user collaborative scenarios; and object-level performance contribution estimates the contribution ratio of the number of rendered faces, materials, and script complexity to the overall load by assigning unique IDs to key business objects. This data provides comprehensive support for subsequent analysis and decision-making.

[0129] The deep analysis decision layer serves as the core processing stage. First, the performance status assessment and trend prediction module performs preliminary analysis of key indicators. This module takes smoothed FPS as input and compares it with a configurable threshold T_fps. If the FPS is not lower than the threshold, normal monitoring continues; if it is lower than the threshold, an abnormal status is determined, triggering an enhanced performance indicator correlation analysis model. Simultaneously, the module adds a trend analysis function, monitoring indicators such as polygon growth rate and sudden increases in data throughput through a sliding window. If it predicts that these indicators will trigger performance thresholds within the next N frames, an early warning is issued and diagnostics are initiated. The enhanced performance indicator correlation analysis model integrates digital twin context and proceeds with diagnosis in three stages. The first stage is bottleneck category separation, distinguishing between CPU and GPU bottlenecks using empirical rules. GPU bottlenecks are determined based on whether the difference between absoluteFPS and FPS exceeds a threshold, while CPU bottlenecks are determined based on the difference not exceeding the threshold and absoluteFPS being lower than T_fps. The second stage is causal chain tracing diagnosis. When a GPU bottleneck is diagnosed, not only is the direct cause determined, but the complete trigger path is also constructed. For example, a surge in twin data throughput triggers the dynamic loading of new models, leading to a sharp increase in the number of active faces and causing GPU overload. The diagnostic conclusion will further trace the source back to the dynamic model loading triggered by real-time data, resulting in a geometric bottleneck and pinpointing the triggering source data stream. If the diagnosis is a CPU bottleneck, it will analyze whether it is caused by twin-specific logic such as physical calculations or data parsing. The third stage is spatial semantics and object-level localization. When the diagnosis involves geometric or rendering bottlenecks, it calls the spatial index information of the digital twin scene to calculate the performance index density of each spatial unit. If the GPU bottleneck is present and the face density exceeds the spatial threshold, it will pinpoint the local spatial bottleneck range and, combined with object-level performance contribution data, locate the top K business objects with the highest contribution to the bottleneck. After completing the three-stage diagnosis, the scene adaptive strategy selection module dynamically adjusts the diagnostic threshold based on the preset or learned scene performance feature profile and prioritizes recommending optimization strategy libraries adapted to the current scene. For example, in a scene with a large number of people, it will improve the early warning sensitivity and prioritize recommending strategies that simplify the LOD of the person model.

[0130] The precise execution output layer, serving as the implementation stage, upgrades the results from the analysis and decision-making layer into a multi-dimensional strategy library. The generated suggestions integrate causal chains, spatial scope, and object information. For example, for a geometric bottleneck in conveyor belt group-05 within production workshop A caused by sensor data triggering model loading, specific suggestions will be provided, including block loading, dynamic LOD settings, and sensor data filtering optimization. Both visualization and automatic optimization output methods are supported. Visualized suggestions precisely pinpoint the optimization direction to specific areas, objects, and data sources and deliver them to developers or maintenance personnel. Automatic optimization adaptively selects strategies based on the scenario type. For instance, in an inspection demonstration scenario encountering a fill rate bottleneck, it automatically reduces shadow quality; in a data analysis scenario, it prioritizes model accuracy and reduces post-processing effects. After strategy execution, the system monitors performance changes. If performance improves, the current strategy is solidified and continuously monitored; if no improvement is achieved, the strategy is rolled back and alternative solutions are tried. The relevant results are fed back to the analysis and decision-making layer to further optimize subsequent diagnosis and strategy selection, ultimately forming a complete closed loop of data collection, analysis and decision-making, and execution feedback.

[0131] The performance optimization apparatus for digital twin systems provided in the embodiments of this application will be described below. The performance optimization apparatus for digital twin systems described below can be referred to in correspondence with the performance optimization method for digital twin systems described above. Figure 2 As shown, this application provides a performance optimization device for digital twin systems, the device comprising:

[0132] The resource bottleneck type determination module 201 is used to determine the resource bottleneck type based on the resource load performance index when the digital twin system is determined to be in a performance risk state based on key performance indicators. The key performance indicators represent the overall operating performance state of the digital twin system, and the resource load performance index represents the load characteristics of the processing resources.

[0133] The causal tracing chain construction module 202 is used to construct the corresponding causal tracing chain based on the business load performance index corresponding to the processing resource bottleneck type and the corresponding target resource load performance index. The business load performance index represents the business load characteristics that trigger the fluctuation of processing resources, and the causal tracing chain represents the triggering path that causes the processing resource bottleneck.

[0134] The bottleneck business object location module 203 is used to calculate the resource load performance index density of each spatial unit of the digital twin system according to the causal tracing chain, determine the bottleneck space range, and within the bottleneck space range, evaluate the load contribution of each business object to the processing of the resource bottleneck, and locate the bottleneck business object.

[0135] The performance optimization strategy generation module 204 is used to generate performance optimization strategies for the digital twin system based on the scenario type of the digital twin system, combined with the processing resource bottleneck type, bottleneck space range, and bottleneck business object.

[0136] In one embodiment, the resource bottleneck type determination module 201 includes:

[0137] The key performance indicator detection unit is used for time-series monitoring of key performance indicators;

[0138] The performance risk status determination unit is used to determine that the digital twin system is in a performance risk state if, based on a preset performance threshold, the key performance indicators meet the instantaneous conditions of performance anomaly, or, based on time-series monitoring results, the key performance indicators meet the conditions of performance anomaly change trend.

[0139] In one embodiment, the resource bottleneck type determination module 201 includes:

[0140] The resource bottleneck type determination unit is used to acquire various resource load performance indicators and determine the resource bottleneck type based on the degree of deviation of each resource load performance indicator from the corresponding preset load threshold. Each resource load performance indicator corresponds to the load characteristics of different processing resources.

[0141] In one embodiment, the causal attribution chain construction module 202 includes:

[0142] The business load performance index determination unit is used to determine the business load performance index corresponding to the processing resource bottleneck type, as well as the corresponding target resource load performance index.

[0143] The indicator correlation analysis unit is used to analyze the temporal sequence relationship and indicator correlation between target resource load performance indicators and business load performance indicators.

[0144] The triggering relationship identification unit is used to identify the triggering relationship that causes changes in the target resource load performance index to occur based on the time sequence and the index relationship. The triggering relationship is used to describe the path that causes fluctuations in the processing resource load due to business load events.

[0145] The causal tracing chain construction unit is used to construct a causal tracing chain corresponding to the type of resource bottleneck based on the triggering relationship.

[0146] In one embodiment, the bottleneck business object location module 203 includes:

[0147] Spatial unit division unit is used to call the spatial index information of the digital twin system and divide the digital twin system into multiple spatial units;

[0148] The resource load performance index density calculation unit is used to statistically analyze the target resource load performance index in each spatial unit based on the causal tracing chain and calculate the resource load performance index density in each spatial unit.

[0149] The bottleneck space range determination unit is used to compare the calculated resource load performance index density with a preset density threshold and determine the space unit that exceeds the preset density threshold as the bottleneck space range.

[0150] In one embodiment, the bottleneck business object location module 203 includes:

[0151] The target resource load performance index acquisition unit is used to acquire the target resource load performance index corresponding to each business object within the bottleneck space.

[0152] The load contribution calculation unit is used to calculate the load contribution of each business object to the processing resource bottleneck based on the target resource load performance index of each business object and the proportion of the total resource load within the bottleneck space.

[0153] The bottleneck business object selection unit is used to select the top K business objects as bottleneck business objects in descending order of their load contribution, where K is a positive integer.

[0154] In one embodiment, the performance optimization strategy generation module 204 includes:

[0155] The scene type identification unit is used to identify the scene type of the digital twin system and call the optimization strategy library corresponding to the scene type;

[0156] The performance optimization strategy generation unit is used to generate performance optimization strategies for the digital twin system based on the type of processing resource bottleneck, the scope of the bottleneck space, and the bottleneck business object, using the optimization strategy library.

[0157] In one embodiment, this application also provides a storage medium storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the performance optimization method for digital twin systems as described in any of the above embodiments.

[0158] In one embodiment, this application also provides a computer device storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the performance optimization method for a digital twin system as described in any of the above embodiments.

[0159] Indicatively, such as Figure 3 As shown, Figure 3 This is a schematic diagram of the internal structure of a computer device 300 provided in an embodiment of this application. The computer device 300 can be provided as a server. (Refer to...) Figure 3The computer device 300 includes a processing component 302, which further includes one or more processors, and memory resources represented by memory 301 for storing instructions, such as application programs, that can be executed by the processing component 302. The application programs stored in memory 301 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processing component 302 is configured to execute instructions to perform the performance optimization methods for digital twin systems described in any of the above embodiments.

[0160] The computer device 300 may also include a power supply component 303 configured to perform power management of the computer device 300, a wired or wireless network interface 304 configured to connect the computer device 300 to a network, and an input / output (I / O) interface 305. The computer device 300 may operate on an operating system stored in memory 301, such as Windows Server™, Mac OS X™, Unix™, Linux™, Free BSD™, or similar.

[0161] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0162] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. In this document, "a," "an," "the," "the," and "its" may also include plural forms unless the context clearly indicates otherwise. "Multiple" refers to at least two, such as 2, 3, 5, or 8, etc. "And / or" includes any and all combinations of the related listed items.

[0163] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can be referred to each other.

[0164] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A performance optimization method for a digital twin system, characterized in that, The method includes: When the digital twin system is determined to be in a performance risk state based on key performance indicators, the processing resource bottleneck type is determined according to the resource load performance indicator. The key performance indicators represent the overall operating performance state of the digital twin system, and the resource load performance indicator represents the load characteristics of the processing resources. Based on the business load performance index and the target resource load performance index corresponding to the processing resource bottleneck type, a corresponding causal tracing chain is constructed. The business load performance index represents the business load characteristics that trigger processing resource fluctuations, and the causal tracing chain represents the triggering path that causes the processing resource bottleneck. The spatial index information of the digital twin system is invoked to divide the digital twin system into multiple spatial units. According to the causal tracing chain, the target resource load performance index in each spatial unit is statistically analyzed, and the resource load performance index density in each spatial unit is calculated. The calculated resource load performance index density is compared with a preset density threshold to determine the spatial units that exceed the preset density threshold as the bottleneck space range. Within the bottleneck space range, the load contribution of each business object to the processing of the resource bottleneck is evaluated, and the bottleneck business object is located. Based on the scenario type of the digital twin system, combined with the processing resource bottleneck type, the bottleneck spatial range, and the bottleneck business object, a performance optimization strategy for the digital twin system is generated.

2. The performance optimization method for digital twin system according to claim 1, characterized in that, The process of determining whether the digital twin system is in a performance risk state based on key performance indicators includes: Time-series monitoring of the aforementioned key performance indicators; If, based on a preset performance threshold, the key performance indicator is determined to meet the instantaneous condition of performance anomaly, or, based on time-series monitoring results, the key performance indicator is determined to meet the condition of performance anomaly trend, then the digital twin system is determined to be in the performance risk state.

3. The performance optimization method for digital twin system of claim 1, wherein, The step of determining the type of resource bottleneck to be addressed based on resource load performance indicators includes: Multiple resource load performance indicators are obtained, and the bottleneck type of the processing resource is determined based on the degree of deviation of each resource load performance indicator from the corresponding preset load threshold, wherein each resource load performance indicator corresponds to the load characteristics of different processing resources.

4. The performance optimization method for digital twin system of claim 1, wherein, The step of constructing a corresponding causal chain based on the business load performance index corresponding to the processing resource bottleneck type and the corresponding target resource load performance index includes: Determine the business load performance index corresponding to the processing resource bottleneck type, and the corresponding target resource load performance index. Analyze the temporal sequence and correlation between the target resource load performance indicators and the service load performance indicators; Based on the time sequence and indicator correlation, identify the triggering correlation that causes the change in the target resource load performance indicator to lead to the change in the business load performance indicator. The triggering correlation is used to describe the path that the business load event leads to the fluctuation of the processing resource load. Based on the triggering relationship, construct the causal tracing chain corresponding to the processing resource bottleneck type.

5. The performance optimization method for digital twin system of claim 1, wherein, The step of assessing the load contribution of each business object to the processing resource bottleneck within the bottleneck space and locating the bottleneck business object includes: Obtain the target resource load performance index for each business object within the bottleneck space; Based on the target resource load performance index of each business object and its proportion of the total resource load within the bottleneck space, calculate the load contribution of each business object to the resource bottleneck. Based on the order of their respective load contributions from largest to smallest, the top K business objects are selected as the bottleneck business objects, where K is a positive integer.

6. The performance optimization method for digital twin system of claim 1, wherein, The step of generating a performance optimization strategy for the digital twin system based on the scenario type of the digital twin system, combined with the processing resource bottleneck type, the bottleneck spatial range, and the bottleneck business object, includes: Identify the scenario type of the digital twin system and call the optimization strategy library corresponding to the scenario type; Based on the bottleneck type, the bottleneck spatial range, and the bottleneck business object, the performance optimization strategy for the digital twin system is generated using the optimization strategy library.

7. A performance optimization apparatus for a digital twin system, the apparatus comprising: a performance optimization module configured to: determine a performance metric of the digital twin system; and optimize the performance metric of the digital twin system. The device includes: The resource bottleneck type determination module is used to determine the resource bottleneck type based on resource load performance indicators when the digital twin system is determined to be in a performance risk state based on key performance indicators. The key performance indicators represent the overall operating performance state of the digital twin system, and the resource load performance indicators represent the load characteristics of the processing resources. The causal tracing chain construction module is used to construct a corresponding causal tracing chain based on the business load performance index and the corresponding target resource load performance index corresponding to the processing resource bottleneck type. The business load performance index represents the business load characteristics that trigger fluctuations in processing resources, and the causal tracing chain represents the triggering path that causes the processing resource bottleneck. The bottleneck business object location module is used to call the spatial index information of the digital twin system, divide the digital twin system into multiple spatial units, count the target resource load performance index in each spatial unit according to the causal tracing chain, calculate the resource load performance index density in each spatial unit, compare the calculated resource load performance index density with a preset density threshold, determine the spatial unit that exceeds the preset density threshold as the bottleneck space range, and within the bottleneck space range, evaluate the load contribution of each business object to the processing of the resource bottleneck, and locate the bottleneck business object. The performance optimization strategy generation module is used to generate a performance optimization strategy for the digital twin system based on the scenario type of the digital twin system, combined with the processing resource bottleneck type, the bottleneck spatial range, and the bottleneck business object.

8. A storage medium characterized by: The storage medium stores computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the performance optimization method for digital twin systems as described in any one of claims 1 to 6.

9. A computer device, comprising: include: One or more processors, and memory; The memory has stored therein computer readable instructions that, when executed by the one or more processors, perform the steps of the method for performance optimization of a digital twin system as claimed in any one of claims 1 to 6.

Images