An industrial software performance optimization method and system based on causal reasoning

Abstract

The application discloses an industrial software performance optimization method and system based on causal reasoning, and relates to the field of industrial software optimization; the method comprises the following steps: predefining an intervention parameter set and an operation environment parameter set of each to-be-optimized parameter; extracting N causal observation events from historical operation logs of the industrial software; constructing a causal event knowledge base based on the N causal observation events; constructing a to-be-completed causal event; wherein the to-be-completed causal event comprises a to-be-optimized parameter of a current system and a target value thereof; extracting K intervention configurations of a to-be-completed causal event candidate in the causal event knowledge base; and obtaining a recommended intervention configuration of the to-be-optimized parameter based on the K intervention configurations; the application generates a recommended intervention configuration that is adapted to a current operation environment, inherits historical effective events, is adapted to the current operation environment, and significantly improves the performance optimization level of the industrial software on the current system.

Description

Technical Field

[0001] This invention relates to the field of industrial software optimization, specifically to a method and system for optimizing industrial software performance based on causal reasoning. Background Technology

[0002] In the field of industrial software performance optimization, traditional methods mainly rely on expert experience for parameter tuning, grid search, or automatic optimization based on statistical models (such as Bayesian optimization and reinforcement learning). Patent document CN114358433A discloses a production planning management optimization method based on vertical federated learning for industrial software integration, which can improve the lag in current enterprise production planning management. However, the aforementioned existing methods generally ignore differences in operating environments, mixing performance data under different hardware configurations or load conditions, causing recommended configurations to fail in new environments. Furthermore, they may mistake highly random performance results for effective interventions, leading to unstable recommendation results. Moreover, when retrieving historical cases, they fail to isolate adjustable parameters from environmental variables, making it difficult to accurately identify the true causal relationship for achieving target performance under the same environment. Therefore, there is an urgent need for an industrial software performance optimization method that can verify and reuse reliable causal experience. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a method and system for optimizing industrial software performance based on causal reasoning. By introducing causal observation events as basic units, a queryable and filterable causal event knowledge base is constructed to solve the technical problems mentioned in the background.

[0004] To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for optimizing the performance of industrial software based on causal reasoning, the method comprising the following steps: S1. A predefined set of intervention parameters and a set of operating environment parameters for each parameter to be optimized; wherein, the parameters are represented as a binary structure of key-value pairs; S2. Extract N causal observation events from the historical operation log of the industrial software; The causal observation events include: intervention configurations that have a causal impact on the parameter to be optimized, operating environment parameters, and observed values ​​of the parameter to be optimized; S3. Based on N causal observation events, construct a causal event knowledge base; where each causal observation event is associated with a unique event identifier; S4. Construct the causal events to be completed; where the causal events to be completed include the parameters to be optimized in the current system and their target values; S5. Extract K intervention configurations from the causal event knowledge base to complete the causal event candidates; S6. Based on K intervention configurations, obtain recommended intervention configurations for the parameters to be optimized; wherein, the recommended intervention configurations consist of a set of key-value pairs composed of several intervention parameters selected from the set of intervention parameters and their recommended values, and the recommended values ​​are generated based on the operating environment parameters of the current system and the parameters to be optimized and their target values.

[0005] In some embodiments, an intervention parameter set and a runtime environment parameter set for the parameter to be optimized are constructed, including: S1-1. Given any parameter to be optimized; S1-2. Based on the given parameters to be optimized, match several corresponding intervention parameters and operating environment parameters in the configuration space of the industrial software. S1-3. Organize several intervention parameters and operating environment parameters into key-value pair sets to generate the intervention parameter set and operating environment parameter set of the parameters to be optimized.

[0006] In some embodiments, a causal event knowledge base is constructed based on N causal observation events, including: S3-1. Calculate the causal stability of N causal observation events; S3-2. Compare the causal stability of N causal observation events with a set threshold. S3-3. If the causal stability exceeds the set threshold, the corresponding causal observation event is marked as a reliable causal event; otherwise, it is marked as a false causal event. S3-4. Store M reliable causal events in a structured manner to construct a causal event knowledge base; In some embodiments, the causal stability of N causal observation events is calculated, including: S3-1-1. Group causal observation events with the same intervention configuration into the same intervention group; S3-1-2. Within each intervention group, causal observation events are clustered based on Euclidean distance of operating environment parameters to form multiple local environment clusters; S3-1-3. For each local environment cluster, calculate the standard deviation of the observed values ​​of the parameter to be optimized within the cluster; S3-1-4. The reciprocal of the standard deviation is used as the causal stability of the local environmental cluster, and this causal stability is assigned to each causal observation event covered within the cluster, until the causal stability of N causal observation events is obtained.

[0007] In some of these embodiments, the causal events to be completed are constructed, including: S4-1. Obtain the current system's operating environment parameters, parameters to be optimized, and their target values; S4-2. Combine the operating environment parameters, the parameters to be optimized, and their target values ​​to obtain the causal events to be completed; In some embodiments, K intervention configurations for candidate causal events to be completed are extracted from the causal event knowledge base, including: S5-1. Input the causal events to be completed into the causal event knowledge base; S5-2. In the causal event knowledge base, extract the operating environment parameter values ​​and the observed values ​​of the parameters to be optimized for M reliable causal events, and combine them to obtain M masked causal events. S5-3. Calculate the similarity of M events between the M masked causal events and the causal events to be completed; S5-4. Based on the similarity of M events, sort the M masked causal events in descending order to generate a sequence of similar events; S5-5. Detect inflection points in similar event sequences, anchor the causal events of event similarity abrupt changes, and record them as event abrupt change inflection points; S5-6. Starting from the masked causal event with the highest event similarity, select masked causal events in sequence until the event before the inflection point of the event mutation is encountered, and obtain K masked causal events; S5-7. Based on the operating environment parameter values ​​of the K masked causal events and the observed values ​​of the parameters to be optimized, locate the corresponding K reliable causal events. S5-8. Extract the corresponding K intervention configurations from the K reliable causal events.

[0008] In some embodiments, inflection point detection is performed on sequences of similar events, including: S5-5-1. In a sequence of similar events, select the event similarity of adjacent mask causal events one by one in descending order. S5-5-2. Calculate the similarity difference between adjacent events based on the event similarity corresponding to adjacent causal events in the mask. S5-5-3. If the similarity difference between adjacent events is greater than a set threshold, then the masked causal event with the smaller similarity among these adjacent events is selected as the event mutation inflection point.

[0009] In some of these embodiments, a recommended intervention configuration for the parameter to be optimized is obtained based on K intervention configurations, including: S6-1. Based on the parameters to be optimized, anchor a predefined set of intervention parameters; S6-2. Select a current intervention parameter from the predefined set of intervention parameters; S6-3. Calculate the frequency of the current intervention parameter's value in the K intervention configurations; S6-4. Select the value that appears most frequently as the recommended value for the current intervention parameter; S6-5. Traverse the predefined set of intervention parameters until the recommended value for each intervention parameter is obtained; S6-6. Combine the recommended values ​​of each intervention parameter to obtain the recommended intervention configuration for the parameter to be optimized.

[0010] This invention provides a method and system for optimizing industrial software performance based on causal reasoning, which has the following beneficial effects: This invention achieves a shift in industrial software performance optimization from experience-based trial and error to causal reuse through structured causal reasoning. It calculates causal stability from causal observation events in historical operation logs and filters out reliable events whose performance can be reproduced under similar operating environments, eliminating false association events and ensuring that the knowledge base contains only real observation events with genuine causal effects. Furthermore, it employs a masking mechanism to hide intervention configurations, thereby calculating event similarity based on operating environment parameters, optimization parameters, and their target values. This focuses the query matching process on the core causal judgment of whether similar performance is achieved in the same environment, avoiding interference from intervention parameters in similarity assessment. Further, by detecting inflection points in similar event sequences, it automatically identifies locations where event similarity significantly decreases, ensuring that recommendations rely only on historically successful events truly similar to the current system. Finally, it generates recommended intervention configurations adapted to the current operating environment, inheriting effective historical events while adapting to the current operating environment, significantly improving the performance optimization level of industrial software for the current system's skills.

[0011] Secondly, the present invention provides an industrial software performance optimization system based on causal reasoning, which executes the industrial software performance optimization method based on causal reasoning described in the first aspect, the system comprising: A parameter set definition unit is used to predefine the intervention parameter set and the operating environment parameter set for each parameter to be optimized; wherein, the parameter is represented as a binary structure of key-value pairs; The observation event extraction unit is used to extract N causal observation events from the historical operation log of the industrial software; The event database construction unit is used to build a causal event knowledge base based on N causal observation events; each causal observation event is associated with a unique event identifier. The event unit to be completed is used to construct the causal events to be completed; wherein, the causal events to be completed include the parameters to be optimized in the current system and their target values; An intervention configuration extraction unit is used to extract K intervention configurations from the causal event knowledge base to complete the causal event candidate; The recommended parameter configuration unit is used to obtain recommended intervention configurations for the parameters to be optimized based on K intervention configurations; The recommended intervention configuration consists of a set of key-value pairs composed of several intervention parameters selected from the intervention parameter set and their recommended values. The recommended values ​​are generated based on the current system's operating environment parameters, the parameters to be optimized, and their target values.

[0012] Compared with the prior art, the beneficial effects of the industrial software performance optimization system based on causal reasoning of the present invention are the same as those of the above-mentioned industrial software performance optimization method based on causal reasoning, so they will not be repeated here. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating an industrial software performance optimization method based on causal reasoning according to the present invention. Figure 2 This is a schematic diagram illustrating the construction process of the causal event knowledge base described in this invention; Figure 3 This is a schematic diagram of the calculation process for the causal stability described in this invention; Figure 4 This is a schematic diagram of the process for selecting the event mutation inflection point described in this invention; Figure 5 This is a structural block diagram of an industrial software performance optimization system based on causal reasoning according to the present invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] Example 1: Please refer to Figures 1 to 4 This invention provides a method for optimizing industrial software performance based on causal reasoning, comprising the following steps: S1. A predefined set of intervention parameters and a set of operating environment parameters for each parameter to be optimized; wherein, the parameters are represented as a binary structure of key-value pairs; Among them, a key-value pair is a binary structure consisting of a parameter name (key) and a specific value (value), in the form of key:value.

[0016] For example, {"thread_count":16,"cache_size":"2GB"} represents an intervention configuration; where thread_count represents the name (key) of the intervention parameter, representing the number of threads, and its value is 16; cache_size represents the cache, and its value is 2GB.

[0017] Specifically, the parameter to be optimized represents the performance indicators (such as throughput and response time) that need to be controlled in industrial software; and when given a specific parameter to be optimized, two sets of parameters need to be defined in advance, including: The intervention parameter set consists of actively adjustable configuration items, such as the number of threads, cache size, and batch size; the runtime environment parameter set consists of uncontrollable environmental characteristics that affect performance, such as the number of CPU cores, total memory, number of concurrent requests, and data size. These two sets of parameters together constitute the basic input for causal reasoning in industrial software.

[0018] S2. Extract N causal observation events from the historical operation log of the industrial software; The causal observation events include: intervention configurations that have a causal impact on the parameter to be optimized, operating environment parameters, and observed values ​​of the parameter to be optimized; specifically, the intervention configurations and operating environment parameters are both represented as sets of key-value pairs. S3. Based on N causal observation events, construct a causal event knowledge base; where each causal observation event is associated with a unique event identifier; Specifically, the causal event knowledge base stores all causal observation events extracted based on historical operation logs, and supports queries by intervention configuration, operating environment parameters, and parameters to be optimized. This knowledge base can be implemented using relational tables or vector storage.

[0019] S4. Construct the causal events to be completed; where the causal events to be completed include the parameters to be optimized in the current system and their target values; S5. Extract K intervention configurations from the causal event knowledge base to complete the causal event candidates; S6. Based on K intervention configurations, obtain recommended intervention configurations for the parameters to be optimized; wherein, the recommended intervention configurations consist of a set of key-value pairs composed of several intervention parameters selected from the set of intervention parameters and their recommended values, and the recommended values ​​are generated based on the operating environment parameters of the current system and the parameters to be optimized and their target values.

[0020] In this embodiment, by organizing historical operation records into causal observation events that include intervention, environmental and performance observations, and constructing a queryable causal event knowledge base, the recommended intervention configuration generated for the current system operating environment and performance goals can correspond to valid events with the same causal context in the past. This enables the recommendation of industrial intervention parameters based on real operational causal experience, and realizes the performance optimization of industrial software from experience-based trial and error to data-driven causal optimization.

[0021] Specifically, in this embodiment, step S1 includes: S1-1. Given any parameter to be optimized; S1-2. Based on the given parameters to be optimized, match several corresponding intervention parameters and operating environment parameters in the configuration space of the industrial software. Each intervention parameter and operating environment parameter consists of its parameter name (as a key) and its corresponding value (as a value), forming a key-value pair structure. S1-3. Organize several intervention parameters and operating environment parameters into key-value pair sets to generate the intervention parameter set and operating environment parameter set of the parameters to be optimized.

[0022] In this embodiment, by defining an intervention parameter set and an operating environment parameter set for each parameter to be optimized, it is clear which configuration items are adjustable and which environmental features need to be included in the comparison. Both sets of parameters are organized in key-value pairs, so that causal observation events extracted from historical logs can be stored in the same structure, thereby accurately matching similar operating environments in history.

[0023] Specifically, in this embodiment, step S3 includes: S3-1. Calculate the causal stability of N causal observation events; S3-2. Compare the causal stability of N causal observation events with a set threshold. S3-3. If the causal stability exceeds the set threshold, the corresponding causal observation event is marked as a reliable causal event; otherwise, it is marked as a false causal event. S3-4. Store M reliable causal events in a structured manner to construct a causal event knowledge base; Specifically, structured storage means that each reliable causal event is organized according to a unified field, including: runtime parameter key-value pairs, intervention parameter key-value pairs, parameters to be optimized and their observations, and associated with a unique event identifier; In this embodiment, the structured storage is implemented using relational database tables, NoSQL documents (such as JSON records), or vector databases.

[0024] In this embodiment, by calculating the causal stability of causal observed events and filtering them according to a threshold, only events that meet the stability standard are stored in the knowledge base as reliable causal events. This excludes historical records with large performance fluctuations or that are not reproducible, so that matching can be based on real and effective causal relationships and avoids recommendations being interfered with by accidental or noisy data.

[0025] Furthermore, step S3-1 also includes: S3-1-1. Group causal observation events with the same intervention configuration into the same intervention group; S3-1-2. Within each intervention group, causal observation events are clustered based on Euclidean distance of operating environment parameters to form multiple local environment clusters; S3-1-3. For each local environment cluster, calculate the standard deviation of the observed values ​​of the parameter to be optimized within the cluster; S3-1-4. The reciprocal of the standard deviation is used as the causal stability of the local environmental cluster, and this causal stability is assigned to each causal observation event covered within the cluster, until the causal stability of N causal observation events is obtained.

[0026] In this embodiment, causal stability is used to represent the impact of a specific intervention configuration on the parameter to be optimized under similar operating environment parameters. If an intervention configuration produces similar observed values ​​of the parameter to be optimized in multiple similar operating environments (i.e., small standard deviation), it indicates that there is a stable causal relationship between the intervention parameter and the parameter to be optimized, rather than being accidental or affected by unobserved confounding factors. Conversely, if the observed values ​​have high dispersion (large standard deviation), it indicates that the effect of the intervention parameter is unreliable and may be affected by noise or other hidden variables. Therefore, using the reciprocal of the standard deviation as causal stability can effectively screen out historical operating events with stable causal effects.

[0027] Specifically, in this embodiment, step S4 includes: S4-1. Obtain the current system's operating environment parameters, parameters to be optimized, and their target values; The parameters to be optimized in the current system are specified by the user (e.g., "desired throughput ≥ 8000 TPS").

[0028] S4-2. Combine the operating environment parameters, the parameters to be optimized, and their target values ​​to obtain the causal events to be completed; Specifically, the causal event to be completed is an incomplete event template with empty intervention configuration fields (to be recommended), containing only known operating environment parameters and parameters to be optimized.

[0029] It should be noted that the causal events to be completed are preferably represented using a triplet structure, i.e., (running environment parameter vector, parameter value to be optimized, empty intervention configuration). Before combination, the running environment parameters and the parameters to be optimized need to be converted into a vectorizable structure using a standardized coding method (such as normalization or one-hot coding).

[0030] Specifically, in this embodiment, step S5 includes: S5-1. Input the causal events to be completed into the causal event knowledge base; S5-2. In the causal event knowledge base, extract the operating environment parameter values ​​and the observed values ​​of the parameters to be optimized for M reliable causal events, and combine them to obtain M masked causal events. Among them, masked causal events indicate that the historical runtime events of the intervention configuration are hidden, and only the runtime environment parameters and the parameters to be optimized are retained.

[0031] S5-3. Calculate the similarity of M events between the M masked causal events and the causal events to be completed; The weighted Euclidean distance is preferred for event similarity, and its calculation formula is as follows: ; in: Represents the weighted Euclidean distance between the causal event to be completed and the causal event of the i-th mask; the smaller the value, the more similar the two are. Representation: The feature vector of the causal event to be completed is composed of the current system's operating environment parameters and the target values ​​of the parameters to be optimized, after encoding; Representation: The feature vector of the i-th masked causal event (i=1,2,…,M) contains only the observation values ​​of the operating environment parameters and the parameters to be optimized in the historical running events, and the intervention configuration has been hidden.

[0032] Represents the total dimension of the feature vector, that is, the total dimension of all parameters involved in the comparison (such as the number of CPU cores, the number of concurrent connections, the throughput, etc.) after encoding.

[0033] Represents: The value (normalized) of the j-th dimension feature of the causal event to be completed. Represents: the value of the i-th masked causal event in the j-th dimension feature (already normalized). This indicates the preset weight of the j-th dimension feature, reflecting the importance of this parameter to performance (e.g., the weight of the number of CPU cores can be higher than the total amount of memory). Represents the scaling factor of the j-th feature, typically the standard deviation or range of values ​​for that feature in historical data, used to eliminate dimensional differences. Representation: The normalized difference of the j-th feature makes parameters of different orders of magnitude comparable. In other words, weighted Euclidean distance calculates the distance between two events in the environmental and target dimensions, under the premise of controlling the dimensions and parameters. The smaller the distance, the higher the similarity between the events.

[0034] Specifically, the masking operation ensures that the similarity is determined solely by the runtime environment parameters and the target parameters to be optimized, avoiding interference from intervention parameters.

[0035] S5-4. Based on the similarity of M events, sort the M masked causal events in descending order to generate a sequence of similar events; After sorting, events at the beginning of the sequence represent those whose historical operating environment is closest to the current scenario and have reached similar performance levels, thus having higher matching value.

[0036] S5-5. Detect inflection points in similar event sequences, anchor the causal events of event similarity abrupt changes, and record them as event abrupt change inflection points; The inflection point of an event mutation marks the end of highly similar events, after which the relevance of subsequent events to the current scenario decreases sharply.

[0037] S5-6. Starting from the masked causal event with the highest event similarity, select masked causal events in sequence until the event before the inflection point of the event mutation is encountered, and obtain K masked causal events; The selected K masked causal events represent historical cases in which the target performance of the parameters to be optimized was successfully achieved under similar historical conditions, and are used to generate recommended configurations.

[0038] S5-7. Based on the operating environment parameter values ​​of the K masked causal events and the observed values ​​of the parameters to be optimized, locate the corresponding K reliable causal events. Specifically, the complete original event (including intervention configuration) can be retrieved from the causal event knowledge base using the event identifier.

[0039] S5-8. Extract the corresponding K intervention configurations from the K reliable causal events.

[0040] Each intervention configuration is a set of key-value pairs, representing a successful parameter setting.

[0041] In this embodiment, by performing a masked comparison between the causal events to be completed and the reliable causal events in the knowledge base, the similarity is calculated based solely on the runtime environment parameters and the parameters to be optimized, thus avoiding interference from the intervention configuration on the matching. After sorting by similarity, historical runtime events with high similarity segments are extracted, and their complete records are retrieved to extract the corresponding intervention configurations, so that the obtained K intervention configurations all come from historical successful cases that are similar to the current system environment and have performance close to the target.

[0042] Furthermore, step S5-5 also includes: S5-5-1. In a sequence of similar events, select the event similarity of adjacent mask causal events one by one in descending order. S5-5-2. Calculate the similarity difference between adjacent events based on the event similarity corresponding to adjacent causal events in the mask. S5-5-3. If the similarity difference between adjacent events is greater than a set threshold, then the masked causal event with the smaller similarity among these adjacent events is selected as the event mutation inflection point.

[0043] In this embodiment, by calculating the similarity difference between adjacent events in a similar event sequence and locating the abrupt inflection point when the decrease exceeds a set threshold, the selection boundary for highly similar historical events can be automatically determined; events after the selected inflection point are excluded because their similarity to the current scene drops sharply, ensuring that the intervention configurations used for recommendation come only from truly similar operating environments.

[0044] Specifically, in this embodiment, step S6 includes: S6-1. Based on the parameters to be optimized, anchor a predefined set of intervention parameters; S6-2. Select a current intervention parameter from the predefined set of intervention parameters; S6-3. Calculate the frequency of the current intervention parameter's value in the K intervention configurations; S6-4. Select the value that appears most frequently as the recommended value for the current intervention parameter; If multiple values ​​have the same frequency, the intervention parameters corresponding to highly similar events can be given higher weights.

[0045] S6-5. Traverse the predefined set of intervention parameters until the recommended value for each intervention parameter is obtained; S6-6. Combine the recommended values ​​of each intervention parameter to obtain the recommended intervention configuration for the parameter to be optimized; Specifically, the recommended key-value pair collection can be output as a configuration suggestion that can be directly applied to the system, such as in JSON / YAML format: {"thread_count":16,"cache_size":"2GB","batch_size":64}.

[0046] In this embodiment, by counting the frequency of each intervention parameter value in the K intervention configurations and selecting the most frequently occurring value as the recommended value, the common configurations in historical successful cases are extracted into a unified recommended intervention configuration. When the frequencies are the same, the values ​​from highly similar events are adopted first, so that the final output key-value pair set can integrate the selection of most effective events and approach the current operating conditions.

[0047] Example 2: This Example 2 differs from Example 1 in that it also provides an industrial software performance optimization system based on causal reasoning. This system is used to implement the above-described method embodiments, and details already described will not be repeated. The terms "module," "unit," and "subunit" used below refer to combinations of software and / or hardware that perform predetermined functions. Although the system described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0048] Figure 5This is a structural block diagram of an industrial software performance optimization system based on causal reasoning according to the present invention. The system includes: A parameter set definition unit is used to predefine the intervention parameter set and the operating environment parameter set for each parameter to be optimized; wherein, the parameter is represented as a binary structure of key-value pairs; The observation event extraction unit is used to extract N causal observation events from the historical operation log of the industrial software; The event database construction unit is used to build a causal event knowledge base based on N causal observation events; each causal observation event is associated with a unique event identifier. The event unit to be completed is used to construct the causal events to be completed; wherein, the causal events to be completed include the parameters to be optimized in the current system and their target values; An intervention configuration extraction unit is used to extract K intervention configurations from the causal event knowledge base to complete the causal event candidate; The recommended parameter configuration unit is used to obtain recommended intervention configurations for the parameters to be optimized based on K intervention configurations; The recommended intervention configuration consists of a set of key-value pairs composed of several intervention parameters selected from the intervention parameter set and their recommended values. The recommended values ​​are generated based on the current system's operating environment parameters, the parameters to be optimized, and their target values.

[0049] In the above system, The system employs a parameter set definition unit to define intervention parameter sets and operating environment parameter sets. An observation event extraction unit extracts N causal observation events. An event library construction unit builds a causal event knowledge base. An event completion unit constructs causal events to be completed. An intervention configuration extraction unit extracts K intervention configurations. A recommended parameter configuration unit obtains recommended intervention configurations for the parameters to be optimized. The recommended intervention configurations consist of a set of key-value pairs composed of several intervention parameters selected from the intervention parameter set and their recommended values. The recommended values ​​are generated based on the current system's operating environment parameters and the parameters to be optimized and their target values, thus solving the problem of accurately identifying true causal relationships.

[0050] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means.

[0051] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for optimizing industrial software performance based on causal reasoning, characterized in that, include: S1. A predefined set of intervention parameters and a set of operating environment parameters for each parameter to be optimized; wherein, the parameters are represented as a binary structure of key-value pairs; S2. Extract N causal observation events from the historical operation log of the industrial software; The causal observation events include: intervention configurations that have a causal impact on the parameter to be optimized, operating environment parameters, and observed values ​​of the parameter to be optimized; S3. Based on N causal observation events, construct a causal event knowledge base; where each causal observation event is associated with a unique event identifier; S4. Construct the causal events to be completed; where the causal events to be completed include the parameters to be optimized in the current system and their target values; S5. Extract K intervention configurations from the causal event knowledge base to complete the causal event candidates; S6. Based on K intervention configurations, obtain recommended intervention configurations for the parameters to be optimized; The recommended intervention configuration consists of a set of key-value pairs composed of several intervention parameters selected from the intervention parameter set and their recommended values. The recommended values ​​are generated based on the current system's operating environment parameters, the parameters to be optimized, and their target values.

2. The industrial software performance optimization method based on causal reasoning according to claim 1, characterized in that, Construct the intervention parameter set and the runtime environment parameter set for the parameters to be optimized, including: S1-1. Given any parameter to be optimized; S1-2. Based on the given parameters to be optimized, match several corresponding intervention parameters and operating environment parameters in the configuration space of the industrial software. S1-3. Organize several intervention parameters and operating environment parameters into key-value pair sets to generate the intervention parameter set and operating environment parameter set of the parameters to be optimized.

3. The industrial software performance optimization method based on causal reasoning according to claim 1, characterized in that, Based on N causal observation events, a causal event knowledge base is constructed, including: S3-1. Calculate the causal stability of N causal observation events; S3-2. Compare the causal stability of N causal observation events with a set threshold. S3-3. If the causal stability exceeds the set threshold, the corresponding causal observation event is marked as a reliable causal event; otherwise, it is marked as a false causal event. S3-4. Store M reliable causal events in a structured manner to build a causal event knowledge base.

4. The industrial software performance optimization method based on causal reasoning according to claim 3, characterized in that, Calculate the causal stability of N causal observation events, including: S3-1-1. Group causal observation events with the same intervention configuration into the same intervention group; S3-1-2. Within each intervention group, causal observation events are clustered based on Euclidean distance of operating environment parameters to form multiple local environment clusters; S3-1-3. For each local environment cluster, calculate the standard deviation of the observed values ​​of the parameter to be optimized within the cluster; S3-1-4. The reciprocal of the standard deviation is used as the causal stability of the local environmental cluster, and this causal stability is assigned to each causal observation event covered within the cluster, until the causal stability of N causal observation events is obtained.

5. The industrial software performance optimization method based on causal reasoning according to claim 1, characterized in that, Construct the causal events to be completed, including: S4-1. Obtain the current system's operating environment parameters, parameters to be optimized, and their target values; S4-2. Combine the operating environment parameters, the parameters to be optimized, and their target values ​​to obtain the causal events to be completed.

6. The industrial software performance optimization method based on causal reasoning according to claim 1, characterized in that, From the causal event knowledge base, extract K intervention configurations for the candidate causal events to be completed, including: S5-1. Input the causal events to be completed into the causal event knowledge base; S5-2. In the causal event knowledge base, extract the operating environment parameter values ​​and the observed values ​​of the parameters to be optimized for M reliable causal events, and combine them to obtain M masked causal events. S5-3. Calculate the similarity of M events between the M masked causal events and the causal events to be completed; S5-4. Based on the similarity of M events, sort the M masked causal events in descending order to generate a sequence of similar events; S5-5. Detect inflection points in similar event sequences, anchor the causal events of event similarity abrupt changes, and record them as event abrupt change inflection points; S5-6. Starting from the masked causal event with the highest event similarity, select masked causal events in sequence until the event before the inflection point of the event mutation is encountered, and obtain K masked causal events; S5-7. Based on the operating environment parameter values ​​of the K masked causal events and the observed values ​​of the parameters to be optimized, locate the corresponding K reliable causal events. S5-8. Extract the corresponding K intervention configurations from the K reliable causal events.

7. The industrial software performance optimization method based on causal reasoning according to claim 5, characterized in that, Inflection point detection for similar event sequences includes: S5-5-1. In a sequence of similar events, select the event similarity of adjacent mask causal events one by one in descending order. S5-5-2. Calculate the similarity difference between adjacent events based on the event similarity corresponding to adjacent causal events in the mask. S5-5-3. If the similarity difference between adjacent events is greater than a set threshold, then the masked causal event with the smaller similarity among these adjacent events is selected as the event mutation inflection point.

8. The industrial software performance optimization method based on causal reasoning according to claim 7, characterized in that, Based on K intervention configurations, recommended intervention configurations for the parameters to be optimized are obtained, including: S6-1. Based on the parameters to be optimized, anchor a predefined set of intervention parameters; S6-2. Select a current intervention parameter from the predefined set of intervention parameters; S6-3. Calculate the frequency of the current intervention parameter's value in the K intervention configurations; S6-4. Select the value that appears most frequently as the recommended value for the current intervention parameter; S6-5. Traverse the predefined set of intervention parameters until the recommended value for each intervention parameter is obtained; S6-6. Combine the recommended values ​​of each intervention parameter to obtain the recommended intervention configuration for the parameter to be optimized.

9. An industrial software performance optimization system based on causal reasoning, executing the industrial software performance optimization method based on causal reasoning as described in any one of claims 1 to 8, characterized in that, The system includes: A parameter set definition unit is used to predefine the intervention parameter set and the operating environment parameter set for each parameter to be optimized; wherein, the parameter is represented as a binary structure of key-value pairs; The observation event extraction unit is used to extract N causal observation events from the historical operation log of the industrial software; The event database construction unit is used to build a causal event knowledge base based on N causal observation events; each causal observation event is associated with a unique event identifier. The event unit to be completed is used to construct the causal events to be completed; wherein, the causal events to be completed include the parameters to be optimized in the current system and their target values; An intervention configuration extraction unit is used to extract K intervention configurations from the causal event knowledge base to complete the causal event candidate; The recommended parameter configuration unit is used to obtain recommended intervention configurations for the parameters to be optimized based on K intervention configurations; The recommended intervention configuration consists of a set of key-value pairs composed of several intervention parameters selected from the intervention parameter set and their recommended values. The recommended values ​​are generated based on the current system's operating environment parameters, the parameters to be optimized, and their target values.

Images