A cloud configuration item intelligent optimization system and method

Abstract

The application provides a cloud configuration item intelligent optimization system and method, and relates to the technical field of cloud configuration data optimization. The system comprises the following steps: mapping configuration item data to different levels of a rule dependency model to construct a dynamic configuration relationship graph, generating a multi-modal evidence body and a dynamic trust chain rule based on the dynamic configuration relationship graph and key attributes of the configuration item, and constructing a cause-effect dependency graph through instantiation, calculating a robustness index of the cause-effect dependency graph, performing conflict resolution arbitration, and determining the state of the configuration item according to the arbitration result; improving data accuracy and self-healing ability through automatic conflict resolution, enhancing the real-time performance and consistency of configuration optimization through cause-effect dependency analysis, and improving optimization efficiency while ensuring reliability.

Description

Technical Field

[0001] This invention relates to a cloud configuration item intelligent optimization system and method, belonging to the field of cloud configuration data optimization technology. Background Technology

[0002] As enterprise IT architectures become increasingly complex, with a vast number of configuration items and intricate relationships, configuration management databases, as the core of IT service management, have expanded their management objects from traditional servers and network devices to complex cloud configuration items such as cloud servers, cloud databases, Kubernetes clusters, containers, and various microservices. However, cloud services cannot utilize the relationships between configuration items for impact analysis in real time and intelligently when dealing with changes, events, and issues related to configuration items, thus affecting the reliability and stability of cloud services.

[0003] In existing technologies, cloud configuration item optimization mainly maintains configuration items and their relationships through a configuration management database, and compares configuration items based on static rule verification to determine differences. However, existing technologies have the following problems: the dimension of configuration item connection relationship is single, which cannot express the business perspective and reflect dynamic service dependency changes. This results in insufficient ability to understand complex business contexts and perceive configuration item risk conflicts, affecting the management efficiency and reliability of cloud configuration items. Summary of the Invention

[0004] The purpose of this invention is to provide a cloud configuration item intelligent optimization system and method to solve the problems of insufficient execution efficiency and reliability in the prior art.

[0005] To solve the above-mentioned technical problems, the present invention is implemented using the following technical solution.

[0006] A cloud configuration item intelligent optimization method includes:

[0007] Specifically, it collects heterogeneous configuration item data in the cloud environment, maps the configuration item data to different levels in a preset rule dependency model, and constructs a dynamic configuration relationship graph.

[0008] Based on the dynamic configuration relationship graph, a multimodal evidence body is constructed for the key attributes of the configuration item, and dynamic trust chain rules based on the rule dependency model are set.

[0009] Based on the dynamic configuration relationship graph and multimodal evidence body, the evidence source node and fact node are instantiated and a causal dependency graph is constructed. The causal dependency graph is backtracked and analyzed through dynamic trust chain rules, and its robustness index is calculated.

[0010] When a data conflict is detected in the causal dependency graph, the minimum evidence source node cut set is solved by the robustness index, the conflict subgraph is arbitrated to obtain the configuration item status after arbitration, and the configuration item is intelligently optimized. The conflict subgraph is obtained by partitioning the causal dependency graph.

[0011] Specifically, based on the dynamic configuration relationship graph, multimodal evidence bodies are constructed for the key attributes of configuration items, and dynamic trust chain rules based on a rule dependency model are set, including:

[0012] For the key attributes of configuration item nodes in the dynamic configuration relationship graph, we extract their directly reported values ​​as direct evidence, extract relationship evidence from adjacent nodes by traversing the relationship edges of configuration item nodes, and extract activity evidence through the event stream related to configuration item nodes.

[0013] Direct evidence, relational evidence, and activity evidence are bound to key attributes in the form of attribute evidence pairs. A multimodal evidence body of the key attributes of the corresponding configuration item node is constructed by combining the evidence quality vector. The evidence quality vector is calculated based on direct evidence, relational evidence, and activity evidence.

[0014] Based on the hierarchical derivation of the dependency relationships in the rule dependency model, the rules for trust transfer and conflict resolution are derived and formalized into a production rule set as the initial dynamic trust chain rules.

[0015] Calculate the historical validity and context relevance coefficients for the initial dynamic trust chain rules, determine the decision weights and priority parameters based on the historical validity and context relevance coefficients, and generate dynamic trust chain rules.

[0016] Calculate historical validity and context relevance coefficients for the initial dynamic trust chain rules, determine decision weights and priority parameters based on the historical validity and context relevance coefficients, and generate dynamic trust chain rules, including:

[0017] Based on the execution status of the initial dynamic trust chain rules, an effectiveness vector is constructed by the number of successful arbitrations and the total number of triggers. The determination of the number of successful arbitrations is based on the fact that the result of this arbitration remains stable within the next time window.

[0018] Set the context features of the configuration items according to the configuration item type and data source type, calculate the distribution similarity of the context features, and normalize the distribution similarity to obtain the context relevance coefficient.

[0019] The distribution characteristics of historical validity and context relevance coefficients are statistically analyzed to determine the critical interval of decision weights. Based on the critical interval, a piecewise nonlinear transformation function is constructed, and the decision weights are calculated in combination with a preset adjustment factor.

[0020] The initial dynamic trust chain rules are prioritized based on the decision weights to obtain the dynamic trust chain rules.

[0021] Specifically, based on the dynamically configured relationship graph and multimodal evidence body, evidence source nodes and fact nodes are instantiated and a causal dependency graph is constructed. The causal dependency graph is then backtracked and analyzed using dynamic trust chain rules to calculate its robustness index, including:

[0022] Based on the dynamic configuration relationship graph and the multimodal evidence body, the key attributes of the configuration items are instantiated as fact nodes, and the evidence data in the multimodal evidence body are instantiated as evidence source nodes. Directed edges are established between the evidence source nodes and fact nodes to construct a causal dependency graph.

[0023] According to the dynamic trust chain rules, the evidence source nodes and fact nodes connected to the fact nodes in the causal dependency graph are traced back to obtain the cross-level dependency path and generate the evidence chain of the fact nodes.

[0024] Extract the structural feature parameters of the evidence chain corresponding to the fact node, including the depth feature parameter determined by the number of configuration levels traversed by the evidence chain, the breadth feature parameter determined by the number of independent evidence source nodes, and the diversity feature parameter determined by the heterogeneity of evidence types.

[0025] The robustness index is calculated using the depth, breadth, and diversity features of the evidence chain.

[0026] Specifically, based on the dynamic trust chain rules, the evidence source nodes and fact nodes connected to the fact nodes in the causal dependency graph are traced back to obtain cross-level dependency paths, and evidence chains of fact nodes are generated, including:

[0027] Starting from the target fact node in the causal dependency graph, trace back the directly connected source evidence nodes in the outward direction and integrate them into a set of direct evidence in the chain of evidence.

[0028] According to the dynamic trust chain rules, the target fact node is taken as the starting point, and the parent fact node it depends on is traced back in the inbound direction. The evidence source nodes connected by the parent fact node are recursively accessed and integrated into a set of relational evidence.

[0029] Based on the causal dependency graph, retrieve the state change event records related to the configuration item attributes represented by the target fact node, and integrate the corresponding evidence source nodes into an active evidence set;

[0030] The direct evidence set, relational evidence set, and activity evidence set are merged and deduplicated. They are then processed according to their respective rule dependency model levels and evidence types to generate an evidence chain containing multimodal evidence and cross-level dependency relationships.

[0031] Specifically, when a data conflict is detected in the causal dependency graph, the minimum evidence source node cut set is solved using the robustness index. Conflict resolution arbitration is then performed on the conflict subgraph to obtain the arbitration status of the configuration items. Intelligent optimization of the configuration items is then performed. The conflict subgraph is obtained by partitioning the causal dependency graph and includes:

[0032] Based on the causal dependency graph, the states of fact nodes pointing to the same key attribute of the configuration item are compared, and the fact nodes with inconsistent states and all their associated evidence source nodes are divided to obtain the conflict subgraph.

[0033] The conflict subgraph is modeled as a flow network. Through flow network analysis, the maximum flow from the evidence source node to the conflict fact node is calculated, the minimum capacity cut set is determined, and the minimum evidence source node cut set is obtained.

[0034] The conflict subgraph is refurbished based on the minimum evidence source node cut set, and the states of the fact nodes are compared. The state of the fact node with the highest robustness index is used as the arbitrated state of the configuration item's key attributes.

[0035] The dynamic configuration relationship graph is updated based on the status of the configuration items after arbitration. The conflict subgraph, the minimum evidence source node cut set, the robustness index, and the status of the configuration items after arbitration are recorded and fused into the multimodal evidence body as feedback data.

[0036] By analyzing multimodal evidence, the dynamic trust chain rules are periodically adjusted, and correction and reset operations are performed on the configuration items based on the status of the configuration items after arbitration and the optimized dynamic trust chain rules.

[0037] Specifically, the conflict subgraph is modeled as a flow network. Through flow network analysis, the maximum flow from the evidence source node to the conflicting fact node is calculated, the minimum capacity cut set is determined, and the minimum evidence source node cut set is obtained, including:

[0038] Set the evidence source node of the conflict subgraph as the source node of the flow, set the conflict fact node with state conflict as the sink node of the flow, and set directed edges according to the source node and sink node of the flow to establish the flow network.

[0039] Based on the dynamic trust chain rules, the edges of nodes in the flow network are assigned capacity weights, where the size of the capacity weight is determined according to the level size, level association and the edge corresponding to the active evidence.

[0040] Based on the source and sink of the flow and the capacity weights assigned to each edge, the maximum flow of the flow network is obtained by iteratively searching for feasible paths from the source to the sink and pushing traffic along the feasible paths until there are no augmenting paths that satisfy the capacity weight constraints.

[0041] The maximum flow identification method based on flow networks identifies all saturated edges whose flow is equal to their capacity weight, and constructs a minimum evidence source node cut set by connecting the evidence source nodes directly to the saturated edges.

[0042] Specifically, conflict resolution is performed on the conflict subgraph based on the minimum evidence source node cut set, and the states of the fact nodes within it are compared. The state of the fact node with the highest robustness index is used as the arbitrated state of the configuration item's key attributes, including:

[0043] Based on the minimum evidence source node segmentation set, the conflict state of evidence source nodes pointing to the same fact node in the conflict subgraph is detected, and the evidence source nodes and their associated edges that are judged to be untrustworthy in the conflict subgraph are eliminated according to the rules of multimodal evidence body and dynamic trust chain, and the corrected conflict subgraph is constructed.

[0044] Based on the state of the fact nodes supported by the evidence source nodes, the consistency degree of the logical relationship of the key attributes of the configuration items is calculated on the corrected conflict subgraph to obtain a global consistency score.

[0045] Based on the current conflict configuration items, preset repair operations are performed. Combined with the global consistency score, simulated repair operations are executed and the state deviation is evaluated to obtain the virtual operation success rate.

[0046] An arbitration utility function is constructed, which takes the global consistency score, the success rate of virtual operations, and the average robustness index of the evidence source node set in the modified conflict subgraph as inputs. The status of the configuration items after arbitration is determined by calculating the arbitration level value.

[0047] Specifically, it collects heterogeneous configuration item data from the cloud environment, maps the configuration item data to different levels in a pre-defined rule dependency model, and constructs a dynamic configuration relationship graph, including:

[0048] The target cloud environment is queried to obtain the raw data streams of computing, network, storage and application services, and structured configuration item data is generated through feature extraction;

[0049] By using a pre-defined rule dependency model, the type and semantic features of configuration items are mapped to the corresponding levels;

[0050] Based on the hierarchical mapping results and the relational attributes contained in the configuration item data, logical edges are established between configuration items at the same or different levels through relation inference, resulting in a dynamic configuration relation graph.

[0051] A cloud configuration item intelligent optimization system includes a configuration perception module, an evidence fusion module, an arbitration decision module, and an execution module;

[0052] The configuration awareness module is used to collect heterogeneous configuration item data in the cloud environment, map the configuration item data to different levels in the preset rule dependency model, and construct a dynamic configuration relationship graph.

[0053] The evidence fusion module is used to construct a multimodal evidence body for the key attributes of the configuration item based on the dynamic configuration relationship graph, and to set dynamic trust chain rules based on the rule dependency model.

[0054] The arbitration decision module is used to instantiate evidence source nodes and fact nodes and construct a causal dependency graph based on the dynamically configured relationship graph and multimodal evidence body, and to perform backtracking analysis on the causal dependency graph through dynamic trust chain rules to calculate its robustness index.

[0055] The execution module is used to solve the minimum evidence source node cut set by robustness index when a data conflict is detected in the causal dependency graph, to perform conflict resolution arbitration on the conflict subgraph, to obtain the configuration item status after arbitration, and to perform intelligent optimization on the configuration item. The conflict subgraph is obtained by partitioning the causal dependency graph.

[0056] Compared with existing technologies, the beneficial effects achieved by this invention are as follows: It maps configuration item data to a preset rule dependency model, generates a dynamic configuration relationship graph, constructs multimodal evidence bodies for key attributes in configuration items, and generates dynamic trust chain rules based on the rule dependency model. By instantiating evidence source nodes and fact nodes to establish a causal dependency graph, it automatically verifies and detects conflicts in configuration item states. Through arbitration using the minimum evidence source node cut set and robustness index, it intelligently resolves configuration item conflicts. This invention solves the problems of insufficient management efficiency and reliability of cloud services in existing technologies, enhances the ability to handle configuration item risk conflicts, and enhances the real-time performance and consistency of configuration management by utilizing causal dependency analysis, thereby improving management efficiency while ensuring reliability.

[0057] It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and are not intended to limit the technical solutions of this disclosure. Attached Figure Description

[0058] Figure 1 A flowchart of a cloud configuration item intelligent optimization method provided by the present invention;

[0059] Figure 2 A flowchart for calculating the robustness index provided by this invention;

[0060] Figure 3 This is a schematic diagram of the flow network structure provided by the present invention;

[0061] Figure 4 This invention provides a structural diagram of a cloud configuration item intelligent optimization system. Detailed Implementation

[0062] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0063] The term "and / or" simply describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0064] Example 1:

[0065] Please see Figures 1-3 This invention provides an embodiment of a cloud configuration item intelligent optimization method, which includes the following specific steps:

[0066] Step S1: Collect heterogeneous configuration item data in the cloud environment, map the configuration item data to different levels in the preset rule dependency model, and construct a dynamic configuration relationship graph.

[0067] The specific steps of step S1 are as follows:

[0068] Step S101: Query the target cloud environment to obtain the raw data stream of computing, network, storage and application services, and generate structured configuration item data through feature extraction.

[0069] In this embodiment, resource status query requests are initiated concurrently to the virtualization platform, container orchestrator, network controller, and cloud management platform based on the target cloud environment to obtain the raw data stream. The data stream is then parsed according to the protocol, and the resource identifier field, resource type field, resource attribute key-value pair set, and resource association description field are separated according to the data source. Through semantic standardization processing, the data is mapped to a unified attribute namespace and converted into a standard metering format to obtain a standardized set of resource attributes, i.e., configuration item data.

[0070] Step S102: Map the type and semantic features of configuration items to the corresponding level through a preset rule dependency model.

[0071] In this embodiment, the semantic features of the type identifier and attribute key-value pairs of the configuration item data are extracted. According to the hierarchical classification rules in the preset rule dependency model, the type of the configuration item is matched with the corresponding level in the rule dependency model to generate a candidate level set and level identifier. By traversing the semantic features of the configuration item, semantic similarity is calculated with the features of each level in the rule dependency model. The semantic similarity and level identifier are mapped to the rule dependency model. It should be noted that the rule dependency model is set by those skilled in the art according to the actual situation.

[0072] Step S103: Based on the hierarchical mapping results and the relational attributes contained in the configuration item data, logical edges are established between configuration items at the same or different levels through relation inference to obtain a dynamic configuration relation graph.

[0073] In this embodiment, based on the hierarchical mapping results and the relational attributes contained in the configuration item data, the connection information, dependency descriptions, and deployment declarations contained in the relational attributes of each configuration item are parsed. By extracting the target configuration item identifier from the relational attributes, initial logical edges are established between configuration item nodes at the same or different levels. Based on the network connection parameters in the configuration item attributes, connection logical edges are established between corresponding configuration item nodes. Based on the deployment path and process information in the configuration item attributes, dependency logical edges are established between corresponding configuration item nodes. Based on the membership description in the configuration item attributes, affiliation logical edges are established between corresponding configuration item nodes. Based on the hierarchical position of the configuration item in the rule dependency model, the directionality of the established logical edges and their conformity with the hierarchical constraints are verified. Edges that violate the hierarchical dependency principle are adjusted in direction or deleted, generating a dynamic configuration relationship graph.

[0074] Step S2: Based on the dynamic configuration relationship graph, construct a multimodal evidence body for the key attributes of the configuration item, and set dynamic trust chain rules based on the rule dependency model.

[0075] The specific steps of step S2 are as follows:

[0076] Step S201: For the key attributes of configuration item nodes in the dynamic configuration relationship graph, extract their directly reported values ​​as direct evidence, extract relationship evidence from adjacent nodes by traversing the relationship edges of configuration item nodes, and extract activity evidence through the relevant event stream of configuration item nodes.

[0077] In this embodiment, the latest directly reported value of the key attribute of the configuration item node is marked as direct evidence. Based on all outgoing and incoming edges of the configuration item node in the dynamic configuration relationship graph, for each adjacent node, it is checked whether there are attribute items in its attributes that have a dependency relationship with the target key attribute. The current state value of the attribute item is extracted as relationship evidence. Event records related to the change of the key attribute state are converted into activity evidence. Evidence collection timestamps are obtained, and the direct evidence, relationship evidence, and activity evidence are timestamp aligned and numerically standardized.

[0078] Step S202: Bind direct evidence, relational evidence, and activity evidence to key attributes in the form of attribute evidence pairs, and construct a multimodal evidence body of the key attributes of the corresponding configuration item node in combination with the evidence quality vector, wherein the evidence quality vector is calculated based on direct evidence, relational evidence, and activity evidence.

[0079] In this embodiment, direct evidence, relational evidence, and activity evidence are bound to their corresponding key attributes to obtain attribute evidence pairs. Evidence freshness is calculated based on the difference between the evidence collection timestamp and the current time. Reliability weights are preset according to the evidence type. A consistency score is obtained by comparing the logical consistency between different pieces of evidence under the same key attribute. Evidence freshness, reliability weights, and consistency scores are weighted and fused to generate an evidence quality vector. The evidence quality vector is then associated and bound to the corresponding attribute evidence pairs to construct a multimodal evidence body containing original evidence data and quality assessment results. It should be noted that the reliability weights are set by those skilled in the art according to the actual situation.

[0080] Step S203: Based on the hierarchical derivation of the dependency relationships in the rule dependency model, derive the trust transfer and conflict resolution rules, and formalize them into a production rule set as the initial dynamic trust chain rules.

[0081] In this embodiment, based on the rule dependency model, pairs of nodes with direct dependencies are identified. The influence of the lower-level evidence status on the upper-level evidence status is analyzed through these pairs. When a lower-level evidence status is detected as verified, a weight enhancement coefficient is set for its corresponding upper-level evidence status. If the confidence level of the lower-level evidence is higher than a preset confidence threshold and directly contradicts the upper-level evidence, a lower-level priority rule is triggered, and the lower-level evidence status covers the upper-level evidence status. If the difference in confidence levels between the upper and lower levels is within a fuzzy range, a secondary adjudication of the reliability and timeliness of the evidence source is performed through evidence tracing. Based on the dependency relationship and conflict decision, production rules are encoded to generate an initial dynamic trust chain rule set. It should be noted that the confidence threshold is set by those skilled in the art according to the actual situation.

[0082] Step S204: Calculate the historical validity and context relevance coefficients for the initial dynamic trust chain rules, determine the decision weights and priority parameters based on the historical validity and context relevance coefficients, and generate dynamic trust chain rules.

[0083] The specific steps of step S204 are as follows:

[0084] Step S2041: Based on the execution status of the initial dynamic trust chain rules, construct an effectiveness vector by the number of successful arbitrations and the total number of triggers. The determination of the number of successful arbitrations is based on the fact that the result of this arbitration remains stable in the next time window.

[0085] In this embodiment, based on the historical execution records of the initial dynamic trust chain rule, the total number of triggers is counted. For the arbitration result generated after each trigger execution, the status changes of configuration items related to the arbitration result and the occurrence of alarm events are monitored through a time window. If a new conflict alarm is detected within the time window, and the conflict alarm is directly related to the configuration item involved in the current arbitration result, the arbitration is determined to be unsuccessful. If the status of the configuration item is forcibly modified within the time window, and the modified status is inconsistent with the arbitration result, the arbitration is determined to be unsuccessful. If neither of the above two situations occurs at the end of the time window, the arbitration is determined to be successful. Based on the determination results, the ratio of the number of successful arbitrations to the total number of triggers is used as the validity measure of the rule, and a vector containing all rule validity measures is constructed.

[0086] Step S2042: Set the context features of the configuration items according to the configuration item type and data source type, calculate the distribution similarity of the context features, and normalize the distribution similarity to obtain the context relevance coefficient.

[0087] In this embodiment, context features are generated by combining the configuration item type identifier and the data source type identifier in the configuration items. The context feature vectors corresponding to the historical triggering scenarios of the initial dynamic trust chain rules are used to construct a historical scenario distribution set. The difference between the context features and the feature vectors in the historical scenario distribution set is calculated to obtain the overall difference measure between the current context and the historical scenario distribution. Based on the overall difference measure, the initial similarity value is calculated through an inverse proportional mapping function, where the larger the difference, the smaller the initial similarity value. The numerical distribution range of the initial similarity value is detected and linearly transformed to a continuous interval between zero and one. The transformed similarity value is used as the context relevance coefficient.

[0088] Step S2043: Statistically analyze the distribution characteristics of historical validity and contextual relevance coefficients, determine the critical interval of decision weights, construct a piecewise nonlinear transformation function based on the critical interval, and calculate the decision weights in combination with preset adjustment factors.

[0089] In this embodiment, the joint probability distribution of historical validity measure and context relevance coefficient is calculated, and the standard deviation is calculated based on the joint probability distribution. The range of the extended standard deviation of the mean of historical validity measure is determined as the critical interval of historical validity, and the range of the extended standard deviation of the mean of context relevance coefficient is determined as the critical interval of context relevance. A piecewise nonlinear transformation function is constructed based on the critical intervals of context relevance and historical validity. The piecewise nonlinear transformation function adopts a quadratic function form to enhance sensitivity within the critical interval and a logarithmic function form to ensure output stability outside the critical interval. The historical validity measure and context relevance coefficient are used as inputs to obtain basic weight values. The basic weight values ​​are multiplied by a preset adjustment factor, and the result of the multiplication is normalized to constrain it within the [0,1] interval to obtain the decision weight. It should be noted that the adjustment factor is set by those skilled in the art according to the actual situation.

[0090] Step S2044: Set the priority of the initial dynamic trust chain rules according to the decision weight to obtain the dynamic trust chain rules.

[0091] In this embodiment, the initial dynamic trust chain rules are arranged in descending order according to their decision weights to generate an initial priority sequence. Conflict resolution is performed on pairs of initial dynamic trust chain rules that have logical conflicts. If the initial dynamic trust chain rules produce opposite conclusions under the same triggering conditions, they are retained or marked as pending based on their decision weights. The priority sequence that has completed conflict resolution is converted into a rule activation queue, where each queue position contains the corresponding dynamic trust chain rule and its current decision weight, thus obtaining the dynamic trust chain rules.

[0092] Step S3: Based on the dynamic configuration relationship graph and multimodal evidence body, instantiate the evidence source node and fact node and construct the causal dependency graph. Perform backtracking analysis on the causal dependency graph through the dynamic trust chain rules and calculate its robustness index.

[0093] The specific steps of step S3 are as follows:

[0094] Step S301: Based on the dynamic configuration relationship graph and the multimodal evidence body, instantiate the key attributes of the configuration items as fact nodes, instantiate the evidence data in the multimodal evidence body as evidence source nodes, establish directed edges between the evidence source nodes and fact nodes, and construct a causal dependency graph.

[0095] In this embodiment, based on each configuration item in the dynamic configuration relationship graph, its key attribute identifier and current state value are extracted, and a unique fact node is created for each key attribute. The fact node contains the attribute identifier, state value, and rule dependency model hierarchy information of the configuration item to which it belongs. The multimodal evidence body is traversed, and an evidence source node is created for each evidence data item. The evidence source node contains evidence content, evidence type, timestamp, and source protocol information. According to the attribute evidence binding relationship recorded in the multimodal evidence body, a directed edge is established between the evidence source node and the corresponding fact node. The edge direction points from the evidence source node to the fact node, representing the support relationship between the evidence and the attribute state. A type identifier is added to the directed edge according to the evidence type, including direct evidence edge, relational evidence edge, and active evidence edge. The fact nodes are connected through the relationship of their respective configuration items to form a causal dependency graph containing attribute state nodes, evidence nodes, and evidence edges.

[0096] Step S302: According to the dynamic trust chain rules, backtrack the evidence source nodes and fact nodes connected to the fact nodes in the causal dependency graph to obtain the cross-level dependency path and generate the evidence chain of the fact nodes.

[0097] The specific steps of step S302 are as follows:

[0098] Step S3021: Starting from the target fact node in the causal dependency graph, backtrack the directly connected source evidence nodes in the outward direction and integrate them into a set of direct evidence for the evidence chain.

[0099] In this embodiment, the target fact node in the causal dependency graph is taken as the starting point, and the directed edges in its outgoing direction are traversed. Edges with the connection type of direct evidence are filtered to obtain the corresponding set of evidence source nodes. The validity of the set of evidence source nodes is verified. The verified evidence source nodes are grouped according to their associated attribute dimensions to obtain direct evidence subsets organized by attributes. These direct evidence subsets are bound to the attribute identifiers of the target fact node to construct a structured storage set of direct evidence.

[0100] Step S3022: Based on the dynamic trust chain rules, take the target fact node as the starting point, backtrack in the inbound direction to the parent fact node it depends on, recursively visit the evidence source nodes connected by the parent fact node, and integrate them into a set of relational evidence.

[0101] In this embodiment, the target fact node is taken as the starting point, and backtracking is performed along the incoming edges of the causal dependency graph to identify directly connected superior fact nodes. Based on the hierarchical dependency relationship in the dynamic trust chain rules, the association strength between the superior fact node and the target fact node is determined. Superior fact nodes with association strength exceeding a preset strength threshold are added to the backtracking path. For each superior fact node in the backtracking path, the backtracking operation is recursively performed until no new superior fact nodes exist. During the recursive backtracking process, evidence source nodes connected to each superior fact node through outgoing edges are collected, and evidence source nodes that already exist in the direct evidence set are removed. The collected evidence source nodes are grouped according to the hierarchical position of their respective superior fact nodes to form a relational evidence set. It should be noted that the strength threshold is set by those skilled in the art based on the actual situation.

[0102] Step S3023: Retrieve state change event records related to the configuration item attributes represented by the target fact node based on the causal dependency graph, and integrate the corresponding evidence source nodes into an active evidence set.

[0103] In this embodiment, based on the resource identifier of the key attribute of the configuration item to which the target fact node belongs as the primary key, the state change event records within the most recent N revision periods are retrieved, where N represents the number of revision periods and the value of N is dynamically calculated based on the stability index of the configuration item. The returned state change event records are filtered according to the operation type, retaining three types of events: attribute-level change operations, resource configuration operations, and dependency update operations. These are instantiated into corresponding evidence source nodes, and each evidence source node is injected with metadata such as change operation type, revision version number, operation checksum, and forward version pointer, to obtain the active evidence set.

[0104] Step S3024: Merge and deduplicate the direct evidence set, relational evidence set, and activity evidence set, and process them according to their respective rule dependency model levels and evidence types to generate an evidence chain containing multimodal evidence and cross-level dependency relationships.

[0105] In this embodiment, the identifiers of the evidence source nodes in the three evidence sets are compared. When multiple evidence source nodes are found to point to the same data entity, the evidence source node with the latest timestamp is retained to complete the merging and deduplication of the evidence sets. According to the hierarchical identifier of the configuration item corresponding to the evidence source node in the rule dependency model, the deduplicated evidence source nodes are grouped by level. Within the same level group, they are classified into three types of evidence: direct evidence, relational evidence, and activity evidence. Based on the inter-level dependency relationship defined in the rule dependency model, a cross-level evidence connection path is established. The classified evidence source nodes are associated and bound with the corresponding fact nodes to construct an evidence chain with hierarchical structure and type characteristics.

[0106] Step S303: Extract the structural feature parameters of the evidence chain corresponding to the fact node, including the depth feature parameter determined by the number of configuration levels traversed by the evidence chain, the breadth feature parameter determined by the number of independent evidence source nodes, and the diversity feature parameter determined by the heterogeneity of evidence types.

[0107] In this embodiment, based on the evidence source nodes and their associated configuration item hierarchy identifiers contained in the evidence chain corresponding to the fact node, the number of different levels appearing in the evidence chain is counted and quantified as the evidence chain depth feature parameter. All evidence source nodes contained in the evidence chain are deduplicated and the same evidence source node repeatedly referenced through different paths is removed. The total number of independent evidence source nodes is quantified as the evidence chain breadth feature parameter. Based on the distribution of evidence types corresponding to the evidence source nodes in the evidence chain, the degree of heterogeneity of evidence types is determined by calculating information entropy and quantified as the evidence chain diversity feature parameter.

[0108] Step S304: Calculate the robustness index of the evidence chain using its depth, breadth, and diversity feature parameters.

[0109] In this embodiment, the depth feature parameters are normalized to convert the number of rule-dependent model levels traversed by the evidence chain into a standard depth value. When the standard depth value is greater than a preset level threshold, a depth gain coefficient is set to weight the standard depth value. The depth gain coefficient changes with the increase in the number of levels traversed. The breadth feature parameters are logarithmically transformed to convert the number of independent evidence source nodes into a breadth value. The breadth value is corrected according to the distribution uniformity of the evidence source nodes. When the evidence source nodes are concentrated in a few levels, a breadth attenuation factor is set. The diversity feature parameters are subjected to dispersion analysis to convert the heterogeneity of evidence types into a diversity coefficient. When the evidence type includes three complete types of evidence—direct evidence, relational evidence, and active evidence—the diversity gain coefficient is activated to adjust the diversity coefficient. The processed depth value, breadth value, and diversity coefficient are nonlinearly fused to obtain the robustness index. It should be noted that the level threshold is set by those skilled in the art according to the actual situation.

[0110] exist Figure 2In this process, depth, breadth, and diversity feature parameters are obtained based on the evidence chain feature parameters. For the depth feature parameter, the number of rule dependency model levels traversed by the evidence chain is normalized to a standard depth value. If this value exceeds a preset level threshold, a depth gain coefficient is activated to weight and amplify it, resulting in the final depth value. For the breadth feature parameter, the number of independent evidence sources is converted into an effective breadth value through logarithmic transformation, and the distribution uniformity of evidence sources across different levels is checked. If the distribution is too concentrated, a breadth attenuation factor is introduced for correction, resulting in the final breadth value. For the diversity feature parameter, the heterogeneity of evidence types is converted into a diversity coefficient. If the evidence chain contains all three types of complete evidence—direct evidence, relational evidence, and active evidence—the diversity gain coefficient is activated to adjust the diversity coefficient, resulting in the final diversity coefficient. The final depth value, final breadth value, and final diversity coefficient are then nonlinearly fused to calculate the robustness index.

[0111] Step S4: When a data conflict is detected in the causal dependency graph, the minimum evidence source node cut set is solved by the robustness index, the conflict subgraph is arbitrated to resolve the conflict, the configuration item status after arbitration is obtained, and the configuration item is intelligently optimized. The conflict subgraph is obtained by partitioning the causal dependency graph.

[0112] The specific steps of step S4 are as follows:

[0113] Step S401: Based on the causal dependency graph, compare the states of fact nodes pointing to the same key attribute of the configuration item, divide the fact nodes with inconsistent states and all their associated evidence source nodes to obtain the conflict subgraph.

[0114] In this embodiment, the set of fact nodes pointing to the same key attribute of the configuration item is searched according to the causal dependency graph. The state value of each node in the set of fact nodes is traversed. Fact nodes with inconsistent state values ​​are marked as conflicting fact nodes. Starting from the conflicting fact nodes, the directed edges in the causal dependency graph are traversed in reverse to collect directly connected evidence source nodes. These nodes are merged with the set of conflicting fact nodes, and the directed edges connecting the corresponding nodes in the causal dependency graph are extracted to construct a conflict subgraph.

[0115] Step S402: Model the conflict subgraph as a flow network. Analyze the flow network to calculate the maximum flow from the evidence source node to the conflict fact node, thereby determining the cut set with the minimum capacity and obtaining the cut set of the minimum evidence source node.

[0116] The specific steps of step S402 are as follows:

[0117] Step S4021: Set the evidence source node of the conflict subgraph as the source node of the flow, set the conflict fact node with state conflict as the sink node of the flow, set directed edges according to the source node and sink node of the flow, and establish the flow network.

[0118] In this embodiment, all evidence source nodes in the conflict subgraph are marked as the source nodes of the flow, and all fact nodes with inconsistent states in the conflict subgraph are identified and marked as the sink nodes of the flow. Based on the directed edge connection relationship in the causal dependency graph, a flow network edge set from the evidence source node to the fact node is established. When the evidence source node is connected to multiple fact nodes, the original multi-way connection structure is maintained in the flow network to obtain the complete flow network.

[0119] Step S4022: Based on the dynamic trust chain rules, assign capacity weights to the edges of nodes in the flow network. The size of the capacity weights is determined according to the level size, inter-level associations, and the edges corresponding to the active evidence.

[0120] In this embodiment, based on the dynamic trust chain rules, a basic capacity weight is assigned to the edges pointing from lower-level nodes to upper-level nodes. The initial value of the basic capacity weight is set according to the trust benchmark value of the corresponding level in the rule dependency model. The trust benchmark value is obtained by statistical analysis of the historical arbitration success rate. A capacity weight after being reduced by the level decay factor is assigned to the edges pointing from upper-level nodes to lower-level nodes. The certainty of the association represented by the edge is analyzed. When the edge corresponds to a strong association, the capacity weight is controlled by the association enhancement coefficient. When the edge corresponds to a weak association, the capacity weight is controlled by the association weakening coefficient. It is detected whether each edge is associated with active evidence. For edges corresponding to recent active evidence, an active evidence gain factor is introduced on the basis of the capacity weight for control. For edges without active evidence support, the current capacity weight remains unchanged.

[0121] Step S4023: Based on the source and sink of the flow and the capacity weights assigned to each edge, iteratively search for feasible paths from the source to the sink and push traffic along the feasible paths until there are no augmenting paths that satisfy the capacity weight constraints, and solve for the maximum flow of the flow network.

[0122] In this embodiment, the current flow value of all edges in the initial flow network is zero. A residual network is established and an initial feasible flow is set. A feasible path from the source to the sink is searched in the residual network. A feasible path must satisfy that the remaining capacity of each edge in the path is greater than zero. When a feasible path is found, the maximum pushable flow of the feasible path is determined. The maximum pushable flow is the minimum value of the remaining capacity of all edges on the path. The maximum pushable flow is pushed along the feasible path, and the current flow value of each edge on the path is updated. At the same time, the remaining capacity of each edge in the residual network is adjusted accordingly. For forward edges, the remaining capacity is reduced, and for reverse edges, the remaining capacity is increased. The feasible path search and flow pushing process is repeated until there is no feasible path from the source to the sink in the residual network. The sum of all flows flowing out of the source is accumulated to obtain the maximum flow value of the flow network.

[0123] Step S4024: Based on the maximum flow identification of the flow network, identify all saturated edges whose flow is equal to their capacity weight, and construct the minimum evidence source node cut set through the evidence source nodes directly connected to the saturated edges.

[0124] In this embodiment, the actual flow value of a directed edge in the flow network is compared with its capacity weight. When the actual flow value of a directed edge is equal to its capacity weight, the edge is marked as a saturated edge and its current flow state is recorded. Based on the evidence source nodes directly connected by the saturated edge, a set of evidence source nodes is obtained. If the same evidence source node is connected to a conflict fact node through multiple saturated edges, the evidence source node is marked as a key conflict source. Redundancy detection is performed on the evidence source node set. When a certain evidence source node can be replaced by other evidence source nodes after removal, the evidence source node is removed from the candidate set. The evidence source node set after redundancy detection is taken as the minimum evidence source node cut set.

[0125] Step S403: Eliminate conflicts in the conflict subgraph based on the minimum evidence source node cut set, compare the states of the fact nodes, and take the state of the fact node with the highest robustness index as the arbitrated state of the configuration item key attribute.

[0126] The specific steps of step S403 are as follows:

[0127] Step S4031: Based on the minimum evidence source node segmentation set, detect the conflict state of evidence source nodes pointing to the same fact node in the conflict subgraph, and eliminate the evidence source nodes and their associated edges that are judged to be untrustworthy in the conflict subgraph according to the multimodal evidence body and dynamic trust chain rules, and construct the corrected conflict subgraph.

[0128] In this embodiment, for each evidence source node in the minimum evidence source node cut set, the state consistency between the evidence source nodes associated with the same fact node is detected. When a state inconsistency is detected, the evidence quality vector of the corresponding evidence source node is extracted from the multimodal evidence body. Combined with the current decision weight of the level to which the evidence source node belongs in the dynamic trust chain rules, the credibility score of the evidence source node is calculated. If the credibility score is lower than the preset credibility threshold, the evidence source node is marked as an untrustworthy node. For evidence source nodes with direct dependencies, when the upper-level evidence source node it supports is marked as an untrustworthy node, the credibility score of the lower-level evidence source node is adjusted according to the strength coefficient of the dependency link. When the adjusted credibility score is lower than the credibility threshold, the downstream evidence source node is marked as an untrustworthy node. All evidence source nodes marked as untrustworthy and their associated edges are removed from the conflict subgraph, generating a conflict subgraph containing only trustworthy evidence source nodes. It should be noted that the credibility threshold is set by those skilled in the art according to the actual situation.

[0129] Step S4032: Based on the state of the fact nodes supported by the evidence source nodes, calculate the consistency of the logical relationship of the key attributes of the configuration items in the corrected conflict subgraph to obtain the global consistency score.

[0130] In this embodiment, based on the fact node pairs connected by logical edges in the conflict subgraph, the compatibility of their respective state values ​​under preset logical relationship constraints is compared. When the state value satisfies the logical relationship constraint, the edge is marked as a consistent state; when the state value violates the logical relationship constraint, the edge is marked as a conflict state. The number of consistent and conflict state edges in the conflict subgraph is counted, and the basic consistency ratio is calculated. Relationship weight coefficients are set for different logical relationship types, with the relationship weight coefficient for cross-level logical edges being higher than that for logical edges at the same level. Based on the hierarchical position of the configuration items involved in the conflict state edge in the rule dependency model, the conflict severity index is calculated. The distribution of the basic consistency ratio, relationship weight coefficients, and conflict severity index are input into the consistency calculation function. The consistency calculation function adjusts the basic consistency ratio based on the relationship weight coefficients and attenuates the adjusted consistency ratio based on the conflict severity index, finally obtaining the global consistency score. It should be noted that the logical relationship constraints are set by those skilled in the art according to the actual situation.

[0131] Step S4033: Based on the current conflict configuration item preset repair operation, combined with the global consistency score, perform simulated repair operation and evaluate the state deviation to obtain the virtual operation success rate.

[0132] In this embodiment, a repair operation is preset based on the current conflicting configuration item. The repair operation specifically includes the operation type, execution parameters, and a set of dependent components. The basic success rate of the repair operation is initialized based on the global consistency score. When the global consistency score is greater than or equal to the consistency threshold, the basic success rate is increased by a preset ratio. When the global consistency score is less than the consistency threshold, the basic success rate is adjusted by a consistency decay coefficient. The repair operation is simulated and its impact on the state of related configuration items is detected. Based on the deviation between the current state and the expected state of the configuration items within the impact range, the local success rate of the repair operation is calculated. The joint success rate is calculated using a forward propagation algorithm to obtain the virtual operation success rate. It should be noted that the repair operation, consistency threshold, and preset ratio are set by those skilled in the art according to the actual situation.

[0133] Step S4034: Construct an arbitration utility function, taking the global consistency score, virtual operation success rate, and the average robustness index of the evidence source node set in the modified conflict subgraph as inputs, and determine the status of the configuration items after arbitration by calculating the arbitration level value.

[0134] In this embodiment, a basic arbitration level value is obtained by calculating the geometric mean of the global consistency score, the virtual operation success rate, and the average robustness index. It is then determined whether the basic arbitration level value is lower than the virtual operation success rate. If it is lower, the arithmetic mean of the basic arbitration level value and the virtual operation success rate is calculated as the adjusted arbitration level value. If it is not lower, the basic arbitration level value is directly used as the arbitration level value. The arbitration level value is compared with a preset decision threshold. If the arbitration level value is greater than or equal to the preset decision threshold, the status of the configuration item after arbitration is determined to be valid. If the arbitration level value is less than the preset decision threshold, the status of the configuration item after arbitration is determined to be invalid. It should be noted that the decision threshold is set by those skilled in the art according to the actual situation.

[0135] Step S404: Update the dynamic configuration relationship graph according to the configuration item status after arbitration, record the conflict subgraph, the minimum evidence source node cut set, the robustness index and the configuration item status after arbitration, and fuse them into the multimodal evidence body as feedback data.

[0136] In this embodiment, the target configuration items involved in arbitration are determined based on the dynamic configuration relationship graph, and their key attribute states are replaced with the configuration item states after arbitration. When there is a cross-level association between configuration items, the topological structure features of the conflict subgraph are extracted according to the rule dependency model, including the hierarchical distribution pattern of conflict fact nodes, the type composition and connection density of evidence source nodes, and are encoded as conflict mode fingerprints. A mapping relationship is established between the minimum evidence source node cut set and the robustness index of the corresponding evidence chain to generate evidence reliability calibration records. Based on the conflict feature vector and the evidence reliability calibration records, the evidence quality vector in the multimodal evidence body is adjusted, and the initial weight estimate of similar evidence in the minimum evidence source node cut set is reduced.

[0137] Step S405: By analyzing the multimodal evidence body, periodically adjust the dynamic trust chain rules, and perform correction and reset operations on the configuration items according to the status of the configuration items after arbitration and the optimized dynamic trust chain rules.

[0138] In this embodiment, the frequency of triggering dynamic trust chain rules during the arbitration process and the consistency of the arbitration results are statistically analyzed. When the triggering frequency of dynamic trust chain rules exceeds the activity threshold and the consistency of the arbitration results is lower than the trust threshold, the dynamic trust chain rules are marked as rules to be optimized. The differences of the rules to be optimized in the context of various configuration items are analyzed, and the weight deviation coefficients under different rule dependency model layer combinations are calculated. When the weight deviation coefficients exceed the adaptive tolerance, the basic weights of the rules to be optimized are recalibrated according to the arbitration success rate of the rules to be optimized on high-level key configuration items. Based on the configuration item status after arbitration and the optimized dynamic trust chain rules, the automated operation and maintenance platform interface is called to perform status reset or configuration push for configuration items in abnormal states, and a service restart or configuration reload process is initiated. The results of all repair operations are verified to confirm that the current state is consistent with the state after arbitration. It should be noted that the activity threshold, trust threshold, and adaptive tolerance are set by those skilled in the art according to the actual situation.

[0139] exist Figure 3 In the diagram, numbers represent capacity weights, arrows represent directed edges, rectangle S represents the source node, rectangle T represents the sink node, and rectangles A, B, C, and D represent intermediate nodes.

[0140] Example 2:

[0141] Please see Figure 4 The present invention provides an embodiment of a cloud configuration item intelligent optimization system, comprising a configuration perception module, an evidence fusion module, an arbitration decision module, and an execution module.

[0142] The configuration awareness module is used to collect heterogeneous configuration item data in the cloud environment, map the configuration item data to different levels in the preset rule dependency model, and construct a dynamic configuration relationship graph.

[0143] The evidence fusion module is used to construct a multimodal evidence body for the key attributes of the configuration item based on the dynamic configuration relationship graph, and to set dynamic trust chain rules based on the rule dependency model.

[0144] The arbitration decision module is used to instantiate evidence source nodes and fact nodes and construct a causal dependency graph based on the dynamically configured relationship graph and multimodal evidence body. It then performs backtracking analysis on the causal dependency graph through dynamic trust chain rules and calculates its robustness index.

[0145] The execution module is used to solve the minimum evidence source node cut set by robustness index when a data conflict is detected in the causal dependency graph, to perform conflict resolution arbitration on the conflict subgraph, to obtain the configuration item status after arbitration, and to perform intelligent optimization on the configuration item. The conflict subgraph is obtained by partitioning the causal dependency graph.

[0146] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0147] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Any improvements and modifications made by those skilled in the art under the guidance of the present invention without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for intelligent optimization of cloud configuration items, characterized in that, include: Collect heterogeneous configuration item data in the cloud environment, map the configuration item data to different levels in the preset rule dependency model, and construct a dynamic configuration relationship graph; Based on the dynamic configuration relationship graph, a multimodal evidence body is constructed for the key attributes of the configuration item, and dynamic trust chain rules based on the rule dependency model are set. Based on the dynamic configuration relationship graph and multimodal evidence body, the evidence source node and fact node are instantiated and a causal dependency graph is constructed. The causal dependency graph is backtracked and analyzed through dynamic trust chain rules, and its robustness index is calculated. When a data conflict is detected in the causal dependency graph, the minimum evidence source node cut set is solved by the robustness index, the conflict subgraph is arbitrated to resolve the conflict, the configuration item status after arbitration is obtained, and the configuration item is intelligently optimized. The conflict subgraph is obtained by partitioning the causal dependency graph. The process involves instantiating source nodes and fact nodes based on a dynamically configured relationship graph and multimodal evidence body, constructing a causal dependency graph, and performing backtracking analysis on the causal dependency graph using dynamic trust chain rules to calculate its robustness index, including: Based on the dynamic configuration relationship graph and the multimodal evidence body, the key attributes of the configuration items are instantiated as fact nodes, and the evidence data in the multimodal evidence body is instantiated as evidence source nodes. Directed edges are established between the evidence source nodes and fact nodes to construct a causal dependency graph. According to the dynamic trust chain rules, the evidence source nodes and fact nodes connected to the fact nodes in the causal dependency graph are traced back to obtain the cross-level dependency path and generate the evidence chain of the fact nodes. Extract the structural feature parameters of the evidence chain corresponding to the fact node, including the depth feature parameter determined by the number of configuration levels traversed by the evidence chain, the breadth feature parameter determined by the number of independent evidence source nodes, and the diversity feature parameter determined by the heterogeneity of evidence types. The robustness index is calculated using the depth, breadth, and diversity features of the evidence chain. When a data conflict is detected in the causal dependency graph, the minimum evidence source node cut set is solved by the robustness index, and the conflict subgraph is arbitrated to obtain the configuration item state after arbitration. The configuration item is then intelligently optimized. The conflict subgraph is obtained by partitioning the causal dependency graph and includes: Based on the causal dependency graph, the states of fact nodes pointing to the same key attribute of the configuration item are compared, and the fact nodes with inconsistent states and all their associated evidence source nodes are divided to obtain the conflict subgraph. The conflict subgraph is modeled as a flow network. Through flow network analysis, the maximum flow from the evidence source node to the conflict fact node is calculated, the minimum capacity cut set is determined, and the minimum evidence source node cut set is obtained. The conflict subgraph is refurbished based on the minimum evidence source node cut set, and the states of the fact nodes are compared. The state of the fact node with the highest robustness index is used as the arbitrated state of the configuration item's key attributes. The dynamic configuration relationship graph is updated based on the status of the configuration items after arbitration. The conflict subgraph, the minimum evidence source node cut set, the robustness index, and the status of the configuration items after arbitration are recorded and fused into the multimodal evidence body as feedback data. By analyzing multimodal evidence, the dynamic trust chain rules are periodically adjusted, and correction and reset operations are performed on the configuration items based on the status of the configuration items after arbitration and the optimized dynamic trust chain rules.

2. The intelligent optimization method for cloud configuration items according to claim 1, characterized in that, The process of constructing multimodal evidence bodies for key attributes of configuration items based on a dynamic configuration relationship graph, and setting dynamic trust chain rules based on a rule dependency model, includes: For the key attributes of configuration item nodes in the dynamic configuration relationship graph, we extract their directly reported values ​​as direct evidence, extract relationship evidence from adjacent nodes by traversing the relationship edges of configuration item nodes, and extract activity evidence through the event stream related to configuration item nodes. Direct evidence, relational evidence, and activity evidence are bound to key attributes in the form of attribute evidence pairs. A multimodal evidence body of the key attributes of the corresponding configuration item node is constructed by combining the evidence quality vector. The evidence quality vector is calculated based on direct evidence, relational evidence, and activity evidence. Based on the hierarchical derivation of the dependency relationships in the rule dependency model, the rules for trust transfer and conflict resolution are derived and formalized into a production rule set as the initial dynamic trust chain rules. Calculate the historical validity and context relevance coefficients for the initial dynamic trust chain rules, determine the decision weights and priority parameters based on the historical validity and context relevance coefficients, and generate dynamic trust chain rules.

3. The intelligent optimization method for cloud configuration items according to claim 2, characterized in that, The process of calculating historical validity and contextual relevance coefficients for the initial dynamic trust chain rules, determining decision weights and priority parameters based on these coefficients, and generating dynamic trust chain rules includes: Based on the execution status of the initial dynamic trust chain rules, an effectiveness vector is constructed by the number of successful arbitrations and the total number of triggers. The determination of the number of successful arbitrations is based on the fact that the result of this arbitration remains stable within the next time window. Set the context features of the configuration items according to the configuration item type and data source type, calculate the distribution similarity of the context features, and normalize the distribution similarity to obtain the context relevance coefficient. The distribution characteristics of historical validity and context relevance coefficients are statistically analyzed to determine the critical interval of decision weights. Based on the critical interval, a piecewise nonlinear transformation function is constructed, and the decision weights are calculated in combination with a preset adjustment factor. The initial dynamic trust chain rules are prioritized based on the decision weights to obtain the dynamic trust chain rules.

4. The intelligent optimization method for cloud configuration items according to claim 3, characterized in that, According to the dynamic trust chain rules, the process of tracing back the evidence source nodes and fact nodes connected to the fact nodes in the causal dependency graph to obtain cross-level dependency paths and generate evidence chains for the fact nodes includes: Starting from the target fact node in the causal dependency graph, trace back the directly connected source evidence nodes in the outward direction and integrate them into a set of direct evidence in the chain of evidence. According to the dynamic trust chain rules, the target fact node is taken as the starting point, and the parent fact node it depends on is traced back in the inbound direction. The evidence source nodes connected by the parent fact node are recursively accessed and integrated into a set of relational evidence. Based on the causal dependency graph, retrieve the state change event records related to the configuration item attributes represented by the target fact node, and integrate the corresponding evidence source nodes into an active evidence set; The direct evidence set, relational evidence set, and activity evidence set are merged and deduplicated. They are then processed according to their respective rule dependency model levels and evidence types to generate an evidence chain containing multimodal evidence and cross-level dependency relationships.

5. The intelligent optimization method for cloud configuration items according to claim 4, characterized in that, The process of modeling the conflict subgraph as a flow network, calculating the maximum flow from the evidence source node to the conflicting fact node through flow network analysis, determining the minimum capacity cut set, and obtaining the minimum evidence source node cut set includes: Set the evidence source node of the conflict subgraph as the source node of the flow, set the conflict fact node with state conflict as the sink node of the flow, and set directed edges according to the source node and sink node of the flow to establish the flow network. Based on the dynamic trust chain rules, the edges of nodes in the flow network are assigned capacity weights, where the size of the capacity weight is determined according to the level size, level association and the edge corresponding to the active evidence. Based on the source and sink of the flow and the capacity weights assigned to each edge, the maximum flow of the flow network is obtained by iteratively searching for feasible paths from the source to the sink and pushing traffic along the feasible paths until there are no augmenting paths that satisfy the capacity weight constraints. The maximum flow identification method based on flow networks identifies all saturated edges whose flow is equal to their capacity weight, and constructs a minimum evidence source node cut set by connecting the evidence source nodes directly to the saturated edges.

6. The intelligent optimization method for cloud configuration items according to claim 5, characterized in that, The process of eliminating conflicts in the conflict subgraph based on the minimum evidence source node cut set, comparing the states of the fact nodes within it, and using the fact node state with the highest robustness index as the arbitrated configuration item state for the key attributes of the configuration item, includes: Based on the minimum evidence source node segmentation set, the conflict state of evidence source nodes pointing to the same fact node in the conflict subgraph is detected, and the evidence source nodes and their associated edges that are judged to be untrustworthy in the conflict subgraph are eliminated according to the rules of multimodal evidence body and dynamic trust chain, and the corrected conflict subgraph is constructed. Based on the state of the fact nodes supported by the evidence source nodes, the consistency degree of the logical relationship of the key attributes of the configuration items is calculated on the corrected conflict subgraph to obtain a global consistency score. Based on the current conflict configuration items, preset repair operations are performed. Combined with the global consistency score, simulated repair operations are executed and the state deviation is evaluated to obtain the virtual operation success rate. An arbitration utility function is constructed, which takes the global consistency score, the success rate of virtual operations, and the average robustness index of the evidence source node set in the modified conflict subgraph as inputs. The status of the configuration items after arbitration is determined by calculating the arbitration level value.

7. The intelligent optimization method for cloud configuration items according to claim 6, characterized in that, The process involves collecting heterogeneous configuration item data from the cloud environment, mapping this data to different levels in a pre-defined rule dependency model, and constructing a dynamic configuration relationship graph, including: The target cloud environment is queried to obtain the raw data streams of computing, network, storage and application services, and structured configuration item data is generated through feature extraction; By using a pre-defined rule dependency model, the type and semantic features of configuration items are mapped to the corresponding levels; Based on the hierarchical mapping results and the relational attributes contained in the configuration item data, logical edges are established between configuration items at the same or different levels through relation inference, resulting in a dynamic configuration relation graph.

8. A cloud configuration item intelligent optimization system, used to implement the cloud configuration item intelligent optimization method according to any one of claims 1-7, characterized in that, It includes a configuration perception module, an evidence fusion module, an arbitration decision module, and an enforcement module; The configuration awareness module is used to collect heterogeneous configuration item data in the cloud environment, map the configuration item data to different levels in the preset rule dependency model, and construct a dynamic configuration relationship graph. The evidence fusion module is used to construct a multimodal evidence body for the key attributes of the configuration item based on the dynamic configuration relationship graph, and to set dynamic trust chain rules based on the rule dependency model. The arbitration decision module is used to instantiate evidence source nodes and fact nodes and construct a causal dependency graph based on the dynamically configured relationship graph and multimodal evidence body, and to perform backtracking analysis on the causal dependency graph through dynamic trust chain rules to calculate its robustness index. The execution module is used to solve the minimum evidence source node cut set by robustness index when a data conflict is detected in the causal dependency graph, to perform conflict resolution arbitration on the conflict subgraph, to obtain the configuration item status after arbitration, and to perform intelligent optimization on the configuration item. The conflict subgraph is obtained by partitioning the causal dependency graph.

Images