A service orchestration system and optimization method based on rule engine

Abstract

The present invention relates to the field of digital service technology, and the present invention discloses a service orchestration system and optimization method based on a rule engine, including a rule parsing module, a service registration center module, an orchestration engine module, an execution engine module, a monitoring feedback module, and a fault-tolerant module. The present invention introduces feature vectors to carry out fine-grained modeling of service attributes, which can comprehensively quantify the performance, resource consumption and dependencies of services, and provide a scientific basis for the precise disassembly of services. With the help of the optimal path search mechanism of reinforcement learning, the system can autonomously optimize the scheduling path in continuous exploration, dynamically balance service quality, resource consumption and risk control, and significantly improve the intelligent decision-making ability of the system. By establishing a fuse model, the fault-tolerant module can respond quickly and adopt corresponding strategies when facing instantaneous and persistent faults to ensure the continuity of services. It supports dual-channel input of natural language description and flowchart sketch, which lowers the threshold for system design.

Description

Technical Field

[0001] The present invention relates to the field of digital service technology, and in particular to a service orchestration system and optimization method based on a rule engine. Background Art

[0002] Service orchestration is a technology that dynamically combines multiple independent services into complete processes based on business logic. It coordinates service execution order and dependencies through a visual rules engine, enabling flexible construction and automated operation of complex business scenarios. It allows non-technical personnel to design processes using a drag-and-drop approach. Combined with real-time data awareness and intelligent scheduling strategies, it automatically optimizes execution paths and provides dynamic fault tolerance and grayscale iterations during operation. Ultimately, this allows distributed services to operate together like a symphony, rapidly responding to business changes while ensuring high availability and execution efficiency.

[0003] In the prior art, Chinese patent document CN113535419A discloses a service orchestration method and device, which obtains preset service configuration information based on the interface name and business input parameters in the call request; wherein the service configuration information includes service mapping information, service call information, and the execution order of the service interface; obtains the corresponding service processing script based on the script address in the service mapping information, groups the service interfaces of the service call information according to the service interface execution order, and uses the service interfaces belonging to the same group as task nodes to generate a task queue; pulls the task node from the task queue according to the service interface execution order of the service interface in the task node, executes the service interface of the task node, and obtains the return result; uses the service processing script to process the return result and return the result. This patent document solves the technical problem that the traditional service call chain needs to modify the code to adapt to business changes, resulting in high development costs and long iteration cycles. The shortcomings are: the writing, compilation, and maintenance of the proposed solution script increase the technical threshold and system complexity, and lacks fault tolerance mechanisms such as circuit breaking and degradation, and the backup link switching logic is imperfect when the service fails.

[0004] Chinese patent document CN117251141A discloses a service orchestration and service regeneration method, platform and device, which creates a table structure based on a service database, and uses a preset custom template engine to directly reference the table structure of a preset computer language pair based on a template language to create business logic. After defining the service by calling the framework customized template, the service is registered to the preset service management platform and the service information and service interface information are configured to form a business tree according to the business identification classification; the service interface is called by the preset process management engine to realize the orchestration creation and regeneration of the service. Complex business verification and business dependencies are formed into a call chain relationship through orchestration to achieve long-term reuse and more flexible orchestration. This patent document solves the problems of high development costs, poor reusability of business logic and rigid service management. The shortcomings are: the solution proposed relies on preset templates and static business trees, lacks a dynamic adjustment mechanism, is difficult to cope with sudden business scenarios, and does not mention fault tolerance strategies.

[0005] Chinese patent document CN117251141A discloses a data sharing service intelligent orchestration technology, including: subdividing the data exchange business logic and abstracting each step into independent and reusable components; connecting the components to further form a higher-level abstract entity as a service; intelligently matching databases of different manufacturers through the abstract combination of components and services, and automatically generating corresponding data exchange services. The present invention effectively encapsulates the program carrier and business logic through the abstraction of components and services, stripping away a large amount of underlying implementation details, so that developers and implementers can focus more on the business itself, greatly simplifying the cost of writing and implementing data sharing services. This patent document solves the problem of poor cross-system compatibility, but the solution it proposes lacks a dynamic load balancing mechanism, and the parallelization efficiency may decrease in high-concurrency scenarios, and no fault tolerance mechanism is mentioned. Summary of the Invention

[0006] In order to solve the current technical problems, the main purpose of the present invention is to provide a service orchestration system and optimization method based on a rule engine. The present invention introduces feature vectors to carry out fine-grained modeling of service attributes, which can comprehensively quantify the performance, resource consumption and dependencies of services, and provide a scientific basis for the precise disassembly of services. With the help of the optimal path search mechanism of reinforcement learning, the system can autonomously optimize the scheduling path in continuous exploration, dynamically balance service quality, resource consumption and risk control, and significantly improve the intelligent decision-making ability of the system. By establishing a fuse model, the fault-tolerant module can respond quickly and adopt corresponding strategies when facing transient and persistent faults to ensure the continuity of services. It supports dual-channel input of natural language description and flowchart sketch, which lowers the threshold for system design.

[0007] In order to overcome the problems existing in the prior art, the technical solution adopted by the present invention is: a service orchestration system based on a rule engine, including the following modules:

[0008] Rule parsing module: including atomic-level parsing unit, combination-level parsing unit, rule prediction unit and multimodal rule input unit;

[0009] The atomic parsing unit is used to disassemble services and parse them into atomic services with functional characteristics;

[0010] The combination-level parsing unit is used to record the number of adjacent occurrences of service pairs and judge the relevance of services. When there are continuous high-frequency adjacent pairs, they are merged into a combined service.

[0011] The rule prediction unit is used to detect whether there is a conflict in the combined service generated by the combination-level parsing unit;

[0012] The multimodal rule input unit is used to receive natural language input and sketch input, and perform preprocessing to obtain a cross-domain perception matrix;

[0013] Service registration center module: including service DNA fingerprint unit and self-healing service unit;

[0014] The service DNA fingerprint unit is used to generate a unique hash fingerprint for each atomic service and composite service;

[0015] Self-healing service unit: used to automatically mark and remove instances of an atomic service or composite service from the available pool when the failure rate exceeds a threshold;

[0016] Orchestration engine module: used to obtain the cross-domain perception matrix and adaptively adjust the optimization strategy through the reward function;

[0017] Execution engine module: This module selects multiple optimal service links based on the results generated by the orchestration engine module, forming a multi-link parallel racing execution technology solution.

[0018] Monitoring feedback module: used to implement feedback strategies for the rule parsing module, service center registration module, and orchestration engine module;

[0019] Fault-tolerant module: used to receive real-time push status from the execution engine module, and quickly respond and adopt corresponding strategies when the execution engine module faces transient and persistent faults.

[0020] A service orchestration optimization method based on a rule engine, which adopts the service orchestration system based on a rule engine, comprises the following steps:

[0021] S1. Perform atomic-level and composite-level analysis on services to obtain atomic services and composite services, and detect potential conflicts in composite services.

[0022] S2, receiving and processing natural language input and sketch input to obtain a cross-domain perception matrix;

[0023] S3. Generate a unique hash fingerprint for each atomic service and composite service;

[0024] S4. Obtain a cross-domain perception matrix, search for the optimal path based on reinforcement learning, and independently design a reward function based on the business;

[0025] S5: Parallel testing: Initiate multiple service links simultaneously and take the first successful result.

[0026] S6. Monitor and provide feedback on each step;

[0027] S7. Receive the real-time push status of the execution engine module, and quickly respond and adopt corresponding strategies when the execution engine module faces transient and persistent failures.

[0028] In S1, services are decomposed into atomic services using atomic parsing units, which have the following functional characteristics: functional singularity, resource independence, idempotence, and temporal and spatial determinism.

[0029] Atomic parsing of a service involves the following steps:

[0030] Perform feature modeling on the service and use feature vectors to describe service attributes. The formula is:

[0031] ;

[0032] Where, The feature vector representing the service, Indicates the CPU consumption of the unit call, Indicates the memory usage. Indicates the delay reference value, Indicates the number of explicitly dependent services, represents the cascading failure coefficient;

[0033] 2) Construct Hamiltonian to define the optimal segmentation, the formula is:

[0034] ;

[0035] Where, represents the Hamiltonian, Indicates the coupling degree between services, Indicates the service atomization tendency parameter, , Indicates the split decision variable, 0 means retaining atomicity, 1 means splitting is required; is a preset weight coefficient;

[0036] The degree of coupling between services Determined by the eigenvector difference, the formula is:

[0037] ;

[0038] Where, represents the natural base, represents the attenuation coefficient, Indicates the vectors Hedi vectors The eigenvector difference of Indicates the call timing angle.

[0039] The service atomization tendency parameter Determined by resource consumption characteristics, the formula is:

[0040] ;

[0041] Where, Indicates the CPU consumption of the unit call The weight coefficient of Indicates memory usage The weight coefficient of Indicates the delay baseline value The weight coefficient of .

[0042] The segmentation is performed using the quantum annealing algorithm, which includes the following steps:

[0043] The Hamiltonian parameters are converted into the physical configuration of the quantum processor, each service is mapped to a physical qubit, and the inter-qubit coupler is set according to the Hamiltonian to apply a local magnetic field to each qubit;

[0044] Set the time evolution path of quantum annealing to improve the probability of optimal solution;

[0045] Quantum annealing is performed to initialize the quantum state, and all quantum bits are initialized to superposition state;

[0046] Screen and optimize the quantum annealing output results, repair broken solutions caused by insufficient chain strength, and sort candidate solutions in ascending order of Hamiltonian energy;

[0047] Determine the final service cutting plan and verify its reliability, and select the candidate solution with the lowest energy and satisfying the constraints.

[0048] In S1, composition-level parsing of services includes the following steps:

[0049] Construct a service co-occurrence matrix to record the number of adjacent occurrences of service pairs;

[0050] An adaptive dynamic threshold is constructed based on the mean and variance. If the number of adjacent occurrences is greater than the adaptive dynamic threshold, it is determined to be a significant association.

[0051] Among them, when there are continuous high-frequency adjacent pairs, they are merged into a combination, and partially overlapping sequences are merged.

[0052] In S1, potential conflicts in the composite service include:

[0053] Resource lock, judged by the length of the circular wait chain;

[0054] Data race, read after write or write after read of the same data item;

[0055] Permission exceeded: the first service calls the second service but does not have permission;

[0056] Temporal paradoxes, where the order of events violates the laws of causality;

[0057] Resource overload, where the cumulative resource demand is greater than the node capacity;

[0058] Business logic conflicts and mutually exclusive rules take effect at the same time.

[0059] In S2, natural language input and sketch input are received and processed to obtain a cross-domain perception matrix, including the following steps:

[0060] Processing of natural language input, including text cleaning and standardization, word segmentation and part-of-speech tagging, domain entity recognition, dependency parsing, semantic role labeling, logical predicate extraction, and intermediate representation generation;

[0061] Processing sketch input, including image preprocessing, graphic element detection, text recognition, topology relationship reconstruction, semantic enhancement and conflict resolution, and intermediate representation generation;

[0062] The processing results of natural language input and sketch input are integrated to obtain a cross-domain perception matrix.

[0063] In S4, the optimal path search based on the reinforcement learning algorithm includes constructing the state space, action space, establishing functions and network structures;

[0064] By obtaining the cross-domain perception matrix and processing it to obtain the state space, the formula is:

[0065] ;

[0066] Where, represents the state space, means flattening the three-dimensional matrix into a vector, represents vector concatenation, Represents LSTM historical state encoding;

[0067] The action space uses discrete actions and continuous parameter adjustments to implement three types of operations: node addition and deletion, path weight adjustment, and fault switching.

[0068] The total reward function is calculated as: ;

[0069] Where, represents the total reward function, represents the basic reward function, represents the domain reward function;

[0070] The formula of the basic reward function is: ;

[0071] Where, Indicates the service level compliance rate The weight coefficient of represents the weight coefficient of resource cost, Indicates the risk event count The weight coefficient of

[0072] The formula of the domain reward function is: ;

[0073] Where, represents the domain reinforcement coefficient, Indicates the number of industry pain point functions, Indicates the Industry pain point function The weight coefficient of

[0074] The network structure of the reinforcement learning algorithm adopts the Actor-Critic framework.

[0075] In S6, the monitoring feedback module feeds back data to the rule parsing module as a heat map of the rule execution path and precompiles high-frequency rules into machine code;

[0076] Dynamically adjust service routing weights based on feedback data from the service center registration module;

[0077] Optimize personalized service links based on feedback from the orchestration engine module;

[0078] By establishing monitoring indicators, threshold ranges and feedback strategies, the system operation status can be monitored and dynamically optimized in real time.

[0079] In S7, the fault-tolerant module receives the real-time status push from the execution engine module. It uses short-term memory to break the circuit and quickly recovers from transient faults. It uses long-term memory to degrade and mark persistent fault services as "long-term untrustworthy" for manual intervention. The steps are as follows:

[0080] Get the real-time status matrix pushed by the execution engine module, as follows:

[0081] ;

[0082] Where, represents the real-time state matrix, Indicates the Services in the time window The first indicator values, including success rate, delay, and error rate;

[0083] Calculate the weighted failure rate for each service using the formula:

[0084] ;

[0085] Where, Indicates the The weighted failure rate of each service, represents the time decay weight, represents the time window, Indicates the characteristic function, which takes 1 when it fails and 0 otherwise. Indicates the The service in The state of the moment;

[0086] Establish the first Abnormal fluctuation of services , construct the health evaluation index based on abnormal fluctuation and weighted failure rate, the formula is:

[0087] ;

[0088] Where, Indicates the Health evaluation indicators of services, represents the weighted failure rate The weight coefficient of The weight coefficient representing abnormal volatility, The weight coefficient indicating the health of the resource;

[0089] The service status is determined based on the health assessment indicator. If the health threshold is lower than the threshold, the circuit breaker is triggered and a backup link is called, which is selected based on its operating speed.

[0090] Construct a decaying abnormal integral to determine whether there is a long-term fault. The formula is:

[0091] ;

[0092] in, Indicates the Services at time The abnormal integral of Indicates the total number of historical abnormal events, represents the time decay coefficient, Indicates the time difference of abnormal events, Indicates the The time when the abnormality occurred, Indicates the The service in Health assessment indicators at the time of abnormality;

[0093] When the decay-type anomaly integral exceeds the threshold, it is marked as "long-term untrustworthy" and enters the manual intervention process.

[0094] Compared with the prior art, the present invention has the following beneficial effects:

[0095] 1. This invention introduces eigenvectors to fine-tune the modeling of service attributes, which can comprehensively quantify the performance, resource consumption, and dependencies of services, providing a scientific basis for the precise disassembly of services. With the help of the Hamiltonian optimization model, the service coupling degree and atomization tendency parameters are fully considered, and a more reasonable service segmentation scheme is achieved. By utilizing the global search capability of the quantum annealing algorithm, the problem of traditional algorithms easily falling into local optimality can be effectively avoided, ensuring that the partitioning results are closer to the optimal solution; in addition, the combination of the broken solution repair mechanism and the energy ascending screening strategy can further improve the stability and accuracy of the analysis.

[0096] 2. With the help of the optimal path search mechanism of reinforcement learning, the present invention enables the system to autonomously optimize the scheduling path during continuous exploration, dynamically balance service quality, resource consumption and risk control, and significantly improve the system's intelligent decision-making capabilities; at the same time, by using the domain reward function to introduce the industry pain point function, it can adaptively adjust the optimization strategy for key indicators of different industries to meet personalized business needs.

[0097] 3. By establishing a circuit breaker model, this invention enables the fault-tolerant module to quickly respond and implement appropriate strategies to transient and persistent faults, ensuring service continuity. The short-term memory circuit breaker mechanism utilizes weighted failure rates and health assessment metrics to achieve rapid circuit breaking and switchover to backup links, effectively reducing the impact of transient faults on the system and improving the real-time responsiveness of services. The long-term memory degradation mechanism uses a decaying anomaly integration method to accurately identify persistent faulty services and trigger manual intervention, further preventing long-term damage to overall system performance caused by abnormal services.

[0098] 4. The present invention supports dual-channel input of natural language description and flowchart sketch. Through deep parsing technologies such as dependency syntax analysis and semantic role labeling, it converts business requirements into structured rules, greatly reducing the threshold for non-technical personnel to participate in system design. BRIEF DESCRIPTION OF THE DRAWINGS

[0099] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0100] Figure 1 This is a structural diagram of a service orchestration system based on a rule engine provided in this application.

[0101] Figure 2 This is the flowchart of the fault-tolerant module.

[0102] Figure 3 This is a flow chart of a service orchestration optimization method based on a rule engine of the present invention. DETAILED DESCRIPTION

[0103] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0104] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0105] Example 1:

[0106] See also Figure 1 The present invention provides a service orchestration system based on a rule engine. The system includes a rule parsing module, a service registration center module, an orchestration engine module, an execution engine module, a monitoring feedback module and a fault tolerance module. The functions and composition of each module will be introduced in detail below.

[0107] Rule parsing module: includes parsing unit, rule prediction unit and multimodal rule input unit. The parsing unit includes atomic-level parsing unit and combination-level parsing unit.

[0108] Furthermore, the atomic-level parsing unit disassembles the service and parses it into atomic services with the following characteristics: functional singularity, that is, each atomic service only completes one business action, resource independence, and can be independently deployed in a container without the need for shared memory / storage, idempotence guarantee, that is, the results of any repeated execution are consistent, and spatiotemporal determinism, that is, the execution time fluctuation range is ≤5%.

[0109] Next, we will introduce how to parse and split the service;

[0110] 1) Model the service features and use a 128-dimensional feature vector to describe the service attributes. The formula is:

[0111] ;

[0112] Where, The feature vector representing the service, Indicates the CPU consumption of the unit call, Indicates the memory usage. Indicates the delay reference value, Indicates the number of explicitly dependent services, represents the cascading failure coefficient;

[0113] 2) Construct Hamiltonian to define the optimal split, the formula is:

[0114] ;

[0115] Where, represents the Hamiltonian, Indicates the coupling degree between services, Indicates the service atomization tendency parameter, , Indicates the split decision variable, 0 means retaining atomicity, 1 means splitting is required; is a preset weight coefficient;

[0116] Among them, the coupling degree between services Determined by the eigenvector difference, the formula is:

[0117] ;

[0118] Where, represents the natural base, represents the attenuation coefficient, Indicates the vectors Hedi vectors The eigenvector difference of Indicates the call timing angle, calculated based on the historical call chain.

[0119] Service atomization tendency parameters Determined by resource consumption characteristics, the formula is:

[0120] ;

[0121] Where, Indicates the CPU consumption of the unit call The weight coefficient of Indicates memory usage The weight coefficient of Indicates the delay baseline value The weight coefficient of .

[0122] Furthermore, the quantum annealing algorithm is used for segmentation, and the specific steps include:

[0123] 1) Convert the Hamiltonian parameters into the physical configuration of the quantum processor, map each service to a physical qubit, set the inter-qubit coupler according to the Hamiltonian, and apply a local magnetic field to each qubit;

[0124] 2) Setting the time evolution path of quantum annealing to improve the probability of the optimal solution;

[0125] 3) Quantum annealing is performed to initialize the quantum state, with all qubits initialized to a superposition state. The quantum tunneling phase lasts for 5 μs at s(t) = 0.4, allowing quantum tunneling to break through the local optimum. The classical annealing phase, when s(t) > 0.4, gradually transitions to the classical Ising model.

[0126] 4) Screen and optimize the quantum annealing output results to repair broken solutions caused by insufficient chain strength. For example, if the results of a service chain are [1, 1, 0] in three measurements, they are corrected to 1. Candidate solutions are sorted in ascending order of Hamiltonian energy.

[0127] 5) Determine the final service cutting plan and verify its reliability, and select the candidate solution with the lowest energy and meeting the constraints.

[0128] By introducing eigenvectors to fine-tune modeling of service attributes, we can comprehensively quantify service performance, resource consumption, and its dependencies, providing a scientific basis for accurate service disassembly. With the help of the Hamiltonian optimization model, we fully consider the service coupling degree and atomization tendency parameters, and achieve a more reasonable service segmentation solution.

[0129] By utilizing the global search capability of the quantum annealing algorithm, we can effectively avoid the problem that traditional algorithms are prone to falling into local optimality, ensuring that the partitioning results are closer to the optimal solution.

[0130] In addition, the fracture solution repair mechanism and energy ascending screening strategy are combined to further improve the stability and accuracy of the solution.

[0131] Next, we'll introduce the compositional parsing unit. We first construct a service co-occurrence matrix, recording the number of adjacent occurrences of service pairs. Table 1 below shows an example matrix. Rows represent the starting service, columns represent the subsequent services, and cell values ​​represent the number of calls from the starting service to the subsequent service. A diagonal value of 0 indicates that self-calls are not recorded. For example, service FW3 appears after service FW2 879 times.

[0132] Then, an adaptive dynamic threshold is constructed based on the mean and variance. If the number of adjacent occurrences is greater than the adaptive dynamic threshold, it is determined to be a significant association.

[0133] Furthermore, when there are continuous high-frequency adjacent pairs, they are merged into a combination, and partially overlapping sequences are allowed to be merged. For example, if combination 1 is ABC and combination 2 is BCD, they can be merged into a new combination ABCD.

[0134] Table 1. Example of service co-occurrence matrix

[0135]

[0136] By constructing a service co-occurrence matrix, we can quantify the temporal correlation between services. By introducing an adaptive dynamic threshold based on mean and variance, we can effectively improve the accuracy of identifying significant correlations, ensure that the judgment criteria can be dynamically adjusted under different scenarios and traffic fluctuations, and enhance the generalization of the method.

[0137] The strategy of merging continuous high-frequency adjacent pairs is adopted to allow the integration of partially overlapping sequences, fully retaining the inherent calling logic of the service and avoiding the problem of excessive service granularity caused by excessive disassembly.

[0138] The rule prediction unit is used to detect whether there are conflicts in the combined services generated by the combination-level parsing unit, covering six core conflict scenarios.

[0139] Including: resource lock, using the length of the circular wait chain to judge;

[0140] Data race, read after write or write after read of the same data item;

[0141] Permission exceeded, service FW1 calls service FW2 but does not have permission;

[0142] Temporal paradoxes, where the order of events violates the laws of causality;

[0143] Resource overload, where the cumulative resource demand is greater than the node capacity;

[0144] Business logic conflicts and mutually exclusive rules take effect at the same time.

[0145] Through the six core conflict scenarios, we can fully identify potential risks in combined services, avoid system instability or functional failure due to abnormal combinations, and help optimize service orchestration and improve system stability, security and operational efficiency.

[0146] The multimodal rule input unit is used to receive natural language input and sketch input, and perform preprocessing to obtain a cross-domain perception matrix. The specific method is:

[0147] First, the natural language input is processed, including text cleaning and standardization, word segmentation and part-of-speech tagging, domain entity recognition, dependency syntax analysis, semantic role labeling, logical predicate extraction and intermediate representation generation.

[0148] Among them, text cleaning and standardization are the first steps in natural language processing, aiming to convert the original text into a unified and clean format. Word segmentation and part-of-speech tagging divide continuous text into meaningful word units.

[0149] Chinese word segmentation is particularly important, for example, "natural language processing" is segmented into "natural language / processing";

[0150] Domain entity recognition refers to identifying entities in a specific field in text, such as disease names in the medical field and stock codes in the financial field. Unlike general named entity recognition (NER), it focuses on domain-specific vocabulary.

[0151] Dependency parsing refers to analyzing the grammatical dependencies between words in a sentence, constructing a syntactic tree, and revealing structures such as subject-predicate, verb-object, etc.

[0152] Semantic role labeling refers to determining the semantic roles of each component in a sentence, such as agent (who does it), patient (what does it), time, place, etc.

[0153] Logical predicate extraction refers to converting sentences into logical expressions, extracting predicates (actions / states) and their parameters, and forming a computable logical form; intermediate representation generation refers to integrating all analysis results to generate a structured intermediate representation (such as JSON, logical expressions, knowledge graph triples) for easy use by subsequent rule engines or business systems.

[0154] The sketch input is then processed, including image preprocessing, graphic element detection, text recognition, topological relationship reconstruction, semantic enhancement and conflict resolution, and intermediate representation generation.

[0155] Image preprocessing includes grayscale, denoising, binarization, and morphological optimization to eliminate sketch noise and enhance key features, providing clear input for subsequent analysis.

[0156] Graphic element detection is used to identify basic graphic elements (nodes, arrows) and their geometric properties;

[0157] Text recognition is used to identify text inside graphic elements;

[0158] Topological relationship reconstruction is used to establish logical connection relationships between graphic elements;

[0159] Semantic enhancement and conflict resolution include node type inference, conditional expression parsing, and conflict detection;

[0160] Intermediate representation generation is used to generate SVG vector graphics for manual secondary confirmation, and can record the semantic differences of each modification.

[0161] By processing the input sketch, it can handle line breaks in hand-drawn sketches (automatically connecting within 10 pixels) and slight overlaps (element spacing > 5 pixels without misjudgment), quickly adapt to the graphic semantic rules of different industries such as finance, medical care, and e-commerce, and complete end-to-end processing from sketch scanning to intermediate representation generation within 1 second.

[0162] The processing results of natural language input and sketch input are integrated to obtain a cross-domain perception matrix.

[0163] The present invention integrates the processing results of natural language input and sketch input into a cross-domain perception matrix, achieves deep fusion of multimodal data, breaks through the limitations of a single data type, and significantly improves the comprehensiveness and accuracy of rule parsing.

[0164] The service registration center module includes a service DNA fingerprint unit and a self-healing service unit.

[0165] The service DNA fingerprint unit generates a unique hash fingerprint for each atomic service and composite service, including interface signatures, performance baselines, dependency library versions, etc., to ensure version consistency;

[0166] The self-healing service unit includes automatically marking an atomic service or composite service instance as "pending repair" and removing it from the available pool when the failure rate is greater than 5%.

[0167] The orchestration engine module is used to build a "business-technology-environment" trinity perception network, including: obtaining a cross-domain perception matrix, searching for the optimal path based on reinforcement learning, and independently designing a reward function based on the business to achieve a domain-adaptive strategy targeting industry pain points.

[0168] Among them, the optimal path search based on the reinforcement learning algorithm includes constructing the state space, action space, establishing functions and network structures; first, obtain the cross-domain perception matrix constructed above, and obtain the state space through processing. The formula is:

[0169] ;

[0170] in, represents the state space, means flattening the three-dimensional matrix into a vector, represents vector concatenation, Represents the LSTM history state encoding.

[0171] The action space uses discrete actions and continuous parameter adjustments to implement three types of operations: node addition and deletion, path weight adjustment, and fault switching.

[0172] The reward function includes the total reward function, the basic reward function and the domain reward function.

[0173] The total reward function is calculated as: ;

[0174] Where, represents the total reward function, represents the basic reward function, represents the domain reward function.

[0175] The formula of the basic reward function is: ;

[0176] Where, Indicates the service level compliance rate The weight coefficient of represents the weight coefficient of resource cost, Indicates the risk event count The weight coefficient of .

[0177] The formula of the domain reward function is: ;

[0178] Where, represents the domain reinforcement coefficient, Indicates the number of industry pain point functions, Indicates the Industry pain point function The weight coefficient of industry pain points includes the success rate of financial anti-fraud, the OEE of manufacturing equipment, etc. Table 2 shows an example of the mapping of industry pain points.

[0179] Table 2. Industry pain point mapping example

[0180]

[0181] In this embodiment, the network structure of the reinforcement learning algorithm adopts the Actor-Critic framework.

[0182] By leveraging the optimal path search mechanism of reinforcement learning, the present invention enables the system to autonomously optimize the scheduling path during continuous exploration, dynamically balance service quality, resource consumption, and risk control, and significantly enhance the system's intelligent decision-making capabilities.

[0183] At the same time, by using the domain reward function to introduce the industry pain point function, it is possible to adaptively adjust the optimization strategy according to the key indicators of different industries to meet personalized business needs.

[0184] The execution engine module is used for parallel testing, initiating three different service links at the same time, namely the main link + backup link 1 + backup link 2, and taking the result of the first successful one.

[0185] Furthermore, after the orchestration engine module generates the service link, the execution engine module selects the three best service links based on the generation results, forming a three-link parallel racing execution technical solution, ensuring that the three links are strictly parallel, without queue waiting delays, and automatically downgrade to dual-link mode when the rejection policy is triggered.

[0186] The monitoring feedback module is used for precise module-oriented feedback strategies. The feedback data from the rule parsing module is a heat map of the rule execution path, and high-frequency rules are pre-compiled into machine code. The service routing weight is dynamically adjusted based on the feedback data from the service center registration module. The personalized service link is optimized based on the feedback from the orchestration engine module.

[0187] The monitoring and feedback module establishes monitoring indicators, threshold ranges, and feedback strategies for the rule parsing module. Some specific feasible strategies are shown in Tables 3, 4, and 5.

[0188] Table 3. Feedback on the rule parsing module

[0189]

[0190] Table 4. Feedback on the service center registration module

[0191]

[0192] Table 5. Feedback on the orchestration engine module

[0193]

[0194] The present invention realizes real-time monitoring and dynamic optimization of the system operation status through a module-oriented precise feedback strategy, can promptly discover and correct potential problems, and ensure stable and efficient operation of each module.

[0195] For the rule parsing module, heat map feedback and pre-compilation strategies significantly improve the efficiency of high-frequency rule execution, reduce latency and optimize resource utilization.

[0196] For the service center registration module, dynamic routing weight adjustment and abnormal self-healing mechanism effectively improve the reliability and stability of service discovery and reduce the impact of failures on the system;

[0197] For the orchestration engine module, personalized service link optimization and dynamic scheduling mechanism improve the balance of resource utilization and the robustness of business links, and enhance the system's adaptability in complex and changing environments.

[0198] The above solution builds an efficient closed-loop monitoring and feedback system, effectively improving the system's stability, performance and service quality.

[0199] Based on the results of the execution engine module, the fault tolerance module is used to adjust and optimize the process. Figure 2 As shown. The fault tolerance module receives the real-time push status from the execution engine module, adopts a fast recovery strategy for transient faults through short-term memory fuse, and marks the persistent fault service as "long-term untrustworthy" through long-term memory degradation, which can be resolved by manual intervention. The specific steps include:

[0200] 1) Get the real-time status matrix pushed by the execution engine module, as follows:

[0201] ;

[0202] Where, represents the real-time state matrix, Indicates the Services in the time window The first indicator values, such as success rate, delay, error rate, etc.

[0203] 2) Calculate the weighted failure rate for each service using the following formula:

[0204] ;

[0205] Where, Indicates the The weighted failure rate of each service, represents the time decay weight, represents the time window, Indicates the characteristic function, which takes 1 when it fails and 0 otherwise. Indicates the The service in The state of the moment.

[0206] 3) Establish the first Abnormal fluctuation of services , construct the health evaluation index based on abnormal fluctuation and weighted failure rate, the formula is:

[0207] ;

[0208] Where, Indicates the Health evaluation indicators of services, represents the weighted failure rate The weight coefficient of The weight coefficient representing abnormal volatility, Indicates the weight coefficient of resource health (normalized value of CPU / memory usage).

[0209] 5) The service status is determined based on the health assessment indicator. If the health threshold is lower than the health threshold, the service is instantly disconnected and backup links 1 and 2 are called, with the selection process based on their operating speeds.

[0210] 6) Construct a decaying abnormality integral to determine whether there is a long-term fault. The formula is:

[0211] ;

[0212] Where, Indicates the Services at time The abnormal integral of Indicates the total number of historical abnormal events, represents the time decay coefficient, Indicates the time difference of abnormal events, Indicates the The time when the abnormality occurred, Indicates the The service in Health assessment indicator during the abnormality.

[0213] 7) When the decay-type anomaly integral exceeds the threshold, it is marked as "long-term untrustworthy" and enters the manual intervention process.

[0214] The manual intervention process includes:

[0215] a. Mark the service status as "Degraded";

[0216] b. Route all requests to the backup cluster;

[0217] c. Triggering a three-level alert (email + SMS + work order);

[0218] d. After manual repair, the average health assessment value of 10,000 tests must be greater than 0.98.

[0219] Short-term memory circuit breakers quickly switch to backup links when transient failures occur, ensuring business continuity and minimizing user experience disruptions. Long-term memory degradation mechanisms effectively identify persistent issues, avoiding resource waste caused by frequent circuit breakers and improving overall system efficiency.

[0220] Example 2:

[0221] This application also provides a service orchestration optimization method based on a rule engine, such as Figure 3 Shown, including:

[0222] S1. Perform atomic-level and composite-level analysis on services to obtain atomic services and composite services; detect potential conflicts in composite services;

[0223] S2, receiving and processing natural language input and sketch input to obtain a cross-domain perception matrix;

[0224] S3. Generate a unique hash fingerprint for each atomic service and composite service, including interface signature, performance baseline, dependency library version, etc., to ensure version consistency;

[0225] S4. Build a "business-technology-environment" trinity perception network, including: obtaining a cross-domain perception matrix, optimal path search based on reinforcement learning, and the ability to independently design reward functions based on business needs to achieve domain-adaptive strategies targeting industry pain points;

[0226] S5: Parallel testing: Initiate three different service links simultaneously: primary link + backup link 1 + backup link 2, and take the result that succeeds first.

[0227] S6. Monitor and provide feedback on each step;

[0228] S7. The fault-tolerance module receives real-time status pushes from the execution engine module. It quickly responds and implements appropriate policies when the execution engine module encounters transient and persistent faults. A rapid recovery policy is used for transient faults, while persistent faults are marked as "long-term untrustworthy" and require manual intervention to resolve.

[0229] This paper uses a quantum annealing algorithm to achieve optimal segmentation of service modules, breaking through the local optimality limitations of traditional optimization methods. It also constructs a Hamiltonian model based on service feature vectors to quantitatively assess the coupling between services and their atomization propensity, and integrates quantum tunneling to achieve global optimization. This technology can triple service segmentation efficiency and reduce resource consumption, making it particularly suitable for dynamic adjustments to highly complex microservice architectures.

[0230] The present invention supports dual-channel input of natural language descriptions and flowchart sketches. Through deep parsing technologies such as dependency syntax analysis and semantic role labeling, it converts business requirements into structured rules. It automatically repairs broken lines (error <10 pixels), identifies overlapping elements (spacing >5 pixels without misjudgment), and generates an executable service chain within 1 second, greatly lowering the threshold for non-technical personnel to participate in system design.

[0231] This invention constructs a three-dimensional perception network of "business-technology-environment" and drives dynamic optimization of service paths through independently designed industry pain point reward functions such as financial anti-fraud rate and industrial equipment OEE; the algorithm supports thousands of concurrent decisions per second, improving key indicators by 15%-38% in complex business scenarios, achieving precise alignment of business needs and technical execution.

[0232] This invention utilizes a parallel racing mechanism between the primary link and dual backup links, along with physical isolation technology using CPU core-level binding, to ensure strict parallel operation of the three service chains without resource contention. Combining a dual strategy of short-term circuit breaking and long-term degradation (when the anomaly score exceeds a threshold), system availability reaches 99.999%, with failover latency controlled within 200 microseconds, perfectly supporting high-sensitivity scenarios such as financial transactions and industrial control.

[0233] This invention establishes a module-level precision feedback network. Through technologies such as rule heat map precompilation, service DNA fingerprint tracking, and dynamic routing weight adjustment, it builds a comprehensive self-healing capability across the entire "monitoring-analysis-repair" chain. When the service failure rate exceeds 5%, rolling updates are automatically triggered, reducing recovery time by 82%. High-frequency business rules can be compiled into machine code in real time, increasing execution efficiency by sixfold and significantly reducing system operation and maintenance complexity.

[0234] Example 3:

[0235] Building on existing fault tolerance mechanisms, we can also explore the introduction of a federated learning framework to enable the co-evolution of health assessment models across business domains. By encrypting and exchanging abnormal volatility and weighted failure rate data across services, we can build a globally shared knowledge base of fault patterns, enabling the system to proactively identify new fault signatures. For example, when a certain type of database connection timeout occurs for the first time in a financial business, this pattern can be quickly synchronized to the e-commerce business fault tolerance module, enabling preemptive cross-domain defense.

[0236] Another approach is to consider applying the quantum random walk algorithm to the dynamic weight adjustment of decay-type anomaly integrals. By using quantum superposition states to concurrently explore the space of different combinations of decay coefficients and health weights, the algorithm seeks the parameter configuration that best fits the integral model to historical faults. Compared to traditional grid search, quantum optimization can improve parameter tuning efficiency by over 10 times, making it particularly suitable for real-time fault-tolerant decision-making in ultra-large-scale service clusters.

[0237] Although the specific embodiments of the present invention are described above, those skilled in the art should understand that the specific embodiments described are merely illustrative and are not intended to limit the scope of the present invention. Any modifications and changes made by those skilled in the art in accordance with the spirit of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A service orchestration system based on a rule engine, characterized in that: Includes the following modules, Rule parsing module: including atomic-level parsing unit, combination-level parsing unit, rule prediction unit and multimodal rule input unit; The atomic parsing unit is used to disassemble services and parse them into atomic services with functional characteristics; The combination-level parsing unit is used to record the number of adjacent occurrences of service pairs and judge the relevance of services. When there are continuous high-frequency adjacent pairs, they are merged into a combined service. The rule prediction unit is used to detect whether there is a conflict in the combined service generated by the combination-level parsing unit; The multimodal rule input unit is used to receive natural language input and sketch input, and perform preprocessing to obtain a cross-domain perception matrix; Receiving and processing natural language input and sketch input to obtain a cross-domain perception matrix includes the following steps: Processing of natural language input, including text cleaning and standardization, word segmentation and part-of-speech tagging, domain entity recognition, dependency parsing, semantic role labeling, logical predicate extraction, and intermediate representation generation; Processing sketch input, including image preprocessing, graphic element detection, text recognition, topology relationship reconstruction, semantic enhancement and conflict resolution, and intermediate representation generation; Integrate the processing results of natural language input and sketch input to obtain a cross-domain perception matrix; Service registration center module: including service DNA fingerprint unit and self-healing service unit; The service DNA fingerprint unit is used to generate a unique hash fingerprint for each atomic service and composite service; Self-healing service unit: used to automatically mark and remove instances of an atomic service or composite service from the available pool when the failure rate exceeds a threshold; Orchestration engine module: used to obtain the cross-domain perception matrix and adaptively adjust the optimization strategy through the reward function; Execution engine module: This module selects multiple optimal service links based on the results generated by the orchestration engine module, forming a multi-link parallel racing execution technology solution. Monitoring feedback module: used to implement feedback strategies for the rule parsing module, service center registration module, and orchestration engine module; Fault-tolerant module: used to receive real-time push status from the execution engine module, and quickly respond and adopt corresponding strategies when the execution engine module faces transient and persistent faults.

2. A service orchestration optimization method based on a rule engine, characterized in that: The service orchestration system based on a rule engine as claimed in claim 1 is adopted, and the optimization method includes the following steps: S1. Perform atomic-level and composite-level analysis on services to obtain atomic services and composite services, and detect potential conflicts in composite services. S2, receiving and processing natural language input and sketch input to obtain a cross-domain perception matrix; S3. Generate a unique hash fingerprint for each atomic service and composite service; S4. Obtain a cross-domain perception matrix, search for the optimal path based on reinforcement learning, and independently design a reward function based on the business; S5: Parallel testing: Initiate multiple service links simultaneously and take the first successful result. S6. Monitor and provide feedback on each step; S7. Receive the real-time push status of the execution engine module, and quickly respond and adopt corresponding strategies when the execution engine module faces transient and persistent failures.

3. A service orchestration optimization method based on rule engine according to claim 2, characterized in that: In S1, services are decomposed into atomic services using atomic parsing units, which have the following functional characteristics: functional singularity, resource independence, idempotence, and temporal and spatial determinism. Atomic parsing of a service involves the following steps: 1) Model the service characteristics and use the characteristic vector to describe the service attributes. The formula is: ； Where, The feature vector representing the service, Indicates the CPU consumption of the unit call, Indicates the memory usage. Indicates the delay reference value, Indicates the number of explicitly dependent services, represents the cascading failure coefficient; 2) Construct Hamiltonian to define the optimal segmentation, the formula is: ； Where, represents the Hamiltonian, Indicates the coupling degree between services, Indicates the service atomization tendency parameter, , Indicates the split decision variable, 0 means retaining atomicity, 1 means splitting is required; is a preset weight coefficient; The degree of coupling between services Determined by the eigenvector difference, the formula is: ； Where, represents the natural base, represents the attenuation coefficient, Indicates the vectors Hedi vectors The eigenvector difference of Indicates the call timing angle; The service atomization tendency parameter Determined by resource consumption characteristics, the formula is: ； Where, Indicates the CPU consumption of the unit call The weight coefficient of Indicates memory usage The weight coefficient of Indicates the delay baseline value The weight coefficient of .

4. A service orchestration optimization method based on rule engine according to claim 3, characterized in that: The segmentation is performed using the quantum annealing algorithm, which includes the following steps: 1) Convert the Hamiltonian parameters into the physical configuration of the quantum processor, map each service to a physical qubit, set the inter-qubit coupler according to the Hamiltonian, and apply a local magnetic field to each qubit; 2) Setting the time evolution path of quantum annealing to improve the probability of the optimal solution; 3) Quantum annealing is performed to initialize the quantum state, and all quantum bits are initialized to a superposition state; 4) Screen and optimize the quantum annealing output results, repair broken solutions caused by insufficient chain strength, and sort candidate solutions in ascending order of Hamiltonian energy; 5) Determine the final service cutting plan and verify its reliability, and select the candidate solution with the lowest energy and meeting the constraints.

5. The service orchestration optimization method based on rule engine according to claim 2 is characterized in that: In S1, composition-level parsing of services includes the following steps: 1) Construct a service co-occurrence matrix to record the number of adjacent occurrences of service pairs; 2) An adaptive dynamic threshold is constructed based on the mean and variance. If the number of adjacent occurrences is greater than the adaptive dynamic threshold, it is considered a significant association. Among them, when there are continuous high-frequency adjacent pairs, they are merged into a combination, and partially overlapping sequences are merged.

6. A service orchestration optimization method based on rule engine according to claim 2, characterized in that: In S1, potential conflicts in the composite service include: Resource lock, judged by the length of the circular wait chain; Data race, read after write or write after read of the same data item; Permission exceeded: the first service calls the second service but does not have permission; Temporal paradoxes, where the order of events violates the laws of causality; Resource overload, where the cumulative resource demand is greater than the node capacity; Business logic conflicts and mutually exclusive rules take effect at the same time.

7. A service orchestration optimization method based on rule engine according to claim 2, characterized in that: In S4, the optimal path search based on the reinforcement learning algorithm includes constructing the state space, action space, establishing functions and network structures; By obtaining the cross-domain perception matrix and processing it to obtain the state space, the formula is: ； Where, represents the state space, means flattening the three-dimensional matrix into a vector, represents vector concatenation, Represents LSTM historical state encoding; The action space uses discrete actions and continuous parameter adjustments to implement three types of operations: node addition and deletion, path weight adjustment, and fault switching. The total reward function is calculated as: ; Where, represents the total reward function, represents the basic reward function, represents the domain reward function; The formula of the basic reward function is: ; Where, Indicates the service level compliance rate The weight coefficient of represents the weight coefficient of resource cost, Indicates the risk event count The weight coefficient of The formula of the domain reward function is: ; Where, represents the domain reinforcement coefficient, Indicates the number of industry pain point functions, Indicates the Industry pain point function The weight coefficient of The network structure of the reinforcement learning algorithm adopts the Actor-Critic framework.

8. The service orchestration optimization method based on rule engine according to claim 2 is characterized in that: In S6, the monitoring feedback module feeds back data to the rule parsing module as a heat map of the rule execution path and precompiles high-frequency rules into machine code; Dynamically adjust service routing weights based on feedback data from the service center registration module; Optimize personalized service links based on feedback from the orchestration engine module; By establishing monitoring indicators, threshold ranges and feedback strategies, the system operation status can be monitored and dynamically optimized in real time.

9. The service orchestration optimization method based on rule engine according to claim 2 is characterized in that: In S7, the fault-tolerant module receives real-time status pushes from the execution engine module. It uses short-term memory to implement a rapid recovery strategy for transient faults and uses long-term memory to degrade services. It marks persistent faults as "long-term untrustworthy" and requires manual intervention to resolve the issue. The specific steps include: 1) Get the real-time status matrix pushed by the execution engine module, as follows: ； Where, represents the real-time state matrix, Indicates the Services in the time window The first indicator values, including success rate, delay, and error rate; 2) Calculate the weighted failure rate for each service using the following formula: ； Where, Indicates the The weighted failure rate of each service, represents the time decay weight, represents the time window, Indicates the characteristic function, which takes 1 when it fails and 0 otherwise. Indicates the The service in The state of the moment; 3) Establish the first Abnormal fluctuation of services , construct the health evaluation index based on abnormal fluctuation and weighted failure rate, the formula is: ； Where, Indicates the Health evaluation indicators of services, represents the weighted failure rate The weight coefficient of The weight coefficient representing abnormal volatility, The weight coefficient indicating the health of the resource; 4) Determine the service status based on the health assessment indicator. If it is less than the health threshold, the circuit breaker is instantaneously disconnected and a backup link is called, which is selected based on its operating speed. 5) Construct a decaying abnormality integral to determine whether there is a long-term fault. The formula is: ； in, Indicates the Services at time The abnormal integral of Indicates the total number of historical abnormal events, represents the time decay coefficient, Indicates the time difference of abnormal events, Indicates the The time when the abnormality occurred, Indicates the The service in Health assessment indicators at the time of abnormality; 6) When the decay anomaly score exceeds the threshold, it is marked as "long-term untrustworthy" and enters the manual intervention process.

Images