A micro-service component combination construction method based on business logic and label classification

Abstract

Aiming at the multi-task complex dynamic and multi-network environment characteristics of current industrial software system, this patent proposes a micro-service component combination construction method based on business logic and label classification to improve the task processing capacity and collaborative service efficiency of components. According to different time sequences of industrial tasks, the method extracts components related to business logic in each collaborative network and constructs a component logic relationship network graph. Then, the importance of components is calculated based on their inherent functional attributes to accurately identify core components and pre-process the labels of their adjacent components. Then, the labels of components are updated, the set of lower-level call relationships of core components is updated, and the labels are propagated and diffused to other components in the micro-service. Finally, a component combination mode associated with business logic is formed to realize identifiable component autonomous collaboration.

Description

Technical Field

[0001] This invention relates to multi-networked industrial software systems. Addressing the complex and dynamic characteristics of multi-task systems, it supports the assembly and scheduling technology of intelligent components within a microservice framework. This improves the efficiency of multi-component collaboration while reducing task blocking and component overload, adapting to diverse industrial network scenarios. Specifically, it relates to a method for constructing microservice components based on business logic and tag classification. This method distinguishes relevant business components in different networks based on business logic and forms a task-oriented software composition pattern through component tag classification and updating, thereby realizing one or more identifiable autonomous business processes. It achieves autonomous collaboration of components while ensuring high component utilization, saving component consideration time and reducing the possibility of task suspension. Background Technology

[0002] In industrial software systems employing intelligent components, the system comprises several autonomous intelligent components that collaborate to complete industrial tasks. Current industrial software systems are typically based on a single network architecture, where interaction and collaboration between intelligent components often occur through a single network relationship. This traditional single-network architecture model leads to significant limitations such as information silos and inefficient software collaboration when facing multi-networked industrial environments. With the prevalence of multi-network scenarios and the increasing complexity of industrial task requirements, the scheduling and collaboration of intelligent components in industrial software systems are affected by multiple network coupling relationships, such as constraints from multiple networks including information flow, control flow, and business flow. Existing industrial software system models only consider single network links between industrial software components, resulting in a lack of collaboration between different industrial software systems on different networks, low task scheduling efficiency, and poor collaborative optimization. Furthermore, the component scheduling and assembly model has a single, fixed granularity, making it difficult to quickly and effectively address the multi-granularity service requirements of multi-networked industrial environments, thus failing to effectively guarantee the efficiency, robustness, and scalability of software systems in multi-networked industrial environments.

[0003] To address the varying task complexity and component functional requirements across multiple industrial network environments, a component composition and construction technique needs to be developed. Based on task-driven or business logic, this technique identifies isomorphic mappings between components from different networks and extracts business logic-related components within a microservice architecture. Through tagging and updating core and adjacent components, a component composition pattern relating to business logic is ultimately formed. This enables the efficient construction of dynamic mapping relationships between task information and component attributes, providing strong technical support for subsequent multi-task execution and component composition pattern switching. This construction technique can also encapsulate a group of collaborative business components into system-level business components, enabling one or more identifiable autonomous business processes. Summary of the Invention

[0004] Technical Problem: The purpose of this invention is to propose a microservice component composition and construction method based on business logic and label classification, utilizing community partitioning techniques from complex networks to construct component composition patterns. Specifically, this method borrows the classic label propagation algorithm from community partitioning techniques to classify components into composition patterns. The method mainly consists of three stages. The first stage is network graph construction, extracting components relevant to the current business logic and constructing a network graph based on the logical relationships between components. The second stage is the component composition pattern preprocessing stage, calculating the importance of components based on their inherent attributes, accurately identifying core components, and preprocessing the labels of adjacent components of the core components. The third stage involves updating component labels and optimizing component composition patterns to make them more reasonable and efficient. Label propagation involves updating the set of lower-level call relationships of the core component. By traversing adjacent components, labels are propagated to all components in the microservice, ultimately forming a business logic-related composition pattern. This method effectively shortens the component identification time and provides a combination pattern reference for subsequent component scheduling. When scheduling components, components in the same combination pattern can be given priority, saving component consideration time. It provides a new approach to solving problems such as the single component scheduling architecture and low component collaboration efficiency in software systems.

[0005] Technical Solution: Intelligent collaboration among components in industrial software systems plays a crucial role in the effective completion of industrial tasks. For large-scale and complex industrial tasks, a single component often fails to complete the entire task due to insufficient functionality or excessive time consumption. Therefore, component collaboration technology is needed to improve task completion efficiency. However, current component collaboration technologies have limited collaborative relationships and cannot adapt to the deployment of intelligent components in coupled scenarios such as multiple workshops, resulting in high collaboration costs or even task failure. Therefore, to solve the technical problems of intelligent component collaboration decision-making and improve component task processing capabilities, this patent proposes a microservice component composition and construction method based on business logic and tag classification. This method implements industrial software component identification and collaboration technology based on a microservice architecture. Through this method, components in the network are divided into multiple combination modes to adapt to the problems of green efficiency improvement, elastic service, and intelligent component deployment in complex coupled scenarios such as multiple products, multiple batches, and multiple workshops. To address the different task complexities and required component functionalities, similar components are integrated to form multiple component combination modes. The technical solution of the microservice component composition and construction method proposed in this patent is as follows:

[0006] (1) Network Graph Initialization. First, a network context model is constructed to provide contextual support for the entire component architecture. Based on business concepts, isomorphic mappings of components in the software design document are identified. A component identification and extraction strategy based on a multi-agent system is used to understand the keywords and semantics in the software design document, and to identify and label business components of different types and functions. Starting from the software design document, multiple agents work together to complete the identification and extraction of components, and the constraints of the component context are fully considered to distinguish the business logic functions of each component. Components related to the current business logic are extracted from the component document, the deployment relationships of the components are examined, the relevant information of the software components is quantified, and the component logical architecture is constructed.

[0007] (2) Component Combination Pattern Preprocessing Stage. This stage, based on the network model constructed using component attributes and scheduling, further studies component scheduling relationships to achieve efficient collaboration between components. In the method designed in this patent, the idea of ​​label propagation algorithm in complex networks is used to divide components into combination patterns. Utilizing label propagation between components, components with the same label are ultimately grouped into the same community. Since the identification of the target component and the priority of scheduling between components affect the time and energy consumption of completing a specified task, and considering the varying efficiency and coordination capabilities of components, this patent adopts a strategy of using the core component in the network to preprocess the labels of neighboring components, reducing the initial number of label types. This stage first uses the PageRank algorithm to find the core component at the center of the network. By preprocessing the labels of adjacent components of the core component, components with high correlation can have the same label before label propagation. After preprocessing, some highly similar components in the network will have the same label, reducing the randomness of component combination patterns and facilitating the construction of component combination patterns.

[0008] (3) Component Call Relationship Set Update Phase. This phase mainly updates the component labels. The label update strategy is to update the lower-level call relationship set of an important component. To make the composition pattern more reasonable and efficient, and to enable better jump scheduling capabilities between components, the component labels are further adjusted to optimize the component composition pattern. After preprocessing the component composition pattern, components with high call frequency and high business logic relevance are assigned the same label. However, the division of the composition pattern is not tight enough at this point, and it is necessary to continuously iterate to spread the labels to components throughout the microservice. By traversing all components in the network, the component labels are updated according to the component's own attributes and the business logic relevance between components, making the component composition pattern more stable. The components are divided into communities according to the business logic relevance and the top-down call relationship. After the division, a component composition pattern for this business logic is formed.

[0009] Beneficial effects:

[0010] (1) Accurately identify and extract technical components, and construct a network graph using graph theory based on the inherent attributes of the components and the scheduling relationships between them. Use the PageRank algorithm to calculate the importance of components and accurately identify core components.

[0011] (2) To reduce component consideration time and improve component utilization, all components were divided into composition patterns based on their own attributes and functions, as well as the degree of correlation between components. The resulting multiple composition patterns can be applied to the multi-task service requirements brought about by different networked environments. The composition pattern is driven by business or task, and the core component that is suitable for the business is selected. Then, the core component schedules other components to complete the task. During the component scheduling process, the core component prioritizes scheduling other components in the same composition pattern, reducing the consideration time for component scheduling.

[0012] (3) Improve component elastic scheduling and collaborative optimization. By calculating the correlation between components, the degree of correlation between components can be digitized, accurately measuring the logical relationship between components, and classifying components into different combination modes. By integrating the scheduling and being scheduled relationships between components and the priority of component scheduling, the preprocessed component combination mode is optimized to improve the collaborative efficiency between components and meet the adaptive switching of service composition mode. Attached Figure Description

[0013] Figure 1 Example diagram of the microservice component composition pattern construction process

[0014] Figure 2 Main principle diagram of the method of this invention Detailed implementation method:

[0015] Building upon the information sharing model, we further investigated microservice component composition patterns to achieve efficient collaboration between components. We abstract microservice components into a network graph based on their own attributes and inter-component scheduling. Components correspond to nodes in the network graph, scheduling relationships between components correspond to edges, and the priority of component scheduling corresponds to the weight of the edges. For a specific business, we extract components related to the business logic, identify core components, and partition the components according to the relevance of the business logic and the top-down calling relationships, ultimately forming a component composition pattern for this business. This composition pattern is driven by the business logic or task. This patent extracts business logic-related components from microservice components to construct a network graph, uses community detection technology in complex networks to partition the network graph, and finally forms a component composition pattern for this business logic. The main innovation of this patent lies in using the idea of ​​complex network community detection to study industrial component composition patterns, comprehensively considering the component retrieval and coordination capabilities based on network topology characteristics, identifying core components, and forming component composition patterns based on the top-down calling relationships of components.

[0016] The specific implementation steps are as follows:

[0017] (1) Network Graph Initialization. Extract components relevant to the current business logic, examine the deployment relationships of components, quantify the relevant information of software components, and construct the component logical architecture. Based on business concepts, identify isomorphic mappings of components in the software design document. Utilize a component identification and extraction strategy based on a multi-agent system to understand the keywords and semantics in the software design document, identify and label business components of different types and functions. Starting from the software design document, fully consider the constraints of the component context and distinguish the business logic functions of each component. Abstract components into nodes in the network graph, and abstract the scheduling and being scheduled process between components into edges in the network graph. The edges start from the scheduling component and point to the scheduled component, forming directed edges in the network graph. The priority of component scheduling is abstracted into the weight of the edges, with a larger weight indicating a higher priority.

[0018] A network graph consists of components V, edges E connecting the components, and edge weights W. A directed weighted network is represented by G = (V, E, W), where V = {v1, v2, v3, ..., v...}. n} represents the set of network components, E = {e1, e2, e3, ..., e m} represents the set of edges in the network. This represents the weight of an edge, where n is the total number of network components and m is the number of edges in the network. By quantifying the scheduling relationships and priorities between components, directed edges and their weights between components with scheduling relationships are obtained. <v1,v2,w 1,2 > indicates that component v1 schedules component v2, w 1,2 The weight of this edge represents the priority of this scheduling. Components that schedule each other are neighboring components. A component can call multiple components and can also be scheduled by multiple components. We will define the scheduling component v. i The component is called an ingress neighbor, using IN. i This indicates, i.e., in-neighbors; the component v i The component that is scheduled is called the outgoing neighbor, and is represented by ON. i This refers to out-neighbors. Edges formed when a component schedules other components are called outgoing edges, and edges formed when a component is scheduled by other components are called incoming edges. The total number of outgoing edges is called the component's v. i The degree of exit, using The total number of incoming edges is called component v. i in-degree, using The total number of incoming and outgoing edges is called the degree of the component, denoted by k. i express.

[0019] (2) Component Combination Pattern Preprocessing Stage. In the method designed in this patent, the idea of ​​label propagation algorithm in complex networks is used to divide components into combination patterns. The basic idea of ​​the label propagation algorithm is to use the label information of neighboring nodes to predict the label information of the node to be updated, that is, to utilize label propagation between components to ultimately group components with the same label into the same community. Since the identification of the target component and the priority of scheduling between components both affect the time and energy consumption of completing a specified task, and considering that the efficiency and coordination capabilities of components vary, this patent adopts a strategy of using the core components in the network to preprocess the labels of neighboring components, thereby reducing the initial label type scale.

[0020] First, the PageRank algorithm is used to calculate the priority of components, that is, the order among components related to the current business logic. The higher the PageRank value of a component, the higher its status in the composition pattern, and the higher it is in the scheduling sequence. The idea of ​​the PageRank algorithm is applied to the network graph of microservice components. The basic idea is that if a component is scheduled by multiple components, it indicates that this component is relatively important, and its PageRank value will be relatively high; if a component with a high PageRank value schedules other components, the PageRank value of the scheduled components will increase accordingly. This patent uses the value calculated by the PageRank algorithm to identify core components, and the calculation formula is as follows:

[0021]

[0022] Among them, PR(v i ) represents node v i PageRank value; L(v i ) represents node v i The output degree is denoted by ; n represents the total number of nodes in the network; d is a coefficient called the damping factor, typically taken as d = 0.85. The PageRank algorithm depends on the network topology. Once the network topology is determined, the PageRank value is also determined, and the PageRank value of a component in the network represents its importance or priority. Initially, all components are considered as components to be processed. For ease of statistics, we set V' to represent the set of components to be processed. After calculating the PageRank values ​​of all components in the network, the results are placed in a PR list. Then, the component with the highest PageRank value is selected from the PR list as the core component, denoted as v. core The calculation formulas for the core components are as follows:

[0023] v core =argmax{PR(v i )|v i ∈V',PR(v i )∈PR}#(2)

[0024] Where V' represents the current set of components to be processed; PR(v i ) indicates component v iThe PageRank (PR) value represents the set of PR values ​​for all components in the network. After identifying the core components from all components to be processed, they are assigned labels. In the first round of processing, the core components selected in each round are those with the highest PR values ​​in the set of components to be processed in that round. Their importance is greater than that of the other components, and the core components selected in the previous round have a higher priority than those selected in the next round. Therefore, we assign labels to the core components in ascending order of their numerical values. The first core component is labeled core_1, the second core component is labeled core_2, and so on.

[0025] After selecting the core component, we identify the components directly related to it, i.e., those with mutual scheduling relationships; these are called neighboring components. Based on business logic relevance, we calculate the tightness between the neighboring components directly scheduled by the core component and the core component, i.e., the degree of association between the components and the core component. Since this patent uses the idea of ​​tag propagation to divide components into composition patterns, and tags are propagated through scheduling relationships between adjacent components, the degree of association calculation only exists between the core component and the components it schedules. The formula for calculating the degree of association of components is as follows:

[0026]

[0027] Among them, w core,i Indicates core component v core With component v i The weights of the directed edges formed, i.e., the weights of the core component v core Scheduling component v i Priority; IN i Indicates component v i The incoming neighbor, i.e., the scheduling component v i Components; N i Indicates component v i The set of neighboring components; |N i | indicates component v i The number of neighboring components, i.e., node v i The degree of z; z represents the component v i With core component v cent Common neighbor components; k z The degree of the common neighbor component z is represented by α; α is a coefficient, which is generally taken as 0.4 to 0.6.

[0028] A threshold δ is set; this threshold δ is an important parameter for determining the degree of association between components and affecting the propagation of tags between components. When Close(v i ,v core When )>δ, it indicates that component v i With core component v coreIf the components have strong interrelationships and close connections, then the core component v... core The tag is given to the component v i If Close(v i ,v core If )≤δ, then it means that component v i With core component v core The correlation between them is weak, and no label propagation occurs.

[0029] During component tag preprocessing, to avoid assigning redundant tags to components and performing unnecessary processing, a residual affinity R_close is set. The residual affinity R_close controls the number of tags a component can have and when it should be removed from component processing. Only when component v... i Close(v) i ,v core )>δ, give component v i After assigning labels to the core components, the component v is then calculated. i The remaining correlation degree, if component v i The affinity with the core component does not satisfy Close(v) i ,v core If )>δ, then component v is not calculated. i The remaining correlation. When Close(v i ,v core When the value is large, it indicates that component v i With core component v core The business logic is highly correlated, and the frequency of calls between components is high. Components can be assigned v i With core component v core The same label places them in the same composition pattern. However, due to component v i There is also the possibility that it could be scheduled by other components, component v i It can also be combined with other components, so the residual association degree of the components is set. The initial value of the residual association degree of each component is set to 1. Component v i Remaining correlation R_close i The calculation formula is as follows:

[0030] R_close i =R_close i -Close(v i ,v core )#(4)

[0031] Set a threshold ε when component v i Close(v) i ,v core )>δ and R_closei When the value is less than ε, it indicates a strong correlation between this component and the core component, suggesting a high probability that they are grouped into the same composition pattern. It also indicates that the core component is calling component v. i Component v has a higher priority and will no longer be considered. i The possibility of being scheduled by other components and then combined with other components, will make component v i Remove component v from the set of components to be processed, V'. i Tag preprocessing is complete. When component v i R_close i When ≥ε, it indicates that component v i There is also the possibility of combining it with other core components, in which case component v will be... i Retained in the collection of call processing components.

[0032] Once the core component selected in this round has completed the tag preprocessing of all adjacent components, it is removed from the component set V' to be processed, and the next round of component preprocessing begins. The component with the highest PR value is selected from the current component set V' as the core component, and the steps of the previous round of component tag preprocessing are repeated until the component set V' is empty, at which point the component composition pattern preprocessing is complete. After preprocessing, all components have at least one tag, and the component composition pattern has been initially formed.

[0033] (3) Component Call Relationship Set Update Phase. After preprocessing the component composition pattern, components with high call frequency and strong business logic relevance are assigned the same label. However, the division of the composition pattern is not tight enough at this point, and iterative updates are needed to allow the labels to spread to components throughout the microservice. To make the composition pattern more reasonable and efficient, and to enable better jump scheduling capabilities between components, this patent also designs a label update strategy to further adjust the labels of components and optimize the component composition pattern. The label update strategy is the update of the lower-level call relationship set of an important component. By traversing all components in the network and reprocessing the component labels, the component composition pattern becomes more stable.

[0034] To ensure that the tags of core components are not affected by components with lower priority, component tags are updated in ascending order of PageRank value. After tag preprocessing, some components may have multiple tags, meaning they may be scheduled by multiple core components. To shorten the component consideration time, a formula for calculating component tag propagation ability is defined. The calculation of component tag propagation ability is used to filter the tags of a component, retaining tags with strong propagation ability among its neighboring components. The formula for calculating component tag propagation ability is as follows:

[0035]

[0036] Among them, P i (l) indicates component v i The probability that the label is updated to label l is the same as the propagation capability of the label; IN i Indicates component v i The set of neighboring nodes, i.e., the scheduling component v i Components; ON i Indicates component v i The set of outgoing neighbor nodes, i.e., the node set of the component v i Scheduling components; lb i Indicates component v i The collection of all tags owned by [the entity / entity]. Indicates component v i The degree of departure; w u,i Represents a directed edge <v u ,v i The weight of >, i.e., component v u Scheduling component v i Priority; w max This represents the highest weight in the network, i.e., the highest priority scheduling among all component scheduling. In the formula for calculating component label propagation capability, label l originates from component v. i The adjacent components have labels, component v i The label can be updated to the label of the component that is scheduling it. The formula is used to represent the importance of components, and the weights of the edges are normalized. The numerator in the formula represents the weights of all outgoing edges of component u, that is, the sum of the priorities of component u in scheduling other components. The larger the weight, the higher the priority. The denominator is the maximum weight multiplied by the outgoing degree of the component. The formula means that component u is related to component v. i The priority of scheduling is determined by the scheduling component v. i The priority percentage of all components. Using the formula... Measuring the performance of a neighboring component in component v i The influence of the component tag propagation capability among all neighboring components. The formula for calculating the propagation capability of a component tag is the scheduling component v. i The propagation capability of the tags carried by all components is closely related to the components carrying those tags. The greater the importance of a component, the stronger the propagation capability of its tags.

[0037] After calculating the propagation capability of component tags, a threshold value γ is set to 0.5 (the value of γ can be adjusted according to actual conditions). γ is used to control the selection of component tags. The set LB is used... i This indicates that for component v i The set of labels for all components that are directly scheduled. For l∈LBi If P i (l)>γ, then component v i The label of P is updated to label l; if P i If (l) ≤ γ, then no label update is performed. Using an iterative approach for label updates ensures the stability of the component partitioning results. During the component label update phase, the algorithm terminates when the labels of all components no longer change after multiple iterations, and components with the same label are grouped into a single composition pattern. Ultimately, the network will be partitioned into various microservice component composition patterns to adapt to elastic scheduling and intelligent component deployment in complex, coupled scenarios such as multi-tasking.

Claims

1. A method for constructing microservice components based on business logic and tag classification, which uses community partitioning technology of complex networks to construct component composition patterns; this method provides a composition pattern reference for subsequent component scheduling, and prioritizes components in the same composition pattern during component scheduling; The network graph is constructed as follows: by extracting components related to the current business logic, examining the deployment relationships of components, quantifying the relevant information of software components, constructing the component logical architecture, identifying isomorphic mappings of components in the software design document based on business concepts, using a component identification and extraction strategy based on a multi-agent system to understand the keywords and semantics in the software design document, identifying and marking business components of different types and functions, and fully considering the constraints of the component context, distinguishing the business logic functions of each component, abstracting the components into nodes in the network graph, abstracting the scheduling and being scheduled process between components into edges in the network graph, with the edges starting from the scheduling component and pointing to the scheduled component, forming directed edges in the network graph, abstracting the priority of component scheduling into the weight of the edges, and abstracting the inherent attributes of components located in the microservice framework into a network graph with a network topology structure by quantifying the inherent attributes of the components. This paper employs the label propagation algorithm in complex networks to partition components into compositional patterns. By utilizing label propagation between components, components with the same label are grouped into the same community. A strategy is adopted where the core component preprocesses the labels of its neighbors. Once the core component has completed preprocessing the labels of all its neighbors in a given round, it is removed from the set of components to be processed. Remove from the current set of components and proceed to the next round of component preprocessing. Select the component with the highest PageRank value as the core component, and repeat the component tag preprocessing steps from the previous round until the set of components to be processed is reached. If empty, the component composition pattern preprocessing is complete. After preprocessing, all components have at least one label, and the basic form of the component composition pattern has been formed. The component call relationship set is updated. After preprocessing the component composition pattern, components with high call frequency and high business logic relevance are assigned the same label. By traversing all components in the network, the component labels are further processed to make the component composition pattern more stable. In the component label update phase, when the component labels have been iterated multiple times until the labels of all components no longer change, the algorithm ends and components with the same label are divided into a composition pattern. Finally, the network will be divided into multiple microservice component composition patterns. The components are tagged in ascending order of PageRank value. After tag preprocessing, some components may have multiple tags, meaning they may be scheduled by multiple core components. A formula for calculating component tag propagation ability is defined. This calculation is used to filter the tags of a component, retaining the tags with stronger propagation ability among its neighboring components. The formula for calculating component tag propagation ability is as follows: in, Representation Component The tag is updated to the tag The probability represents the extent to which the tag can spread. Representation Component The set of incoming neighbor nodes; Representation Component The set of out-neighbor nodes; Representation Component The collection of all tags it possesses; Representation Component The degree of departure; Represents a directed edge The weights; Represents a directed edge The weights; This represents the largest weight in the network; Represents a directed edge The weights; After calculating the propagation capability of component tags, a threshold is set. The value is 0.

5. Used to control the selection of component tags, using a collection Indicates the component The set of tags for all components that are directly scheduled. ,like Then the component The tag is updated to the tag ;like If not, then no tag update will be performed.

2. The method for constructing microservice components based on business logic and tag classification according to claim 1, characterized in that: First, the PageRank algorithm is used to calculate the importance of components. The higher the PageRank value of a component, the higher its status in the composition pattern and the higher it is in the scheduling sequence. This patent uses the PageRank algorithm to calculate the importance of components, and the calculation formula is as follows: in, Represents a node PageRank value; Represents a node The degree of departure; Indicates the total number of nodes in the network; It is a coefficient, called the damping factor.

3. The method for constructing microservice components based on business logic and tag classification according to claim 1, characterized in that: Network diagrams consist of components Connections between components and the weight of the connected edges Composition, using Let represent a directed weighted network, where Represents a collection of network components. Represents the set of edges in the network. This represents the weight of the connected edge, where The total number of network components. To determine the number of network edges, we quantify the scheduling relationships and priorities between components, obtaining the directed edges and their weights between components with scheduling relationships. Representation Component For components Perform scheduling. The weight of the edge indicates the priority of this scheduling. Components that are scheduled to each other are neighboring components. A component can call multiple components and can also be scheduled by multiple components. The scheduling component will be... The component is called an ingress neighbor, using Indicates; the component The scheduling component is called the outgoing neighbor, and is used... The term "outgoing edge" refers to the connection formed by a component scheduling other components, while the connection formed by a component being scheduled by other components is called an incoming edge. The total number of outgoing edges is called the component's total number of incoming edges. The degree of exit, using The total number of incoming edges is called the component's total number of edges. in-degree, using The total number of incoming and outgoing edges is called the degree of the component. express.

4. The method for constructing microservice components based on business logic and tag classification according to claim 1, characterized in that: After calculating the PageRank value of all components in the network, the results are put into... From the list, then from The component with the highest PageRank value in the list is selected as the core component, denoted as . The calculation formulas for the core components are as follows: in, This represents the set of components currently awaiting processing. Representation Component PageRank value, This represents the set of PageRank values ​​corresponding to all components in the network. After identifying the core components from all components to be processed, a label is assigned to the core components. In the first round of processing, the core components selected in each round are those from the set of components to be processed in that round. The component with the highest value is more important than the other components, and the core components selected in the previous round have a higher priority than the core components selected in the next round. Therefore, the core components are labeled in ascending order of their numerical values.

5. The method for constructing microservice components based on business logic and tag classification according to claim 1, characterized in that: Based on business logic relevance, the tightness between the core component and its directly scheduled neighboring components is calculated, i.e., the degree of association between the components and the core component. Since this patent uses the idea of ​​tag propagation to divide the components into composition patterns, and the tags are propagated through the scheduling relationship between adjacent components, the degree of association calculation only exists between the core component and the components it schedules. The formula for calculating the degree of association of components is as follows: in, Indicates core components With components The weights of the directed edges formed, i.e., the core components Scheduling component Priority; Representation Component The collection of neighboring components; Representation Component The number of neighboring components, i.e., the number of nodes. The degree; Representation Component With core components Common neighbor components; Represents common neighbor components The degree; This is the correlation coefficient; Represents a directed edge The weights; core components The collection of neighboring components.

6. The method for constructing microservice components based on business logic and tag classification according to claim 5, characterized in that: Rules for component tag preprocessing, setting thresholds threshold It is an important parameter for determining the degree of association between components and affecting the propagation of tags between components. When, describe the component With core components If the components are highly interconnected and collaborative, and the connection between two components is close, then the core component... The tag is assigned to the component ,like This indicates that the component With core components The correlation between them is weak, and no label propagation occurs.

7. The method for constructing microservice components based on business logic and tag classification according to claim 6, characterized in that: Remaining correlation Used to control the number of tags a component can have and when to remove them from the call handling component, only when the component... of Give components After assigning labels to the core components, the component calculations are then performed. The remaining correlation degree, if the component The correlation with the core components is not satisfied. Then the component is not calculated. The remaining correlation degree, the initial value of the remaining correlation degree of each component is set to 1, and the component Remaining correlation The calculation formula is as follows: Set threshold When the component of and When, the component From the collection of components to be processed Remove from the middle, component Once the tag preprocessing is complete, when the component... of When, describe the component There is also the possibility of combining it with other core components, in which case the components will be... Retained in the collection of call processing components.

Images