Method for importance identification and scheduling of multiple networked industrial software components

Abstract

The component scheduling coordination and assembly technology in current industrial software system mainly aims at single industrial network architecture, ignores the constraint of multiple network coupling relationship, causes the lack of correlation collaboration between software systems, single mode granularity, and lack of data fusion, making it difficult to solve the coupling effect, correlation constraint, resource competition and other problems of industrial software system in multiple network environment. This patent proposes a multiple network industrial software component importance identification technology. First, the relative influence ability and the response ability of the component in the current layer are obtained by calculating the network core degree of the component, the corresponding grade and the influence distance of the component. Then, the total benefit of the component is obtained by the intra-layer and inter-layer benefit, and the dynamic benefit function of the component is obtained with the evolution of time. Finally, the utilization rate of the component is maximized by adding usage bit and modification bit to the task to be processed and making cyclic judgment.

Description

Technical Field

[0001] This invention relates to a technology for identifying intelligent components that support cross-network layer communication and collaborative decision-making in multi-networked industrial systems, improving the efficiency of multi-component collaboration and adapting to multi-industry network scenarios. Specifically, it relates to a technology for identifying key software components and optimizing the scheduling and allocation of component resources in multi-network environments. Facing the increasingly complex multi-network industrial environment, this method identifies key components that provide solutions for different tasks based on the role of components in each network layer and the collaborative relationships arising from cross-network layer correlations, thereby shortening processing cycles and improving operational efficiency while minimizing time and space overhead. Background Technology

[0002] With each leap forward in my country's industrial technology, the country's economic development has been greatly propelled, leading to higher demands for industrial technology. Industrial tasks are often broken down into sub-tasks, which are then processed in parallel or sequentially depending on their interrelationships. This not only improves overall efficiency but also facilitates later inspection and maintenance. Similarly, industrial software is gradually being broken down into components. By adhering to the principle of "high coupling and low cohesion," components that are easy to update and maintain are created. The introduction of these industrial software components has once again dramatically increased the speed of task processing.

[0003] In industrial software systems employing intelligent components, these components collaborate to complete industrial tasks. Current industrial software systems are typically based on a single network architecture, meaning that the interaction and collaboration between intelligent components occur through a single network relationship. However, with the increasing dynamic complexity of industrial environments, the connections between components in today's industrial software systems exhibit multiple networked patterns, including information flow, control flow, and business flow. For example, software components may collaborate simultaneously through supply chain networks, production workflow networks, and scheduling and control networks. Furthermore, the complex and dynamic demands of industrial tasks and the dynamic changes in networks place higher functional requirements on current intelligent software components. Therefore, current industrial software systems exhibit typical characteristics such as "multiple interrelationships in industrial software," "diverse component functions," and "complex and dynamic tasks." Addressing the new challenges brought by multi-layered industrial network environments to industrial software systems necessitates resolving crucial issues such as industrial software system architecture adaptable to multi-layered network environments, information sharing, and collaborative scheduling. However, traditional industrial software system models typically consider only a single network architecture based on the needs of a single industrial task, lacking integration, fusion, and optimized collaboration among industrial software components, and exhibiting fixed and unidirectional patterns and granularities in component scheduling and collaboration. This traditional single-network architecture industrial software system model leads to significant drawbacks such as information silos and inefficient software collaboration when faced with multi-networked industrial environments.

[0004] Therefore, building a collaborative and real-time feedback optimization mechanism for the assembly and scheduling of industrial software components, and realizing the intelligent identification and scheduling of industrial software components in a multi-networked environment, is the key and difficult point in promoting efficient industrial operations.

[0005] If we continue to use an architecture that considers a single network based on the needs of a single industrial task, the components will be highly integrated but lack flexibility, often failing to handle complex new tasks. Furthermore, the component scheduling and assembly model has a single, fixed granularity, making it difficult to quickly and effectively respond to the multi-granularity service demands of industrial software systems in multi-layered industrial network environments. This results in the software system's efficiency, robustness, and scalability not being effectively guaranteed. Specific situations are as follows... Figure 1 As shown in the figure, the industrial software system model only considers the single network link relationship between industrial software components. The components have a high degree of integration. Although it can accurately solve the given tasks 1, 2 and 3 within the scope, it cannot smoothly handle the fuzzy task 4, which is different from the previous ones. The integrated components, which are information silos and lack correlation and cooperation, cannot solve the problem smoothly.

[0006] Today, tasks are more complex than ever before. For massive tasks, industry needs to break them down into smaller parts so that different components can handle them, which requires more coordinated scheduling between these components. Figure 2 As shown, modern components can reside in different network layers based on their own attributes, functions, and the degree of correlation with other components under different circumstances. This granularity facilitates scheduling and collaboration between components and allows for flexible handling of various tasks. However, the complex relationships between network layers and between components within a network layer cause industrial systems to spend additional time making judgments and decisions, indirectly affecting efficiency and increasing energy consumption.

[0007] Therefore, it is necessary to research and design component identification and scheduling technologies that can adapt to multiple network architectures, so as to effectively solve problems such as green efficiency improvement, elastic service and intelligent component deployment in complex coupled scenarios such as multiple products, multiple batches and multiple workshops. Summary of the Invention

[0008] Technical Problem: The purpose of this patent is to propose a method for identifying and scheduling the importance of multiple networked industrial software components, in order to solve the problems mentioned in the background. Specifically, the method is mainly divided into three stages. The first stage is the calculation of the component's own attributes within the network layer. Based on the inherent attributes of the component, the coreness of the component in the current layer is calculated, and its influence distance and related intra-layer benefits are deduced. The second stage is the calculation of inter-layer benefits. By calculating inter-layer benefits, the component's effectiveness is accurately measured, thereby accurately identifying the component and facilitating subsequent scheduling. The third stage is component scheduling, using an improved clock permutation algorithm to operate the task. This method effectively shortens the component identification time, provides a reliable technical guarantee for subsequent component scheduling, and further saves industrial costs through the improved clock permutation algorithm. It provides a new approach to solving component scheduling in complex coupled industrial production scenarios such as multiple products, multiple batches, and multiple workshops.

[0009] Technical Solution: Current software component scheduling, coordination, and assembly technologies in industrial software systems are primarily designed for single industrial network architectures. This technology, by neglecting the constraints and influences of multiple network coupling relationships, results in a lack of inter-system collaboration, fixed and singular granularity, and a lack of data integration and sharing. This leads to information silos and inefficient software collaboration between different software systems on different networks, making it difficult to resolve the coupling effects, inter-constraints, and resource competition problems brought about by multi-networked environments to industrial software systems. Therefore, this patent proposes a multi-networked industrial software component importance identification technology, based on a microservice architecture, to achieve flexible scheduling, coordination, and intelligent processing of industrial software components in multi-networked environments. Compared to traditional single-networked multi-component industrial software system scheduling and coordination technologies, the new technology proposed in this patent deals with industrial software components located within a microservice framework. Because components can be independently developed, deployed, run, upgraded, and interconnected, they satisfy adaptive switching of service composition modes and dynamic adjustment of service granularity to form multi-granularity component services. To achieve accurate identification and efficient scheduling of components in complex, coupled industrial production scenarios involving multiple products, batches, and workshops, and to address the challenges of green efficiency improvement, flexible service, and intelligent deployment of industrial software components, this patent provides strong technical support for enhancing the efficiency, robustness, and scalability of industrial software systems in multi-network environments. The identification and scheduling technical solution of this patent is as follows:

[0010] (1) Data processing within the horizontal layer. By quantifying the relevant information of industrial software components in the microservice framework, a multi-network relationship graph that is easy to process is abstracted, thereby calculating the heterogeneous attributes of related components. By quantifying information such as the call frequency and response time between components, weighted edges are generated. In addition to obtaining the inherent graph attributes of components based on graph theory, the network coreness value SP of the components is calculated based on effective information. v Based on this, the components are divided into the kernel group G of the current network layer. N Or non-kernel group NG N Calculate the corresponding Rank(v) and influence distance D for components located in different groups. v and ND v By calculating from different dimensions, the relative influence capability (IF) of the component is obtained. v The comprehensive evaluation coefficient RF of the response capability at the current layer v .

[0011] (2) Vertical Inter-Layer Revenue Calculation. Components have different relationships at different layers, enabling them to flexibly handle different types of specific tasks. In other words, because of the multi-layer component architecture, the current component can be combined with other components to achieve multiple functions, resulting in more flexible working capabilities and higher revenue. The total revenue of a component can be obtained through intra-layer and inter-layer revenue, and the dynamic revenue function of the component can be obtained as time evolves.

[0012] (3) Task component scheduling stage. Each task is marked by adding a use bit and a modify bit. The task is processed differently according to four different cases. The utilization rate of the component is maximized by iterative judgment.

[0013] Beneficial effects:

[0014] (1) Accurate measurement and identification of technical components: By combining the inherent attributes of components with known information and based on graph theory, the differences between components can be accurately grasped. Accurate measurement of components is essential for a good understanding of their use, allowing for the differentiation of components and shortening the component consideration time in the process.

[0015] (2) Improving Component Utilization: To maximize component utilization, the elasticity between components needs to be maximized. Based on accurate component identification, components are divided into kernel clusters and non-kernel clusters, and gradient processing is applied to calculate intra-layer and inter-layer benefits. This quantifies the correlation between components, achieving a cascading effect from points to lines and from lines to surfaces. This improves component utilization within a given time period, enabling adaptive switching of service composition modes and dynamic adjustment of service intensity, facilitating later maintenance.

[0016] (3) Reduce the total system processing time by improving component utilization and further reducing the task loop waiting time during the process. Use an improved clock replacement algorithm to solve complex coupled industrial production problems such as multiple products and multiple batches. By dynamically judging and shortening the waiting time, more tasks can be processed by the target components, thereby achieving green, efficient and intelligent deployment of components as a whole. Attached Figure Description

[0017] Figure 1 This is an example diagram of traditional industrial components and task processing;

[0018] Figure 2 It is a schematic diagram of the scheduling, combination and task processing of multiple networked industrial components;

[0019] Figure 3 This is a schematic diagram of the main principle of the method of the present invention. Detailed Implementation

[0020] Based on the industrial processes and corresponding operational granularity, we define a multi-layer industrial software network as a multi-network system containing n network layers, N = (N1, N2, ..., N...). n Different network layer components are responsible for executing the processing of modules corresponding to different tasks. For a given task, components are scheduled both within the same layer and across different layers; some components are highly coupled, while others are loosely coupled. To complete a given task, the identification of the target component, the execution order between components, and the determination of priorities all affect the time and energy consumption required to complete the task. The main innovation of this patent lies in comprehensively considering the efficiency and coordination capabilities of the components, identifying key components, and scheduling them.

[0021] The specific implementation steps are as follows:

[0022] (1) Intra-layer identification and scheduling phase. We divide the industrial software components placed in the microservice framework according to specified functional standards, and generate a multi-layer network with weighted values ​​based on the scheduling frequency and response time between components. Each component with heterogeneous attributes is defined as component v = {E} v W v ,K v SP v ,S v ,IF v ,RF v}. Among them, E v W is the set of all associated edges connecting component v, including intra-layer edges and inter-layer edges. v K represents the weighted value of all edges corresponding to component v. v SP indicates the link degree of component v, including links within the intra-layer network and between the inter-layer network.v S is the network coreness value of component v in the network. v This represents the state of component v. Component scheduling occurring between and within network layers is divided into two types: intra-layer scheduling and cross-layer scheduling. IF v When indicating whether a component belongs to a kernel cluster or a non-kernel cluster, it represents the influence of components within that cluster. RF v This represents the responsiveness of component v to these schedules. In this scheduling model of the identification process, some properties of components in the multi-layer network can be further described as follows. The SP of each component... v It mainly depends on its own correlation degree and the weighted value of all related edges. The network coreness value of component v in network N1 is defined as follows:

[0023]

[0024] For component v, its network coreness value in the current layer N1 is... The network core strength value is obtained by comparing the sum of the weighted values ​​of the edges associated with the current component with the sum of the weighted values ​​of the edges associated with all components. Furthermore, the components in network N1 can be divided into two groups: the core group and the subgroup. or non-kernel group When a component's network core strength value is greater than the average network core strength value, it is classified as a kernel group. Nodes whose network core strength value is less than the average network core strength value will belong to the non-core cluster. Each component in different groups can adopt a different strategy in each scheduling. For each strategy, its ranking in different groups needs to be calculated based on the network core strength values ​​of all components and the benefits they gain in each scheduling:

[0025]

[0026] Among them, P v This represents the potential benefit that component v can obtain in each scheduling based on previous statistics. To assess the degree of influence between components in the current scheduling, the influence distance of component v needs to be defined.

[0027]

[0028] Here, d represents the shortest weighted path between components, which can be calculated using Dijkstra's algorithm in graph theory. It is the total number of components in the corresponding clique of network layer N1. Using a similar method as described above, the corresponding influence distance in the non-kernel clique of the network layer can be obtained.

[0029] The influence of component v from the core group is defined as follows:

[0030]

[0031] Here, σ1 and σ2 are two parameters representing relative importance, and f is a monotonically increasing function. Using the same idea, the formula for calculating the influence of components in non-core clusters can be obtained. In this scheduling model, component v represents the comprehensive evaluation coefficient of its responsiveness at the current layer.

[0032]

[0033] (2) Inter-layer identification and scheduling stage. Based on the interrelationships and collaborations between industrial software components, component v can indirectly access components in other layers through key components within or between layers, thereby achieving more flexible operational capabilities and higher returns, as detailed below:

[0034]

[0035] NE v It is the directly related component of component v in the current network layer, while CNE v It is an indirectly related component across layers of component v, and the resulting P v1 With P v2 These are referred to as the intra-layer and inter-layer correlations of component v. Here, we use a key parameter α to adjust and reflect the degree of close connection between the two networks. Then, after intra-layer and inter-layer interactions, the total gain of component v is P. v :

[0036] P v =αP v1 +(1-α)P v2 (7)

[0037] Clearly, a close interlayer interaction between the two networks implies a small parameter α. In particular, when α = 1, it indicates that there is no interlayer interaction between the two networks.

[0038] Since the benefit of calling subsequent components after the target component has been called is still measured by the parameter α, the benefit is either primarily provided by intra-layer associations or by inter-layer associations. Therefore, the benefits between components are as follows:

[0039]

[0040] Where t represents the response time.

[0041] (3) Task component scheduling phase. The traditional clock replacement algorithm, for a group of tasks to be processed, if there are sufficient operating resources, will load the target task. If it is to be component-scheduled and processed, its usage is set to 1, and it will still be 1 for secondary use. For tasks that have been loaded but have higher priority or higher priority calls of the component itself, the task will be suspended and wait. The tasks to be swapped out are linked in a loop into a queue. If all tasks to be processed are 1 (i.e., all have been processed), the task is cyclically set to 0 and the first 0-state task is found (the first recently unused task). If all tasks to be loaded are 0 (nearly unused), the first 0-state task is found. If there are tasks in 1 and 0 states, the first 0-state task is found and the 1-state tasks along the way are set to 0.

[0042] Our proposed component replacement algorithm adds a modification bit (1 indicates usage, 0 indicates non-use) to the existing usage bit (1 indicates modification, 0 indicates no modification). Therefore, the task has four possible states: (0,0) state: not recently loaded and not processed; (0,1) state: target component loaded but no task processing action; (1,0) state: task processing action performed but target component not loaded; (1,1) state: target component loaded and corresponding task processing performed.

[0043] During the replacement, among tasks of the same priority, first look for tasks with status code 00, that is, tasks that have not been recently loaded and have not been processed; if there are no tasks with status code 00, then look for tasks with status code 01. When looking for tasks with status code 01, set the usage bits of the tasks along the way to 0 (that is, unrelated to the modification bits); if there are no tasks with status code 01, then look for tasks with status code 00 first, then look for tasks with status code 01, and repeat the search (because the bits along the way are set to 0).

[0044] The improved clock replacement algorithm addresses interruptions in the console's component scheduling caused by high-priority tasks by suspending currently delayed tasks—that is, swapping out the most recently unused task—rather than selecting an incoming task that meets certain criteria. Essentially, it finds the most recently unused swap-out. The core of the proposed component replacement algorithm lies in subdividing the scheduling process into target component incoming tasks and target components that have not yet been loaded.

[0045] The unimproved replacement algorithm swaps out a task after a maximum of two scans, while the improved replacement algorithm requires a maximum of four scans before swapping out a page. This reduces the frequency of task swapping in and out, and through secondary consideration, shortens the overall industrial process time.

Claims

1. A method for identifying the importance of multi-networked industrial software components. This method identifies key components that provide solutions for different tasks by reducing time and space overhead, based on the role of components in each network layer and the collaborative relationship generated by the correlation between network layers. By quantifying the inherent information of components in a microservice framework, including interaction frequency and response time, a network graph with a multi-layered network topology is abstracted. Based on graph theory, the inherent topological attributes of the components are calculated, and the influence distance reference parameters of the components are obtained by calculating intra-layer and inter-layer benefits. Intra-layer refers to the same network layer, inter-layer refers to different network layers; influence distance refers to the degree of influence between currently scheduled components; Based on the industrial processes and corresponding operational granularity, a multi-layer industrial software network is defined as a multi-network system containing n network layers: N = (N1, N2, ..., N...). n Different network layer components are responsible for executing the processing of different task modules. For a given task, components are scheduled between the same layer and between different layers. Some components are highly coupled, while others are loosely coupled. Industrial software components placed within a microservice framework are divided according to specified functional standards. A multi-layered network with weighted values ​​is generated based on the scheduling frequency and response time between components. Each component with heterogeneous attributes is defined as component v = {E}. v W v ,K v SP v ,S v ,IF v ,RF v }; where E v W is the set of all associated edges connecting component v, including intra-layer edges and inter-layer edges; v K represents the weighted value of all edges corresponding to component v. v The link degree of component v refers to the total number of associated links, including links in intra-layer and inter-layer networks; SP v S is the network coreness value of component v in the network; v This represents the state of component v; component scheduling occurring between and within layers of the network is divided into two types: intra-layer scheduling and cross-layer scheduling; IF v When indicating whether a component belongs to a kernel cluster or a non-kernel cluster, it represents the influence of components within that cluster. RF v This indicates the responsiveness of component v to scheduling; The network coreness value of component v in network N1 is: For component v, its network coreness value in the current layer N1 is... The network core strength value is obtained by comparing the weighted sum of the associated edges of the current component with the weighted sum of the associated edges of all components. Furthermore, the components in network N1 are divided into two groups: the core group and the subgroup. or non-kernel group When a component's network core strength value is greater than the average network core strength value, it is classified as a kernel group. Nodes whose network core strength value is less than the average network core strength value will belong to the non-core cluster. Each component in different groups employs a different strategy for each scheduling; The formula for influence distance is: d represents the shortest weighted path between components, calculated using Dijkstra's algorithm in graph theory; It is the total number of components in the corresponding clique of network layer N1; The formulas for intra-layer correlation and inter-layer correlation are: NE v It is the directly related component of component v in the current network layer, while CNE v It is an indirectly related component across layers of component v, P v1 With P v2 These are the intra-layer correlation and inter-layer correlation of component v, respectively; Using a key parameter α to adjust and reflect the degree of close connection between the two networks, the total gain of component v after intra-layer and inter-layer interactions is P. v =αP v1 +(1-α)P v2 When α = 1, it indicates that there is no inter-layer interaction between the two networks. Since the benefit of calling subsequent components after the target component has been called is still measured by the parameter α, the benefit between components is: t represents the response time; For each strategy, it is necessary to calculate their ranking in different groups based on the network coreness values ​​of all components and the benefits they gain in each scheduling. Among them, P v The revenue gained by component v in each scheduling is calculated statistically. The influence of component v from the core group is defined as follows: Where σ1 and σ2 are two parameters representing relative importance, and f is a monotonically increasing function; in this scheduling model, the comprehensive evaluation coefficient of component v's response capability at the current layer is:

Images