Component scheduling method based on load balancing strategy and improved ant colony algorithm in multiple industrial networks

Abstract

In today's complex industrial network environment, the efficiency of industrial task completion depends on the coordination and interaction between multiple networks. The invention proposes a component deployment method based on load balancing and ant colony algorithm in multiple industrial networks. Through improving the ant colony algorithm, the search and scheduling of different subnet components are realized. Through the load balancing strategy, the problem of excessive load in multiple networks is solved. According to the business logic relationship of specific tasks, the crawling of ants is used to simulate the scheduling process of components within and across the network. After multiple iterations, the best path of related component scheduling is planned, and the self-assembly of components within and across the network is realized. Without affecting the execution of multiple target tasks, this method can effectively reduce the scheduling time and cost, improve the utilization of components, and improve the parallel execution efficiency of multiple industrial software systems under the micro-service framework.

Description

Technical Field

[0001] This invention relates to searching and scheduling relevant components in various subnets within a multi-industry network environment, based on the business logic and processes of specific tasks. Based on load balancing strategies and an improved ant colony algorithm, the optimal path for component scheduling is planned, enabling scheduling and self-assembly of components within and across networks. This method balances efficiency and cost, effectively reducing scheduling time and cost without affecting multiple target tasks, and improving the parallel execution efficiency of multiple industrial software systems within a microservices framework. Background Technology

[0002] Existing industrial software systems are typically based on a single network architecture. While different industrial networks can currently share data to some extent, complete integration and collaboration between industrial software systems in different network environments are lacking. With the development of industrial technology and enterprise information intelligence, today's industrial software systems are simultaneously constrained and influenced by multiple network flows, including information flow, control flow, and business flow. Furthermore, complex industrial tasks urgently require collaboration between different industrial software systems across multiple industrial networks. However, the lack of effective integration and collaboration between traditional single-network architecture industrial software systems leads to poor collaborative optimization and excessive local load. Therefore, how to solve the component scheduling and assembly problems based on complex industrial tasks through component collaboration and interaction between different software systems in multiple industrial networks has become a key focus and challenge in improving the parallel execution efficiency of industrial software systems.

[0003] In recent years, with the rapid development of computer technology, microservices, as a service-oriented approach, essentially involves modularizing large monolithic application systems. This involves using fine-grained services, replacing the original large monolithic application with a series of independently designed, asynchronously developed, distributed, and independently maintained software service units. Each service uses its own database, avoids strong dependencies, and collaborates without high coupling. The multi-network built under the microservice model integrates independently, with services divided at a granular level and deployed in a distributed manner. Task scheduling spans multiple different component networks, each providing different component services. This clear logic and flexible expansion make it highly suitable for today's rapidly developing industrial software systems.

[0004] When a localized portion of an industrial system experiences excessive load, task reallocation is necessary to ensure the overall system's task completion. Tasks that these overloaded components cannot complete due to overload are then reassigned to lighter or idle components. However, differences in scheduling algorithms across systems lead to variations in scheduling order and schemes. Furthermore, different components have varying usage frequencies and execution efficiencies. Therefore, industrial systems are prone to localized overload. If this situation is not addressed promptly, continuing task scheduling will further exacerbate the component burden and hinder the execution of multiple target tasks.

[0005] In industrial systems, uneven load distribution is typically handled manually to allocate tasks that are overloaded. However, this approach is not always effective in detecting and addressing overload promptly; it often only becomes apparent after the overload has already impacted tasks. This invention addresses this issue by employing a component scheduling method based on load balancing and an improved ant colony algorithm within a microservice framework. The method uses the crawling process of ants to simulate the assembly process of components within and across networks, applying specific parameters to relevant components. The heuristic function of the ants depends on these parameters, and the ant crawling process is equivalent to component scheduling. Once all ants have crawled, the final path obtained is the component assembly path that meets the task requirements, and components are self-assembled according to the assembly sequence. This method introduces a load balancing strategy during path search, so even in cases of uneven load distribution, this strategy helps to balance the load through several scheduling iterations. Summary of the Invention

[0006] Technical Problem: The purpose of this invention is to propose a component scheduling method based on load balancing and ant colony optimization in multi-network industrial systems, enabling task allocation of components within and across networks, thereby addressing the problems mentioned in the background. This method primarily improves the traditional ant colony optimization algorithm using a load balancing strategy, employing the crawling motion of ants to simulate the scheduling of components within and across networks, ensuring it meets the requirements of specific working conditions and tasks. The final path reached by the ants after several iterations is the optimal path for component scheduling. This method can alleviate the problem of excessive load in component scheduling in industrial networks to a certain extent and can find the optimal scheduling path that satisfies the task in a relatively short time. It improves the utilization rate of components in the industrial network environment to a certain degree, and also improves the parallel execution efficiency of multiple industrial software systems to a certain extent.

[0007] Technical Solution: In a complete multi-industry network environment, each component is responsible for a considerable number of tasks, and the completion of each industrial task depends on the orderly scheduling of components. However, the performance of components is not static. As the scheduling environment, such as usage time and call intensity, changes in the component load will increase, and its performance will be affected to some extent. At this time, it is necessary to allocate the tasks processed by the components, reassigning some pending tasks to components with lighter loads or no load.

[0008] To ensure orderly and efficient scheduling among components, this invention is implemented through the following steps: First, initialize the number of ants m, pheromone heuristic factor α, heuristic function factor β, path length L, pheromone evaporation factor ρ, pheromone constant Q, maximum number of iterations T, and pheromone concentration τ. ij Construct the solution space G for component scheduling.<V,E> The components are likened to ants, and m ants are placed on nodes in a network. Each ant's path selection uses a roulette wheel algorithm. Based on the cost N of the currently scheduled component, the assembly time T for different component designs, a load balancing coefficient X is defined, and the average load value X_AVE of i nodes is recorded. ij Calculate the load variance X_s of the i nodes. A larger load variance X_s indicates a heavier workload and longer scheduling time for that node (component), thus indicating a higher network load. Conversely, a smaller load variance X_s indicates a lighter workload and shorter scheduling time for that component, thus indicating a relatively lower network load. Based on this data, calculate the probability of the ant crawling to the next node.

[0009]

[0010] Where i and j are the starting point and the next node of an ant's crawling during the execution process, respectively. η represents the probability that an ant will crawl from node i to node j. ij This represents the reciprocal of the path distance between two points i and j. k This represents a node that has not yet been visited. Under constraints, P... ij This will change with the load balancing value. When the load value X_s of a node is smaller (i.e., the node's load is lower), the probability P of the ant choosing this node increases. ij The higher the load value X_s of a node (i.e., the heavier the node's load), the greater the probability P that an ant will choose that node. ij It will decrease.

[0011] The ants crawl according to the roulette wheel algorithm to select the next node. After each path movement, the ant adjusts the pheromone concentration τ.ij An update is performed, and after crawling for one cycle, the ants return to the initial component, thus completing a scheduling path and a local update of the pheromone. After all ants have completed one cycle of crawling, a global pheromone update is performed on the network. Due to the influence of the pheromone, during several iterations, all ants converge on a single path. The resulting assembly path is the optimal path for component allocation, covering both intra-network and cross-network search ranges. This path is derived under the condition that all nodes in the entire scheduling system are under load balance. Therefore, it alleviates the pressure on some related components to a certain extent.

[0012] Beneficial effects:

[0013] 1) Reduced scheduling costs and time: This invention fully utilizes the characteristics of the ant colony algorithm, analogizing the component scheduling process to the foraging process of ants, and using heuristic information to complete path searching. This not only broadens the search scope of scheduling schemes, covering both intra-network and cross-network search spaces, but also accurately identifies the optimal scheduling scheme, optimizing the structure of the scheduling mode to a certain extent, improving scheduling efficiency, and reducing the time required for component scheduling.

[0014] 2) Improved security and stability of component assembly system. The present invention is based on a load balancing scheduling scheme, so there is no situation where individual nodes are overloaded. It significantly improves the overall stability of scheduling and local optima, and improves the utilization rate of components in multiple task cycles, which facilitates the self-assembly of components that meet the service composition mode.

[0015] 3) Balances the load of each component in the network. This invention introduces a load balancing strategy to affect the component scheduling process of the entire industrial network, avoiding the negative impact of some components being overloaded, ensuring that the scheduling task can be carried out in a better environment, and improving the parallel execution efficiency of multiple industrial software systems to a certain extent. Attached Figure Description

[0016] Figure 1 This is a schematic diagram illustrating the scheduling of industrial task components in a traditional single network.

[0017] Figure 2 This diagram illustrates component scheduling driven by multiple network tasks in a microservices model.

[0018] Figure 3 This is a schematic diagram of a multi-network component scheduling method based on load balancing and ant colony algorithm. Detailed Implementation

[0019] In a multi-industry network, in order for component scheduling to proceed normally, it is first necessary to detect the performance and load of each component in the network.

[0020] In a multi-industry network composed of microservice architectures, a multi-industry network consists of several industrial networks, each composed of components with similar characteristics. When a series of tasks need to be scheduled for related components, the efficiency depends on the coordination and interaction between multiple industrial networks. This method improves the traditional ant colony algorithm with a load balancing strategy, using the crawling of ants to simulate the scheduling process of components within and across networks. Each component, based on specific task requirements, completes the allocation and assembly within and across networks through a series of scheduling instructions, thereby completing the corresponding task. When the load of a scheduled component far exceeds the average load value of others, this method reduces the probability of that component being scheduled in each iteration. After multiple iterations, the load of the component will approach the average load value. Until the load of the scheduled component is relatively good, this method will search for an optimal path that meets the scheduling requirements for the task, and the self-assembly of components within and across networks can be completed according to this path.

[0021] • Initialization of parameters: Suppose there are n different types of components, and define a K*K matrix A. o To represent the number of components required for the finished product, we further define a K*K matrix A. n Let O represent the number of nodes in the component library at a certain moment, if matrix O = A n -A o If any eigenvalue of matrix O is negative, the system will provide feedback indicating that the task cannot be completed. If no negative eigenvalues ​​exist in matrix O, meaning the assembly requirements of various components are met, the system will provide a signal indicating that assembly path searching is possible. Initialize pheromone concentration τ. ij =τ0i,j∈K. Initialize the number of ants m, pheromone heuristic factor α, heuristic function factor β, path length L, pheromone evaporation factor ρ, pheromone constant Q, maximum number of iterations T, and pheromone concentration τ. ij The pheromone heuristic factor, heuristic function factor, pheromone evaporation factor, pheromone constant, maximum number of iterations, and pheromone concentration are important characteristic factors in the ant colony algorithm, determining its execution process. Setting these values ​​too high or too low has a serious impact on the results. Therefore, when initializing these parameters, appropriate settings should be made according to the actual situation.

[0022] • Construct a component scheduling space: Set the total number of ants to m. Define K industrial networks, and select components as nodes in each industrial network. Randomly place m ants on the nodes of the K industrial networks. Each time an ant crawls, it selects its next node using a roulette wheel algorithm. When an ant reaches node i at time t, first calculate the cost N of that component. t The time T required for this component to be scheduled t The load factor X of this component:

[0023] X = η1 × N t +η2×T t η1 and η2 represent the weights of the available quantity of components and the assembly time.

[0024] The load mean and load variance are calculated from the above values.

[0025] Average load of i nodes

[0026] Because variance can effectively reflect the degree of fluctuation of a system, it can more intuitively reflect the load of the entire network.

[0027] Therefore, the load variance of the i nodes

[0028] This is used to assess the load of the current node and adjust the network scheduling structure accordingly. The load factor is commonly used to analyze network load; however, in this invention, since the load factor consists of component cost and component assembly time—both of which are target factors for component scheduling—it also serves as a heuristic function in the ant colony algorithm, acting as one of the decisive factors for selecting the next node.

[0029] Ant colony optimization (ACO) is a unique heuristic algorithm, and its heuristic function determines the direction of execution. Therefore, the heuristic function for this algorithm is defined as follows:

[0030]

[0031] Where X i Let X_s represent the load value of node i, and X_AVE represent the average load value of nodes i. The average load value refers to the average of the load balancing values ​​of nodes i. After constructing the heuristic function, the probability of reaching subsequent nodes can be calculated using this function. The selection probability of subsequent nodes can be calculated using the pheromone concentration τ, the pheromone heuristic factor α, the heuristic function X_s, and the heuristic function factor β.

[0032]

[0033] As can be observed from the above formulas, the lower the cost and assembly time of a component, the lower its complexity balance coefficient, and the more likely it is to become a node in the final scheduling sequence. When a component's load coefficient is too high, its probability of being selected as the next node is lower, and its load will gradually decrease after several iterations. Conversely, when a component's load coefficient is too low, its probability of being selected as the next node will gradually increase, thus significantly improving the component's utilization rate.

[0034] Each iteration is equivalent to all ants completing a path search, selecting the shortest path that satisfies the scheduling requirements, and then performing a local pheromone update. The pheromone update rules are as follows:

[0035] τ ij (t+1)=τ ij (t)*(1-ρ)+△τ ij , 0 < ρ < 1, Δτ ij =Q / L

[0036] Where ρ represents the pheromone evaporation coefficient, Q represents the pheromone constant (i.e., the total amount of pheromone released by a single ant), L represents the path length of the ant, and Δτ ij This indicates the change in pheromone levels.

[0037] Ants release pheromones along their paths, but the pheromone concentration evaporates in a ρ-proportional ratio over time. Therefore, as the number of iterations increases, the pheromone concentration on the scheduling path that best suits the task will be higher, while the pheromone concentration on other, more complex paths will gradually decrease due to pheromone evaporation. Since ants always crawl along the path with the highest pheromone concentration, the final crawling route of the entire ant colony is the scheduling path required by the task.

[0038] • Once all ants have completed their pathfinding, the optimal path for this iteration can be determined, at which point the global pheromone level is updated:

[0039]

[0040] In the formula, D is a constant, and L best This represents the optimal path length in this iteration.

[0041] After the global pheromone update, check if the number of iterations has reached the set maximum value. If it is not less than the maximum value, output the optimal path. Otherwise, it indicates that there is a problem in the iteration process, and the path search needs to be performed again.

Claims

1. A component-level scheduling method based on load balancing and an improved ant colony algorithm in a multi-industry network. This method, in a complex and dynamic industrial network environment, uses an improved ant colony algorithm to achieve path search for scheduling components in different subnets, and employs a load balancing strategy to address the overload problem in multi-industry networks. This yields an optimal path that meets task requirements, and based on this path, self-assembly of components within and between networks is completed. The multi-industry network consists of K industrial networks, where components are selected as nodes within each industrial network, using a K*K matrix A. o To represent the number of each component required for the finished product, we further define a K*K matrix A. n Let O represent the number of nodes in the component library at a certain moment, if matrix O=A n -A o If any eigenvalue of matrix O is negative, the system will provide feedback indicating that the task cannot be completed. If no negative eigenvalues ​​exist in matrix O, meaning the assembly requirements of various components are met, the system initializes the following parameters: number of ants m, pheromone heuristic factor α, heuristic function factor β, path length L, pheromone evaporation factor ρ, pheromone constant Q, maximum number of iterations T, and pheromone concentration τ. ij The value of m is set to 1.5 times the number of components; the pheromone constant is selected according to the actual industrial environment [10, 1000], the maximum number of iterations T is selected [200, 500], the pheromone heuristic factor α reflects the relative importance of the amount of pheromone accumulated on the path during the ant movement in the guided ant colony search, and the value range is between [1, 4], the heuristic function factor β reflects the relative importance of heuristic information in guiding the ant search process, the strength of the a priori and deterministic factors in the ant colony optimization process, and the value range is between [0, 5], the pheromone evaporation factor reflects the level of pheromone disappearance, and conversely, 1-ρ reflects the level of pheromone retention, and the value range is between [0.2, 0.5]. As a unique heuristic algorithm, the heuristic function determines the execution direction of the algorithm, so the heuristic function is defined as follows: ; wherein X i X_i represents the load value of the i th node, X_AVE ij X_AVE represents the average load value of i nodes, the load average value refers to the average of i node load balancing values, m is the total number of component nodes, after constructing the heuristic function, the probability of reaching each subsequent node is calculated, and the pheromone concentration , pheromone heuristic factor α, heuristic function X_s, heuristic function factor β, to calculate the selection probability of the subsequent node: ; where i, j are the start and next node of the crawling of an ant in the execution process; denotes the probability of an ant crawling from i to j; allowed k denotes the node that has not been visited; The crawling of ants is used to simulate the intra-network and cross-network scheduling process of components: m ants are randomly placed on each node in K industrial networks, and each time the ants choose the next node according to the roulette algorithm, the heuristic function is used to represent the visibility of the ants from one node to another, and the shorter the distance between the two places, the greater the probability of the path being selected by the greater concentration of pheromone, from the probability calculation formula, it can be seen that the pheromone heuristic factor alpha is the index of the concentration of pheromone, and the heuristic function factor beta is the index of the heuristic function, which determines the contribution degree of the concentration of pheromone and the transfer expectation to the possibility of the ants k from i to j; When the ant reaches the i node at time t, first calculate the cost N of the component t , the time T required for the component to be scheduled t , the load factor X of the component, X = η1 x N t + η2 x T t， η 1、 η2 represents the weight of the available amount of the component and the assembly time; the load mean and the load variance are calculated by the above values, the average load value of the i nodes , and therefore the load variance of the i nodes ; The load coefficient is used to evaluate the load condition of the current node, and the network scheduling structure is adjusted according to the load condition, and the load coefficient is used to analyze the load condition of the network.

2. The method of claim 1, wherein, Each completed iteration, equivalent to all the ants complete a path search, from which to select a shortest path to meet the scheduling requirements, a iteration refers to m pointer ants walk through all the nodes, that is, there are m search path, all paths are compared, select the shortest path, make the iteration of the visualization results, update pheromone, and the current shortest path with the past shortest path length comparison, while the number of iterations plus 1. Then judge the current iteration number is equal to the set iteration number, if equal to stop iteration, otherwise the next iteration, pheromone local update, pheromone update rule as follows: ; wherein, ρ represents the pheromone volatile factor, Q represents the pheromone constant, that is, the total amount of information released by a single ant, L represents the path length of the ant walking, The change of pheromone is represented by Δη, when all the ants complete the path search, the optimal path in this iteration is obtained, at this time the global pheromone is updated: ; The optimal path length in this iteration is represented by Lopt.

Images