Distributed heterogeneous task scheduling method and system based on ant colony algorithm

Abstract

The invention provides a distributed heterogeneous task scheduling method and system based on an ant colony algorithm, and the method is constructed on a scheduling framework of a distributed heterogeneous task scene, and when a new task instance exists, a scheduler firstly checks whether the resource of each node meets the demand of the task instance; if the resources of the nodes meet the requirements, resource utilization rate heuristic values of the nodes are calculated; and in combination with the pheromone concentration and the heuristic value of the node, calculating to obtain a mixing score of the node and carrying out normalization processing. By comparing the mixed scores of the nodes, selecting the node with the highest score as a target node of the to-be-scheduled task instance; and updating the pheromone concentration of the node, and persistently storing the updated pheromone concentration table. Through a dynamic pheromone updating mechanism and a heuristic search strategy of the ant colony algorithm, the task scheduling decision is optimized, the cluster resource utilization rate and the task execution efficiency are improved, the global optimization capability of the ant colony algorithm is fully utilized, and more intelligent and more efficient resource scheduling is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of task scheduling in distributed systems, and in particular to a distributed heterogeneous task scheduling method and system based on the ant colony algorithm. Background Art

[0002] With the rapid development of information technology, distributed systems are increasingly widely used in fields such as cloud computing, big data processing, artificial intelligence, and high-performance computing. Task scheduling, as a key link in distributed systems, is directly related to the utilization rate of system resources, the execution efficiency of tasks, and the overall performance. In these applications, efficient and intelligent task scheduling methods play a crucial role in improving the overall performance of the system.

[0003] Traditional task scheduling methods mainly rely on static rules and instantaneous resource states for decision-making, lacking the ability to perceive historical load trends. This leads to the problem that in a dynamic heterogeneous environment, nodes may be continuously selected due to short-term idleness and then suddenly overloaded, while new nodes are idle for a long time due to no historical records, and the problem of uneven resource allocation occurs frequently. In addition, problems such as resource contention and scheduling delay are also common.

[0004] Although current scheduling technologies have the ability of global search, their convergence speed is slow and the computational cost is high, which cannot meet the requirements of real-time scheduling and large-scale clusters. Traditional meta-heuristic algorithms (such as genetic algorithms and particle swarm optimization) perform poorly in these problems. For example, when solving the task scheduling problem, the genetic algorithm has the ability to globally and quickly search for the optimal solution, but the algorithm itself requires more parameters to be set, the programming implementation is complex, and it is easy to fall into a local optimal solution when the computational scale is large. The particle swarm optimization algorithm has a fast convergence speed in the early stage and a slow convergence speed in the later stage. Although it is superior to the genetic algorithm in the ability to search for the optimal solution, it also has the problems of complex parameter setting and large computational cost. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present invention provides a distributed heterogeneous task scheduling method and system based on the ant colony algorithm. The present invention combines local and global pheromone updates, dynamically adjusts the pheromone increment and decay rate, avoids premature convergence to the local optimal solution, thereby optimizing the task scheduling strategy, improving the rationality of resource allocation and scheduling efficiency, reducing the task waiting time, and enhancing the stability and availability of the system.

[0006] The technical solution of the present invention is as follows: A distributed heterogeneous task scheduling method based on the ant colony algorithm, including the following steps:

[0007] S1), construct a scheduler applicable to the distributed heterogeneous task scenario and integrate the ant colony algorithm;

[0008] S2), Initialize the parameters required by the scheduler and the ant colony algorithm, and obtain the relevant information of each node and task instance in the distributed heterogeneous task scenario;

[0009] S3), When a new task instance is submitted to the cluster, the scheduler first checks whether the resources of each node meet the requirements of the task instance;

[0010] S4), For the eligible nodes, calculate the heuristic value of their resource utilization rate;

[0011] S5), Combine the pheromone concentration and heuristic value of the node, and use the ant colony algorithm to calculate the mixed score of the node;

[0012] S6), According to the normalized score, select the node with the highest score as the target node for the task instance to be scheduled, and schedule the task instance to this node; at the same time, calculate the reward value according to the resource usage of the node, and update the pheromone concentration of the node in combination with the pheromone decay rate and the global average pheromone concentration;

[0013] S7), Check the updated pheromone concentration and persist the updated pheromone concentration table to the storage.

[0014] Preferably, in step S3), the scheduler calculates the total resource request of the task instance and compares it with the allocable resources of each node. Nodes that do not meet the resource requirements will be excluded from the optional nodes.

[0015] Preferably, in step S4), the calculation expression of the heuristic value of the resource utilization rate of the node is:

[0016]

[0017] In the formula, h represents the heuristic value of the resource utilization rate of the node; CPU usage is the usage of the CPU of the node; Memory usage is the usage of the memory of the node; ∈ is a constant to prevent division by zero;

[0018] The lower the resource utilization rate of the node, the higher the heuristic value will be obtained.

[0019] Preferably, in step S5), the pheromone concentration of the node is used to simulate the preference of ants in path selection, and the mixed score score of the node is expressed as:

[0020] score = (p α ) × (h β ) (2)

[0021] In the formula, p is the pheromone concentration of the node; h is the heuristic value of the node, and α and β represent weight parameters.

[0022] Preferably, in step S6), after normalizing the mixed scores of all nodes to ensure the comparability of the mixed scores between different nodes, the normalization process is specifically as follows:

[0023] S61), find the highest score maxScore and the lowest score minScore of all nodes;

[0024] S62), adjust the score of each node to the range of 0-100 according to non-linear compression; if the highest score maxScore and the lowest score minScore are the same, set the scores of all nodes to the middle value 50.

[0025] Preferably, in step S6), during the process of scheduling the task instance to the node, update the status information of the node to ensure the correct execution of the scheduling decision. After a task instance is successfully scheduled to the node, the scheduler immediately obtains the current resource usage and task execution status of the node.

[0026] Preferably, in step S6), calculate the reward value according to the resource usage of the node, and update the pheromone concentration of the node in combination with the pheromone decay rate and the global average pheromone concentration. Specifically:

[0027] p new = p old ×(1 - ρ)+Δτ * localWeight+globalAvg * globalWeight

[0028] In the formula, p new represents the updated pheromone concentration of the node; p old represents the current pheromone concentration of the node; ρ represents the pheromone decay rate; Δτ represents the reward value of this scheduling; localWeight and globalWeight respectively represent the local pheromone weight and the global pheromone weight; globalAvg represents the average value of the pheromone of all nodes.

[0029] The formula for calculating the reward value is specifically:

[0030] Δτ = 100×(1.0 - currentLoad)

[0031] In the formula, currentLoad represents the real-time load of the node.

[0032] Preferably, in step S7), check the updated pheromone concentration. If the updated pheromone concentration exceeds the preset upper and lower bounds, adjust the updated pheromone concentration to the default value; and persist the updated pheromone concentration table to the storage.

[0033] The present invention provides a distributed heterogeneous task scheduling system based on the ant colony algorithm. The system is based on a scheduler and an ant colony algorithm framework in a distributed heterogeneous task scenario, and includes:

[0034] An initialization module, which is used to initialize the parameters in the scheduling process and obtain the relevant information of each node and task instance in the distributed heterogeneous task scenario;

[0035] A node screening module, which is used to screen the nodes that meet the scheduling conditions. When a new task instance is submitted to the cluster, the node screening module checks whether the resources of each node meet the requirements of the task instance, takes the nodes that meet the requirements as optional nodes, and deletes the nodes that do not meet the conditions;

[0036] A node heuristic value calculation module, which is used to calculate the heuristic value of a node according to the usage of the CPU and memory resources of the node;

[0037] A node mixed score calculation module, which is used to calculate the mixed score of a node according to the heuristic value of the node and the pheromone concentration of the node;

[0038] A node score normalization processing module, which is used to perform normalization processing on the mixed scores of all nodes;

[0039] A binding module, which is used to select the node with the highest score as the target node of the task instance to be scheduled according to the normalized score, and bind the task instance to this node;

[0040] An update module, which is used to calculate the reward value according to the resource usage of the node, update the pheromone concentration of the node in combination with the pheromone decay rate and the global average pheromone concentration, and check the updated pheromone concentration and persist the updated pheromone concentration table to the storage.

[0041] The beneficial effects of the present invention are as follows:

[0042] 1. By introducing the dynamic pheromone update mechanism and heuristic search strategy of the ant colony algorithm, the present invention optimizes the task scheduling decision, improves the cluster resource utilization rate and task execution efficiency. At the same time, by combining the global optimization ability of the ant colony algorithm and the flexibility of the distributed scheduling framework, it realizes a more intelligent and efficient resource scheduling;

[0043] 2. Based on the global optimization characteristics of the ant colony algorithm, the present invention simulates the group behavior of ants, comprehensively considers the pheromone concentration and heuristic information of all nodes, and searches for the global optimal solution; thus, it can more effectively utilize the cluster resources, improve the task execution efficiency, and reduce the scheduling delay;

[0044] 3. By integrating the dynamic pheromone update mechanism of the ant colony algorithm, the present invention realizes the real-time perception of the node resource usage situation. By dynamically adjusting the pheromone concentration, it reflects the resource load status of the nodes, thereby scheduling tasks to nodes with abundant resources and avoiding resource waste and overloading problems.

[0045] 4. The present invention utilizes the pheromone update mechanism to quickly identify faulty nodes. When a node fails, by reducing its pheromone concentration, the probability of subsequent tasks being scheduled to the faulty node is reduced; effectively solving the problem of delay in node failure handling, improving the availability and reliability of the system, and reducing the possibility of tasks failing due to node failures. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic flow chart of the scheduling method according to Embodiment 1 of the present invention;

[0047] Figure 2 It is a schematic framework diagram of the scheduling system according to Embodiment 2 of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0048] The following further describes the specific embodiments of the present invention in conjunction with the accompanying drawings:

[0049] Embodiment 1

[0050] As Figure 1 shown, this embodiment provides a distributed heterogeneous task scheduling method based on the ant colony algorithm. This embodiment is built on a scheduling framework for distributed heterogeneous task scenarios, and specifically includes the following steps:

[0051] S1). Construct a scheduler applicable to the distributed heterogeneous task scenario and integrate the ant colony algorithm;

[0052] S2). Initialize the parameters required for the scheduler and the ant colony algorithm, and obtain the relevant information of each node and task instance in the distributed heterogeneous task scenario;

[0053] In this embodiment, the initialized parameters include pheromone, heuristic factor, and pheromone decay rate;

[0054] S3). When a new task instance is submitted to the cluster, the scheduler first checks whether the resources of each node meet the requirements of the task instance;

[0055] The scheduler calculates the total resource request of the task instance and compares it with the allocable resources of each node. Nodes that meet the conditions are used as optional nodes, while nodes that do not meet the resource requirements are excluded from the optional nodes.

[0056] S4). For the nodes that meet the conditions, calculate their heuristic values of resource utilization rate;

[0057] In this embodiment, the heuristic value reflects the current resource tension of the node. Nodes with lower resource utilization will obtain higher heuristic values. By analyzing the usage of resources such as CPU and memory of the node, a heuristic value is generated for each node, specifically as follows:

[0058]

[0059] In the formula, h represents the heuristic value of the resource utilization rate of the node; CPU usage is the usage of the CPU of the node; Memory usage is the usage of the memory of the node; ∈ is a constant to prevent division by zero.

[0060] S5), Combine the pheromone concentration and heuristic value of the node, and use the ant colony algorithm to calculate the mixed score of the node;

[0061] The mixed score of the node consists of two parts: one part is the heuristic value calculated based on the resource utilization rate, which reflects the current resource tension of the node; the other part is the pheromone concentration of the node, which synthesizes the successful experience of historical scheduling and reflects the advantages of the node in long-term scheduling. The pheromone concentration of the node is used to simulate the preference of ants in path selection. The mixed score score of the node is expressed as:

[0062] score=(p α )×(h β ) (2)

[0063] In the formula, p is the pheromone concentration of the node; h is the heuristic value of the node, and α and β represent weight parameters.

[0064] During the scheduling process, this embodiment can better balance resource utilization and task execution efficiency. In a cluster with both high-priority urgent tasks and ordinary long-term tasks, the scheduler can quickly find nodes that can not only meet the resource requirements of urgent tasks but also reasonably arrange ordinary tasks according to the pheromone concentration, avoiding frequent scheduling adjustments caused by simple resource matching, thereby significantly improving the average completion speed of tasks and the overall utilization rate of resources.

[0065] S6), According to the normalized score, select the node with the highest score as the target node for the task instance to be scheduled, and schedule the task instance to this node; at the same time, calculate the reward value according to the resource usage of the node, and update the pheromone concentration of the node in combination with the pheromone decay rate and the global average pheromone concentration;

[0066] In this embodiment, by performing normalization processing on the mixed scores of all nodes, it is ensured that the mixed scores between different nodes are comparable. The specific normalization processing is as follows:

[0067] S61), find the maximum score maxScore and the minimum score minScore of all nodes;

[0068] S62), adjust the scores of each node to the range of 0 - 100 according to non - linear compression. Specifically, the non - linear compression is implemented through the Sigmoid function, which is used to suppress the influence of extreme values on the scheduling decision. First, determine the score range through the maximum score maxScore and the minimum score minScore, and then normalize the score of each node to map it to the range of 0 to 1. Next, apply the Sigmoid function to perform non - linear compression on the normalized value, making the score change smoothly near the middle value and steeply at both ends, thus effectively suppressing extreme values. Finally, map the compressed value to the range of 0 to 100 to obtain the final normalized score. If the maximum score maxScore and the minimum score minScore are the same, set the scores of all nodes to the middle value of 50.

[0069] During the process of scheduling task instances to nodes, update the status information of the nodes to ensure the correct execution of the scheduling decision. When a task instance is successfully scheduled to a node, the scheduler immediately obtains the current resource usage and task execution status of the node.

[0070] Calculate the reward value according to the resource usage of the node, and update the pheromone concentration of the node in combination with the pheromone decay rate and the global average pheromone concentration. Specifically:

[0071] p new =p old ×(1 - ρ)+Δτ*localWeight+globalAvg*globalWeight

[0072] In the formula, p new represents the updated pheromone concentration of the node; p old represents the current pheromone concentration of the node; v represents the pheromone decay rate; p old ×ρ simulates the evaporation of pheromone in nature; Δτ represents the contribution value of this scheduling; localWeight and globalWeight represent the local pheromone weight and the global pheromone weight respectively; globalAvg represents the average pheromone of all nodes.

[0073] The formula for calculating the reward value is specifically:

[0074] Δτ=100×(1.0 - currentLoad)

[0075] In the formula, currentLoad represents the real - time load of the node.

[0076] The update mechanism of this embodiment not only considers the success or failure of the current scheduling, but also takes into account the scheduling history and trends of the entire cluster, enabling the pheromone concentration to dynamically reflect the long-term scheduling value of nodes. During the scheduling process, as the number of scheduling times increases, nodes with higher pheromone concentrations will gradually stand out and become the preferred choices for subsequent scheduling, thus achieving self-optimization of the scheduling strategy. By simulating the behavior of ants releasing pheromones on paths, self-optimization and continuous improvement of scheduling decisions are achieved.

[0077] S7), Check the updated pheromone concentration and persist the updated pheromone concentration table to storage.

[0078] In this embodiment, the updated pheromone concentration is checked. If the updated pheromone concentration exceeds the preset upper and lower bounds, the updated pheromone concentration is adjusted to the default value; and the updated pheromone concentration table is persisted to storage.

[0079] Embodiment 2

[0080] As Figure 2 shown, this embodiment provides a distributed heterogeneous task scheduling system based on the ant colony algorithm. The system is based on a scheduler and an ant colony algorithm framework in a distributed heterogeneous task scenario, including:

[0081] An initialization module, used to initialize the parameters in the scheduling process and obtain the relevant information of each node and task instance in the distributed heterogeneous task scenario;

[0082] A node screening module, used to screen nodes that meet the scheduling conditions. When a new task instance is submitted to the cluster, the node screening module checks whether the resources of each node meet the requirements of the task instance, and takes the nodes that meet the requirements as optional nodes, and deletes the nodes that do not meet the conditions;

[0083] A node heuristic value calculation module, used to calculate the heuristic value of a node according to the usage of the CPU and memory resources of the node;

[0084] In this embodiment, the heuristic value reflects the current resource tension of the node. The lower the resource utilization rate of the node, the higher the heuristic value will be obtained. By analyzing the usage of resources such as the CPU and memory of the node, a heuristic value is generated for each node, specifically:

[0085]

[0086] In the formula, h represents the heuristic value of the node's resource utilization rate; CPU usage is the usage of the node's CPU; Memory usage is the usage of the node's memory; ∈ is a constant to prevent division by zero.

[0087] A node hybrid score calculation module, which is used to calculate the hybrid score of a node according to the heuristic value of the node and the pheromone concentration of the node;

[0088] The hybrid score of the node consists of two parts: one part is the heuristic value calculated based on resource utilization, which reflects the current resource tension of the node; the other part is the pheromone concentration of the node, which synthesizes the successful experience of historical scheduling and reflects the advantages of the node in long-term scheduling. The pheromone concentration of the node is used to simulate the preference of ants in path selection, and the hybrid score score of the node is expressed as:

[0089] score = (p α ) × (h β ) (2)

[0090] In the formula, p is the pheromone concentration of the node; h is the heuristic value of the node, and α and β represent weight parameters.

[0091] A node score normalization processing module, which is used to normalize the hybrid scores of all nodes; the specific normalization processing is as follows:

[0092] First, find the highest score maxScore and the lowest score minScore of all nodes;

[0093] Then, adjust the score of each node to the range of 0 - 100 according to non-linear compression. The non-linear compression is implemented through the Sigmoid function, which is used to suppress the influence of extreme values on the scheduling decision. First, determine the score range through the highest score maxScore and the lowest score minScore, and then normalize the score of each node and map it to the range of 0 to 1. Then, apply the Sigmoid function to perform non-linear compression on the normalized value, so that the score changes smoothly near the middle value and changes steeply at both ends, thereby effectively suppressing extreme values. Finally, map the compressed value to the range of 0 to 100 to obtain the final normalized score.

[0094] If the highest score maxScore and the lowest score minScore are the same, then set the scores of all nodes to the middle value of 50.

[0095] A binding module, which is used to select the node with the highest score as the target node of the task instance to be scheduled according to the normalized score, and schedule the task instance to this node; when a task instance is successfully bound to a node, the scheduler immediately obtains the current resource usage and task execution status of the node.

[0096] An update module, configured to calculate a reward value according to the resource usage of a node, update the pheromone concentration of the node in combination with the pheromone decay rate and the global average pheromone concentration, and check the updated pheromone concentration and persist the updated pheromone concentration table to storage.

[0097] Specifically:

[0098] p new = p old ×(1 - ρ)+Δτ*localWeight+globalAvg*globalWeight

[0099] In the formula, p new represents the pheromone concentration of the node after update; p old represents the pheromone concentration of the current node; ρ represents the pheromone decay rate; p old ×ρ simulates the volatilization of pheromone in nature; Δτ represents the contribution value of this scheduling; localWeight and globalWeight respectively represent the local pheromone weight and the global pheromone weight; globalAvg represents the average pheromone of all nodes.

[0100] In this embodiment, the update module described in this embodiment dynamically updates the pheromone concentration of a node according to the scheduling performance of the node, and realizes self-optimization and continuous improvement of the scheduling decision by simulating the behavior of ants releasing pheromone on the path.

[0101] The above embodiments and descriptions in the specification only illustrate the principles and the best embodiments of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed.

Claims

1. A distributed heterogeneous task scheduling method based on the ant colony algorithm, characterized in that, It includes the following steps: S1), construct a scheduler applicable to the distributed heterogeneous task scenario and integrate the ant colony algorithm; S2), initialize the parameters required by the scheduler and the ant colony algorithm, and obtain the relevant information of each node and task instance in the distributed heterogeneous task scenario; S3), when a new task instance is submitted to the cluster, the scheduler first checks whether the resources of each node meet the requirements of the task instance; S4), for the qualified nodes, calculate the heuristic value of their resource utilization rate; S5), combine the pheromone concentration and heuristic value of the node, and use the ant colony algorithm to calculate the mixed score of the node; S6), according to the normalized score, select the node with the highest score as the target node for the task instance to be scheduled, and schedule the task instance to this node; at the same time, calculate the reward value according to the resource usage of the node, and update the pheromone concentration of the node in combination with the pheromone decay rate and the global average pheromone concentration; S7), check the updated pheromone concentration and persist the updated pheromone concentration table to the storage.

2. The distributed heterogeneous task scheduling method based on the ant colony algorithm according to claim 1, wherein: In step S3), the scheduler calculates the total resource request of the task instance and compares it with the allocable resources of each node. The nodes that do not meet the resource requirements will be excluded from the optional nodes.

3. The distributed heterogeneous task scheduling method based on the ant colony algorithm according to claim 2, characterized in that: In step S4), the calculation expression of the heuristic value of the resource utilization rate of the node is: where h represents the heuristic value of the resource utilization of the node; CPU usage is the usage of the CPU of the node; Memory usage is the usage of the memory of the node; ∈ is a constant to prevent division by zero.

4. A distributed heterogeneous task scheduling method based on the ant colony algorithm according to claim 3, characterized in that: In step S5), the mixed score score of the node consists of two parts: one is the heuristic value calculated based on the resource utilization rate, and the heuristic value reflects the current resource tension of the node; the other is the pheromone concentration of the node. The pheromone concentration of the node is used to simulate the preference of ants in path selection. The mixed score score of the node is expressed as: score=(p α )×(h β ) (2) In the formula, p is the pheromone concentration of the node; h is the heuristic value of the node, and α and β represent weight parameters.

5. A distributed heterogeneous task scheduling method based on the ant colony algorithm according to claim 4, characterized in that: In step S6), after normalizing the mixed scores of all nodes to ensure the comparability of the mixed scores between different nodes, the specific normalization process is: S61), find the highest score maxScore and the lowest score minScore of all nodes; S62), adjust the scores of each node to the range of 0-100 according to the non-linear compression.

6. A distributed heterogeneous task scheduling method based on the ant colony algorithm according to claim 5, characterized in that: In step S6), if the highest score maxScore and the lowest score minScore are the same, set the scores of all nodes to the intermediate value 50.

7. A distributed heterogeneous task scheduling method based on the ant colony algorithm according to claim 4, characterized in that: In step S6), during the process of scheduling the task instance to the node, update the status information of the node to ensure the correct execution of the scheduling decision. When a task instance is successfully scheduled to the node, the scheduler immediately obtains the current resource usage and task execution status of the node.

8. A distributed heterogeneous task scheduling method based on the ant colony algorithm according to claim 4, characterized in that: In step S6), calculate the reward value according to the resource usage of the node, and update the pheromone concentration of the node in combination with the pheromone decay rate and the global average pheromone concentration. Specifically: p bew = p old × (1 - ρ) + Δτ * localWeight + globalAvg * globalWeight where p new represents the pheromone concentration after node update; p old represents the pheromone concentration of the current node; ρ represents the pheromone decay rate; p old ×ρ simulates the evaporation of pheromone in nature; Δτ represents the contribution value of this scheduling; localWeight and globalWeight represent the local pheromone weight and the global pheromone weight respectively; glpbalAvg represents the average value of pheromone of all nodes.

9. A distributed heterogeneous task scheduling method based on the ant colony algorithm according to claim 1, characterized in that: In step S7), check the updated pheromone concentration. If the updated pheromone concentration exceeds the preset upper and lower bounds, adjust the updated pheromone concentration to the default value; And persist the updated pheromone concentration table to the storage.

10. A distributed heterogeneous task scheduling system based on the ant colony algorithm, characterized in that, The system performs task scheduling by using the method described in any one of claims 1-9. The system is based on a scheduler and an ant colony algorithm framework in a distributed heterogeneous task scenario, and includes: An initialization module, which is used to initialize the parameters in the scheduling process and obtain the relevant information of each node and task instance in the distributed heterogeneous task scenario; A node screening module, which is used to screen the nodes that meet the scheduling conditions. When a new task instance is submitted to the cluster, the node screening module checks whether the resources of each node meet the requirements of the task instance, takes the nodes that meet the requirements as optional nodes, and deletes the nodes that do not meet the conditions; A node heuristic value calculation module, which is used to calculate the heuristic value of a node according to the usage of the CPU and memory resources of the node; A node hybrid score calculation module, which is used to calculate the hybrid score of a node according to the heuristic value of the node and the pheromone concentration of the node; A node score normalization processing module, which is used to perform normalization processing on the hybrid scores of all nodes; A binding module, which is used to select the node with the highest score as the target node for the task instance to be scheduled according to the normalized score, and bind the task instance to this node; An update module, which is used to calculate the reward value according to the resource usage of the node, update the pheromone concentration of the node in combination with the pheromone decay rate and the global average pheromone concentration, and check the updated pheromone concentration and persist the updated pheromone concentration table to the storage.

Images