Multi-level computing resource multi-objective dynamic scheduling method and system under algorithm network fusion

Abstract

The application belongs to the technical field of computer resource scheduling, and provides a multi-level computer resource multi-target dynamic scheduling method and system under algorithm-network fusion, comprising: obtaining a task request of a user; based on the obtained task request, constructing a global resource-network state distribution view; according to the parameters of the task request and the global resource-network state distribution view, taking minimizing task completion delay, optimizing load balancing and minimizing energy consumption as the target, generating a task-to-computing resource node allocation scheme and a task data transmission path planning scheme through a multi-target optimization scheduling model; according to the generated allocation scheme and path planning scheme, distributing the task to the corresponding computing resource node, and monitoring the task execution state and resource load in real time; when it is detected that the load of the computing node or the network link exceeds the preset threshold, triggering a task migration mechanism to complete the multi-target dynamic scheduling of the multi-level computing resource under algorithm-network fusion.

Description

Technical Field

[0001] This invention belongs to the field of computer resource scheduling technology, specifically relating to a multi-level, multi-objective dynamic scheduling method and system for computing resources under the integration of computing and network. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] With the development of the digital economy, computing power has become a core production factor. The convergence of computing and network resources, integrating multi-level computing and network resources such as edge computing, cloud computing, and supercomputing to build an integrated infrastructure, is an inevitable trend in the development of computing infrastructure. The advancement of national strategies such as the "Eastern Data, Western Computing" project has accelerated the implementation of this convergence. However, the current multi-level computing resource scheduling faces many challenges: existing scheduling schemes often focus on single-level resources, lacking full-level collaboration between edge computing, cloud computing, and supercomputing; scheduling decisions do not fully consider the strong coupling relationship between network and computing power, resulting in excessively high transmission latency; dynamic adaptability is weak, making it difficult to cope with fluctuations in resource load and changes in task requirements; and multi-objective optimization is unbalanced, failing to simultaneously meet requirements for latency, energy consumption, and load balancing.

[0004] While existing computing-network convergence architectures clearly define hierarchical divisions, they lack specific implementation schemes for dynamic scheduling algorithms. Cloud computing scheduling technologies ignore the impact of network latency, edge scheduling lacks heterogeneous resource adaptation mechanisms, and supercomputing scheduling fails to achieve deep collaboration with network resources. Research on scheduling for computing-network convergence is still in its early stages, applicable only to small-scale systems or single scenarios, lacking versatility and scalability. Therefore, there is an urgent need for a computing resource scheduling scheme that can achieve full-level collaboration, deep coupling of network and computing power, dynamic adaptation, and multi-objective optimization to support the efficient operation of computing-network convergence systems. Summary of the Invention

[0005] To address the aforementioned issues, this invention proposes a multi-level, multi-objective dynamic scheduling method and system for computing resources under the convergence of computing and network. Through multi-objective collaborative optimization of minimizing latency, load balancing, and minimizing energy consumption, it achieves deep collaboration and dynamic adaptation between computing power and network resources, effectively improving the scheduling efficiency, resource utilization, and overall robustness of the converged computing and network system under multi-objective constraints. This solves the problems of single scheduling objectives, low resource-task matching, slow algorithm convergence, and insufficient system stability in existing technologies.

[0006] According to some embodiments, the first solution of the present invention provides a multi-level computing resource multi-objective dynamic scheduling method under computing-network convergence, which adopts the following technical solution: A multi-level, multi-objective dynamic scheduling method for computing resources under the convergence of computing and network technologies includes: Obtain the user's task request; Based on the acquired task requests, construct a global resource-network status distribution view; Based on the parameters of the task request and the global resource-network status distribution view, with the goals of minimizing task completion latency, optimizing load balancing, and minimizing energy consumption, a multi-objective optimization scheduling model is used to generate an allocation scheme for tasks to computing resource nodes and a path planning scheme for task data transmission. Based on the generated allocation scheme and path planning scheme, tasks are distributed to the corresponding computing resource nodes, and the task execution status and resource load are monitored in real time. When the load of computing nodes or network links exceeds the preset threshold, the task migration mechanism is triggered to complete the multi-objective dynamic scheduling of multi-level computing resources under the computing-network convergence.

[0007] As a further technical limitation, the multi-objective optimization scheduling model adopts a multi-objective optimization algorithm based on a dual-population hybrid genetic-ant colony approach, including the following steps: Based on task latency constraints, tasks are divided into first priority, second priority, and third priority; Chromosomes are initialized using a three-dimensional encoding method of "task-resource-path" and two populations are constructed. The first population is initialized hierarchically according to task priority, while the second population is initialized randomly. The population is iteratively optimized through genetic operations, including selection operations based on non-dominated sorting and crowding calculation, crossover operations based on task priority segments, and differential mutation operations for low-priority tasks on nodes with excessive load. During the iteration process, the shortest path algorithm is used to dynamically replan the transmission paths of low-priority tasks in high-load links; An elite retention strategy is adopted, in which elite individuals within the population are retained in the early stage of iteration, and in the later stage of iteration, Pareto frontier individuals from the two populations are merged to form global elite individuals to participate in evolution. When the preset termination conditions are met, the optimal task-resource allocation mapping and transmission path planning scheme is output.

[0008] Furthermore, the first priority task is assigned to an edge computing node, the second priority task is assigned to a cloud computing node, and the third priority task is assigned to a supercomputing layer computing node.

[0009] Furthermore, during the hierarchical initialization process based on task priority, the first group uses an improved ant colony algorithm to reselect computing resource nodes for the tasks to be mutated, while during the random initialization process of the second group, the tasks to be mutated are randomly reallocated.

[0010] As a further technical limitation, the objective function of the multi-objective optimization scheduling model is a combination of a latency minimization function, a load balancing function, and an energy consumption minimization function; the latency minimization function comprehensively considers task access latency, network transmission latency, and computing node processing latency; the load balancing function simultaneously considers the load balancing of computing nodes and the load balancing of network links; and the energy consumption minimization function simultaneously considers the energy consumption of computing nodes and the energy consumption of network transmission.

[0011] As a further technical limitation, the triggered task migration mechanism includes: Identify compute nodes whose load exceeds the limit and tasks to be migrated; In computing nodes where the load is not exceeded, the suitability of each node to the task to be transferred is evaluated based on the weighted cosine similarity function. The computing node with the highest adaptability and that satisfies the constraint that the total task latency does not exceed its maximum latency is selected as the target node for task migration.

[0012] According to some embodiments, the second aspect of the present invention provides a multi-level computing resource multi-objective dynamic scheduling system under computing-network convergence, which adopts the following technical solution: A multi-level, multi-objective dynamic scheduling system for computing resources under the convergence of computing and network technologies includes: The acquisition module is configured to acquire user task requests; The building module is configured to construct a global resource-network status distribution view based on the acquired task requests; The generation module is configured to generate an allocation scheme for tasks to computing resource nodes and a path planning scheme for task data transmission based on the parameters of the task request and the global resource-network status distribution view, with the goals of minimizing task completion latency, optimizing load balancing and minimizing energy consumption, through a multi-objective optimization scheduling model. The scheduling module is configured to distribute tasks to corresponding computing resource nodes according to the generated allocation scheme and path planning scheme, and monitor the task execution status and resource load in real time. When the load of computing nodes or network links exceeds a preset threshold, the task migration mechanism is triggered to complete the multi-objective dynamic scheduling of multi-level computing resources under the convergence of computing and network.

[0013] According to some embodiments, a third aspect of the present invention provides a computer-readable storage medium, employing the following technical solution: A computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of the multi-level computing resource multi-objective dynamic scheduling method under the convergence of computing and network as described in the first aspect of the present invention.

[0014] According to some embodiments, the fourth aspect of the present invention provides an electronic device, which adopts the following technical solution: An electronic device includes a memory, a processor, and a program stored in the memory and running on the processor. When the processor executes the program, it implements the steps in the multi-level computing resource multi-objective dynamic scheduling method under the convergence of computing and network as described in the first aspect of the present invention.

[0015] According to some embodiments, the fifth aspect of the present invention provides a computer program product, which adopts the following technical solution: A computer program product includes software code, wherein the program in the software code performs the steps of the multi-level computing resource multi-objective dynamic scheduling method under the computing network convergence as described in the first aspect of the present invention.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention achieves multi-objective collaborative optimization of the computing-network convergence system by constructing an integrated optimization function encompassing latency, load balancing, and energy consumption, thus solving the problem of single-objective algorithms. Based on task priority and the heterogeneous characteristics of the three-layer resources, a hierarchical initialization strategy is designed, prioritizing the allocation of high-priority, low-latency tasks to edge nodes to improve task execution efficiency. A dual-population evolution mechanism is employed, combining segmented crossover, improved ant colony algorithm mutation, and a global elite retention strategy to accelerate algorithm convergence and avoid getting trapped in local optima. A task migration mechanism is introduced, dynamically unloading overloaded tasks through similarity evaluation and latency judgment, ensuring that node and link loads do not exceed limits and improving system robustness. It is applicable to a three-tiered heterogeneous computing power network environment of edge-cloud-supercomputing, meeting the diverse needs of tasks with different priorities and adapting to the development trend of computing-network convergence technology. Attached Figure Description

[0017] The accompanying drawings, which form part of this embodiment, are used to provide a further understanding of this embodiment. The illustrative embodiments and their descriptions are used to explain this embodiment and do not constitute an improper limitation of this embodiment.

[0018] Figure 1 This is a flowchart of the multi-level computing resource multi-objective dynamic scheduling method under computing-network integration in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of an architecture for a multi-level computing resource multi-objective dynamic scheduling method under computing-network integration in Embodiment 1 of the present invention; Figure 3 This is a detailed schematic diagram illustrating the steps of the multi-level computing resource multi-objective dynamic scheduling method under computing-network integration in Embodiment 1 of the present invention; Figure 4 This is a schematic diagram of another architecture of the multi-level computing resource multi-objective dynamic scheduling method under computing-network integration in Embodiment 1 of the present invention; Figure 5This is a schematic diagram of the computing network in Embodiment 1 of the present invention; Figure 6 This is a structural block diagram of the multi-level computing resource multi-objective dynamic scheduling system under the computing network integration in Embodiment 2 of the present invention. Detailed Implementation

[0019] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0020] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0021] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0022] In this invention, terms such as "upper," "lower," "left," "right," "front," "back," "vertical," "horizontal," "side," and "bottom" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are used only to facilitate the description of the structural relationships of the various components or elements of this invention and do not specifically refer to any component or element in this invention. They should not be construed as limiting the invention.

[0023] In this invention, terms such as "fixed connection," "connected," and "linked" should be interpreted broadly, indicating a fixed connection, an integral connection, or a detachable connection; a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can determine the specific meaning of these terms in this invention based on the specific circumstances, and they should not be construed as limitations on the invention.

[0024] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0025] Example 1 Embodiment 1 of this invention introduces a multi-level, multi-objective dynamic scheduling method for computing resources under the convergence of computing and network.

[0026] like Figure 1 The method for multi-level, multi-objective dynamic scheduling of computing resources under the convergence of computing and network, as shown, includes: Obtain the user's task request; Based on the acquired task requests, construct a global resource-network status distribution view; Based on the parameters of the task request and the global resource-network status distribution view, with the goals of minimizing task completion latency, optimizing load balancing, and minimizing energy consumption, a multi-objective optimization scheduling model is used to generate an allocation scheme for tasks to computing resource nodes and a path planning scheme for task data transmission. Based on the generated allocation scheme and path planning scheme, tasks are distributed to the corresponding computing resource nodes, and the task execution status and resource load are monitored in real time. When the load of computing nodes or network links exceeds the preset threshold, the task migration mechanism is triggered to complete the multi-objective dynamic scheduling of multi-level computing resources under the computing-network convergence.

[0027] like Figure 2 , Figure 3 and Figure 4 As shown, this embodiment achieves collaborative work and functional integration through close cooperation between various layers. The specific process is as follows: Users initiate task requests through the application-layer user service platform. The requests clearly include core parameters such as the computing power required for the task, the amount of data, the latency constraints, and priority preferences. After receiving a task request, the SDN control platform combines the resource management module and the network awareness module to collect real-time data on the global computing resource status and network resource status of the edge layer, cloud layer, and supercomputing layer, and constructs a global resource-network status distribution view. The scheduling module takes into account task priority, resource heterogeneity, and network link status, and generates the optimal task-resource allocation scheme and transmission path planning with the goals of "minimizing latency, balancing load, and minimizing energy consumption". Based on the scheduling decision results, the execution control module completes the accurate distribution of tasks, node deployment and dynamic execution; it monitors the task running status and node / link load changes in real time, and triggers the task migration mechanism when the node or link load exceeds the limit.

[0028] It should be noted that the resource model in this embodiment is as follows: Define the resource set of the computer-network converged system as R= ,in For edge layer resource collection , Cloud resource collection , For supercomputing layer resource collection Each resource node The state can be represented as = { , , , } ,in The remaining computing power (GFLOPS) This represents the maximum computing power of the node. The available network bandwidth (Mbps) and the power consumption of a computing node in processing a task are related to the amount of task and the parameters of the device itself. This represents the energy coefficient of the computing node's power consumption in processing tasks. Represents the node's static power. Load rate (%) This indicates the maximum load capacity of the node.

[0029] It should be noted that the task model in this embodiment is as follows: Define the task set as T={ }, each task The requirement can be represented as a tuple ,in This is a quantification of the computing resources required for the task. For the size of the data, The latency requirement is in milliseconds (ms). In addition, the task priority is calculated based on the latency requirement; the lower the latency requirement, the higher the priority.

[0030] It should be noted that the network model in this embodiment is: The computing nodes use dynamic data transmission links, and there may be multiple routes from one resource node to another. Assuming that the data transmission rate of the dynamic links is perceptible, the entire network architecture can be constructed as an undirected connected graph. , where V ;Right now

[0031]

[0032]

[0033] Where W represents the set of network links, and M represents the total number of resource nodes. This indicates the maximum capacity of the link. Indicates the remaining bandwidth of the current network link. For transmission energy consumption coefficient, This indicates the load rate of the link. To meet the load limit, the three-layer computing resource network is distributed as follows: Figure 5 As shown.

[0034] The multi-level resource scheduling problem can be described by the scheduling mapping function formula. ; making each task Assigned to resource nodes that meet their needs The goal is to find the optimal transmission path p and simultaneously optimize the system objective function.

[0035] Considering the multi-objective optimization requirements of the computer-network converged system, the following objective function is constructed: (1) Minimize task completion delay: Task latency includes access latency Transmission delay Computation delay Total latency ; Access latency refers to the time it takes for a terminal device to generate a task and upload it to an edge node via a wireless network. It primarily depends on the amount of data in the task and the uplink speed. Therefore, access latency can be expressed as: ; in, Given the uplink rate (in bits per second), the wireless transmission link from the terminal device to the edge node experiences signal attenuation, which is derived using Shannon's formula as follows: ; in, It is the channel bandwidth. It is transmission power. and These are the channel gain and the Gaussian white noise power, respectively.

[0036] When a task is transmitted along a path P from one routing node or resource node to another resource node, the transmission delay is expressed as: ; ; The system average latency target is .

[0037] (2) Load balancing computing nodes The load can be expressed as ; in, Represents a node The total amount of computing resources used by the received tasks. This means that a computing node can only handle computing tasks that do not exceed its own computing resources.

[0038] The standard deviation is used to evaluate the load balancing performance of all computing nodes, i.e. ; Where M represents the total number of computing nodes, This represents the average load across all computing nodes.

[0039] Network Link The load can be expressed as ; in, It refers to the total amount of task data traversed by network link n. This means that a network path can only be loaded up to its own transmission capacity.

[0040] The standard deviation is also used to evaluate the load balancing performance of all network links, i.e. ; in, This represents the average load across all computing nodes.

[0041] Therefore, the load balancing of the entire system can be expressed as:

[0042] (3) Minimize energy consumption System energy consumption includes resource node energy consumption. Network transmission energy consumption Here is set The total energy consumption target is: ; ; ; The scheduling process must satisfy the following constraints: The scheduling problem can be described as follows: for a set of tasks, select the set of nodes in a supercomputing cloud-edge heterogeneous computing network and adjust the transmission path of each task to obtain an optimal system operating state that minimizes the total cost of latency, energy consumption, and load balancing. Then, under the overall constraints, the multi-objective optimization problem can be expressed as: ; ； ； ； ； ； (1) This means that the computing power required for task k must not exceed the remaining computing power of the allocated resource nodes; (2) This means that the total amount of data transmitted by task k on network link n must not exceed the maximum capacity of that link; (3) This means that the total delay of task k must not exceed the maximum tolerable delay of that task; (4) and This indicates that individual resource nodes and network links must not exceed their load limits to prevent overload.

[0043] To address the multi-objective optimization requirements of computing network fusion scheduling (delay minimization, utilization maximization, and energy consumption minimization), this invention provides an improved non-dominated sorting genetic algorithm (Dual-Population Hybrid Genetic-AntColony Optimization Algorithm) and a dual-population hybrid genetic-ant colony multi-objective optimization algorithm (DP-HGA).

[0044] Before the algorithm starts, tasks are divided into three priorities: low, medium, and high, based on their latency requirements.

[0045] A three-dimensional coding method of "task-resource-path" is adopted, with each chromosome corresponding to a scheduling scheme. The coding structure is as follows:

[0046] in, For the task gene fragments, For the allocated resource nodes, This corresponds to the network path. For example, chromosome. Indicates task Assigned to edge nodes ,path ,Task Assigned to cloud nodes ,path .

[0047] The algorithm employs a dual-population mechanism. Population 1 uses a hierarchical initialization mechanism, where high-priority tasks have low latency requirements and are more likely to be assigned to nearby edge nodes, medium-priority tasks are more likely to be assigned to the cloud, and low-priority tasks are more likely to be assigned to the supercomputing layer. Population 2 uses a random initialization mechanism.

[0048] After each iteration, elite individuals are exchanged between the two populations to accelerate convergence.

[0049] The Floyd algorithm is used to calculate the shortest path between nodes in the chromosome network. The weight calculation formula is as follows:

[0050] in + .

[0051] Floyd algorithm flow: Initialize the distance matrix D, where D[i][j] represents the direct distance from node i to node j. If there is a direct connection between i and j, then D[i][j] is the weight of the edge; if i and j are not directly connected, then D[i][j] is infinity; D[i][i] = 0. For each node k (as an intermediate node), iterate through all node pairs (i,j); If passing through k can shorten the path from i to j, i.e., D[i][j] > D[i][k] + D[k][j], then update D[i][j] = D[i][k] + D[k][j]. When all nodes are considered as intermediate nodes, the value in matrix D is the final shortest path distance.

[0052] After chromosome initialization, the distribution density of individuals in the target space within the multi-level non-dominated solution set is further calculated: ; in, Chromosomes The value under the k-th objective space function, Chromosomes The value under the k-th objective space function, This represents the maximum value of the entire population under k objective functions. This represents the minimum value of the entire population under k objective functions.

[0053] This embodiment employs a segmented crossover operation, where chromosome segments are divided into high, medium, and low segments according to task priority. One segment from each of the two parent chromosomes is randomly selected for exchange to generate offspring.

[0054] In this embodiment, a lower-priority task from several computationally busy nodes in the chromosome allocation scheme is selected for mutation. The two populations employ different mutation methods: population two uses random redistribution, while population one uses a small-scale ant colony algorithm.

[0055] Initialize pheromones: ; ; Where i represents the task to be mutated, and j represents the computing node. This indicates the load rate of the computing node. The remaining computing power of the node. The computing power required for the task. This is a penalty factor, replacing transmission cost.

[0056] The formula for ant k in city i to select the next city j according to the pseudo-random ratio rule is as follows: ; in, It is a random number controlled within the range [0,1] and conforms to a normal distribution. q is a constant belonging to the interval [0,1]. When q≤q0, the next node is selected from all selectable nodes. The maximum value is calculated; conversely, when q > q0, the probability of choosing city j from city i is calculated using the roulette wheel selection strategy C. probability The definition is as follows: ; ; in, It's Ant Search. The set of subsequent tasks at that time This is a heuristic function that takes into account factors such as computational latency, load rate, and energy consumption. This is the penalty coefficient for transmission costs.

[0057] Pheromone Update. To prevent the pheromone on each path from accumulating indefinitely, the pheromone on each path is continuously updated using the pheromone evaporation coefficient ρ until the optimal number of iterations is reached, thus obtaining the optimal task scheduling and resource allocation scheme.

[0058] ; in, Let m ants be the number of ants left on the path in this search. The pheromone increment represents the sum of the fitness of all ants that pass through this path.

[0059] ; ; It is important to note that, considering the complexity of the algorithm, crossover and mutation operations do not involve changes to the gene segments involved in path planning within the chromosome.

[0060] After crossover and mutation, the new population is used to plan the transmission paths of lower priority tasks in high-load links using the Dijkstra algorithm.

[0061] The specific process of Dijkstra's algorithm is as follows: (1) Initialize the distance array dist[], set the distance of the source node s to 0, set the distance of other nodes to infinity, and mark the source node as the current node; (2) Create a set (or priority queue) to store unvisited nodes and put all nodes into it; (3) When the set of unvisited nodes is not empty, select the node u with the smallest distance from it (in the first loop, u is the source node s). (4) Mark node u as visited (remove it from the unvisited set); (5) For each neighbor node v of node u, if the distance from u to v is shorter than the distance recorded in the current record, update dist[v] = dist[u] + w(u, v) and update the predecessor node of node v to u; (6) Repeat steps (3) to (5) until the target node has been visited.

[0062] In this embodiment, individuals in the first-layer Pareto front are selected as elite individuals. In the early stages, only a single elite individual is retained in each generation of the two populations to participate in the next generation of evolution, allowing the algorithm to fully explore the solution space and preventing the algorithm from getting trapped in local optima. In the later stages of the algorithm, a fast non-dominated sort is performed on the individuals in the first-layer Pareto front of the two populations, that is, all elite individuals in the two populations are sorted. The obtained first-layer Pareto front is called the global elite, and the two populations share the global elite to participate in the next generation of evolution.

[0063] In the early stages, the fast non-dominated sorting algorithm sorts the two populations separately. If the global elite remains unchanged for several consecutive generations or the maximum number of iterations is reached, the algorithm terminates.

[0064] When a node's load rate exceeds the limit, a task migration mechanism is triggered. This mechanism selects nodes with lower load rates for offloading. Specifically: (1) Traverse available servers: The system identifies servers whose load has not yet exceeded the limit; (2) Similarity Assessment: A weighted cosine similarity function is used to determine whether the target server is suitable for receiving tasks. The similarity score is expressed as: ; (3) Delay Judgment: Task migration depends on ensuring that the execution time of the task on the target server does not exceed its maximum delay constraint. Specifically, the conditions for task migration can be expressed as: ; (4) Select the server with the highest score for migration.

[0065] This embodiment achieves multi-objective collaborative optimization of the computing-network convergence system by constructing an integrated optimization function encompassing latency, load balancing, and energy consumption, thus addressing the issue of single-objective algorithms in traditional systems. A hierarchical initialization strategy is designed based on task priority and the heterogeneous characteristics of the three-layer resources, prioritizing the allocation of high-priority, low-latency tasks to edge nodes to improve task execution efficiency. A dual-population evolution mechanism is employed, combining segmented crossover, improved ant colony algorithm mutation, and a global elite retention strategy to accelerate algorithm convergence and avoid getting trapped in local optima. A task migration mechanism is introduced, dynamically unloading overloaded tasks through similarity evaluation and latency judgment, ensuring that node and link loads do not exceed limits and improving system robustness. It is suitable for edge-cloud-supercomputing three-tier heterogeneous computing power network environments, meeting the diverse needs of tasks with different priorities and adapting to the development trend of computing-network convergence technology.

[0066] Example 2 Embodiment 2 of the present invention introduces a multi-level computing resource multi-objective dynamic scheduling system under the integration of computing and network.

[0067] like Figure 6 The system shown is a multi-level, multi-objective dynamic scheduling system for computing resources under the convergence of computing and network, comprising: The acquisition module is configured to acquire user task requests; The building module is configured to construct a global resource-network status distribution view based on the acquired task requests; The generation module is configured to generate an allocation scheme for tasks to computing resource nodes and a path planning scheme for task data transmission based on the parameters of the task request and the global resource-network status distribution view, with the goals of minimizing task completion latency, optimizing load balancing and minimizing energy consumption, through a multi-objective optimization scheduling model. The scheduling module is configured to distribute tasks to corresponding computing resource nodes according to the generated allocation scheme and path planning scheme, and monitor the task execution status and resource load in real time. When the load of computing nodes or network links exceeds a preset threshold, the task migration mechanism is triggered to complete the multi-objective dynamic scheduling of multi-level computing resources under the convergence of computing and network.

[0068] The detailed steps are the same as those of the multi-level computing resource multi-objective dynamic scheduling method under computing network integration provided in Example 1, and will not be repeated here.

[0069] Example 3 Embodiment 3 of the present invention provides a computer-readable storage medium.

[0070] A computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of the multi-level computing resource multi-objective dynamic scheduling method under computing-network convergence as described in Embodiment 1 of the present invention.

[0071] The detailed steps are the same as those of the multi-level computing resource multi-objective dynamic scheduling method under computing network integration provided in Example 1, and will not be repeated here.

[0072] Example 4 Embodiment 4 of the present invention provides an electronic device.

[0073] An electronic device includes a memory, a processor, and a program stored in the memory and running on the processor. When the processor executes the program, it implements the steps in the multi-level computing resource multi-objective dynamic scheduling method under computing network convergence as described in Embodiment 1 of the present invention.

[0074] The detailed steps are the same as those of the multi-level computing resource multi-objective dynamic scheduling method under computing network integration provided in Example 1, and will not be repeated here.

[0075] Example 5 Embodiment 5 of the present invention provides a computer program product.

[0076] A computer program product includes software code, wherein the program in the software code executes the steps of the multi-level computing resource multi-objective dynamic scheduling method under the convergence of computing and network as described in Embodiment 1 of the present invention.

[0077] The detailed steps are the same as those of the multi-level computing resource multi-objective dynamic scheduling method under computing network integration provided in Example 1, and will not be repeated here.

[0078] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0079] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0080] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0081] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0082] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0083] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

[0084] The above description is merely a preferred embodiment of this practice and is not intended to limit the scope of this practice. Various modifications and variations can be made to this practice by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of this practice should be included within the protection scope of this practice.

Claims

1. A multi-level, multi-objective dynamic scheduling method for computing resources under the convergence of computing and network, characterized in that, include: Obtain the user's task request; Based on the acquired task requests, construct a global resource-network status distribution view; Based on the parameters of the task request and the global resource-network status distribution view, with the goals of minimizing task completion latency, optimizing load balancing, and minimizing energy consumption, a multi-objective optimization scheduling model is used to generate an allocation scheme for tasks to computing resource nodes and a path planning scheme for task data transmission. Based on the generated allocation scheme and path planning scheme, tasks are distributed to the corresponding computing resource nodes, and the task execution status and resource load are monitored in real time. When the load of computing nodes or network links exceeds the preset threshold, the task migration mechanism is triggered to complete the multi-objective dynamic scheduling of multi-level computing resources under the computing-network convergence.

2. The method for multi-level, multi-objective dynamic scheduling of computing resources under the convergence of computing and network as described in claim 1, characterized in that, The multi-objective optimization scheduling model adopts a multi-objective optimization algorithm based on a dual-population hybrid genetic-ant colony approach, including the following steps: Based on task latency constraints, tasks are divided into first priority, second priority, and third priority; Chromosomes are initialized using a three-dimensional encoding method of "task-resource-path" and two populations are constructed. The first population is initialized hierarchically according to task priority, while the second population is initialized randomly. The population is iteratively optimized through genetic operations, including selection operations based on non-dominated sorting and crowding calculation, crossover operations based on task priority segments, and differential mutation operations for low-priority tasks on nodes with excessive load. During the iteration process, the shortest path algorithm is used to dynamically replan the transmission paths of low-priority tasks in high-load links; An elite retention strategy is adopted, in which elite individuals within the population are retained in the early stage of iteration, and in the later stage of iteration, Pareto frontier individuals from the two populations are merged to form global elite individuals to participate in evolution. When the preset termination conditions are met, the optimal task-resource allocation mapping and transmission path planning scheme is output.

3. The method for multi-level, multi-objective dynamic scheduling of computing resources under the convergence of computing and network as described in claim 2, characterized in that, The first priority task is assigned to the edge computing node, the second priority task is assigned to the cloud computing node, and the third priority task is assigned to the supercomputing layer computing node.

4. The method for multi-level, multi-objective dynamic scheduling of computing resources under the convergence of computing and network as described in claim 2, characterized in that, During the hierarchical initialization process based on task priority, the first swarm uses an improved ant colony algorithm to reselect computing resource nodes for the tasks to be mutated. During the random initialization process of the second swarm, the tasks to be mutated are randomly redistributed.

5. The method for multi-level, multi-objective dynamic scheduling of computing resources under the convergence of computing and network as described in claim 1, characterized in that, The objective function of the multi-objective optimization scheduling model is a combination of a latency minimization function, a load balancing function, and an energy consumption minimization function. The latency minimization function comprehensively considers task access latency, network transmission latency, and computing node processing latency. The load balancing function simultaneously considers the load balancing of computing nodes and the load balancing of network links. The energy consumption minimization function simultaneously considers the energy consumption of computing nodes and the energy consumption of network transmission.

6. The method for multi-level, multi-objective dynamic scheduling of computing resources under the convergence of computing and network as described in claim 1, characterized in that, The triggered task migration mechanisms include: Identify compute nodes whose load exceeds the limit and tasks to be migrated; In computing nodes where the load is not exceeded, the suitability of each node to the task to be transferred is evaluated based on the weighted cosine similarity function. The computing node with the highest adaptability and that satisfies the constraint that the total task latency does not exceed its maximum latency is selected as the target node for task migration.

7. A multi-level, multi-objective dynamic scheduling system for computing resources under the convergence of computing and network, characterized in that, include: The acquisition module is configured to acquire user task requests; The building module is configured to construct a global resource-network status distribution view based on the acquired task requests; The generation module is configured to generate an allocation scheme for tasks to computing resource nodes and a path planning scheme for task data transmission based on the parameters of the task request and the global resource-network status distribution view, with the goals of minimizing task completion latency, optimizing load balancing and minimizing energy consumption, through a multi-objective optimization scheduling model. The scheduling module is configured to distribute tasks to corresponding computing resource nodes according to the generated allocation scheme and path planning scheme, and monitor the task execution status and resource load in real time. When the load of computing nodes or network links exceeds a preset threshold, the task migration mechanism is triggered to complete the multi-objective dynamic scheduling of multi-level computing resources under the convergence of computing and network.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the multi-level computing resource multi-objective dynamic scheduling method under the convergence of computing and network as described in any one of claims 1-6.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the program, it implements the steps of the multi-level computing resource multi-objective dynamic scheduling method under the computing-network convergence as described in any one of claims 1-6.

10. A computer program product, comprising software code, characterized in that, The program in the software code executes the steps of the multi-level computing resource multi-objective dynamic scheduling method under the computing network integration as described in any one of claims 1-6.

Images