Intelligent agent-based cloud task scheduling method and device, equipment and medium

Abstract

The application provides an agent-based cloud task scheduling method, device, equipment and medium, which can be applied to the technical field of complex cloud task scheduling. The cloud task scheduling method comprises the following steps: in response to a task request, obtaining task characteristics and user preferences; based on the task characteristics and the user preferences, generating an action plan for task scheduling by using an agent; according to the action plan, sequentially calling each tool by using the agent to generate a task scheduling scheme, comprising: generating a task description based on the task characteristics and the user preferences; collecting resource cost, resource usage information and link condition information to generate a network topology graph; predicting a task load fluctuation condition by using a time sequence prediction model; selecting a link node based on the network topology graph and the task load fluctuation condition; selecting a task scheduling method based on the task type and the user preferences; and generating a task scheduling scheme by using the task scheduling method based on the task description and the link node.

Description

Technical Field

[0001] The present invention relates to the technical field of complex cloud task scheduling, and more specifically to an agent-based cloud task scheduling method, device, equipment and medium. Background Art

[0002] In today's digital age, cloud computing has become a core technology driving the development of various industries. With the continuous expansion of application scenarios and the increasing complexity of business needs, the cloud computing environment has gradually evolved into a complex ecosystem interwoven with on-premises clouds, public clouds, and edge clouds. On-premises clouds, with their advantage of being deployed within enterprises, can provide enterprises with highly customized resource services, ensuring data security and controllability. This makes them particularly suitable for industries with strict data privacy requirements, such as finance and healthcare. However, their resource scale is often limited by an enterprise's own hardware investment, making resource expansion relatively limited when responding to large-scale, sudden tasks. Public clouds, with their powerful resource pools, flexible resource allocation, and convenient pay-as-you-go models, have attracted a large number of cost-sensitive small and medium-sized enterprises with volatile resource demands. However, in public cloud environments, multi-tenant resource sharing can lead to resource contention and uncertain service quality. The emergence of edge clouds effectively addresses the real-time processing shortcomings of on-premises and public clouds. Located close to data sources and users, they can rapidly process data at the edge of the network, significantly reducing data transmission latency. They play a key role in areas with extremely high real-time response requirements, such as the Internet of Things, real-time video processing, and autonomous driving. However, edge cloud node resources are relatively scattered, resource management is difficult, and the resource capacity of a single node is limited.

[0003] Faced with such a complex and diverse cloud environment, achieving unified resource management and efficient task scheduling has become a key challenge in improving cloud computing performance. On-premises, public, and edge clouds utilize a diverse range of resources, encompassing computing resources (such as CPUs and GPUs), storage resources (such as hard drives and memory), and network resources (such as bandwidth and IP addresses). These resources not only differ in physical location and performance parameters, but their usage is also constantly changing. For example, CPU utilization fluctuates in real time with task load, and memory usage constantly adjusts depending on the running of different applications. Traditional resource management models often operate in a fragmented manner, lacking a unified coordination mechanism across cloud platforms. This leads to both idle and wasted resources, making it difficult to improve resource utilization efficiency.

[0004] At the same time, complex cloud environments present a wide variety of tasks, each exhibiting unique characteristics in terms of resource requirements, task structure, and scheduling objectives. Some tasks are urgent and non-interruptible, placing extremely high demands on the performance of computing and network resources. These tasks require ensuring sufficient resources are deployed in the shortest possible time to achieve the task objectives, effectively completing the task as quickly as possible and ensuring smooth progress. Scientific research tasks, on the other hand, are computationally intensive and have long task cycles. They require extensive computing resources to process massive amounts of data over extended periods. These tasks are typically interruptible under certain conditions, and their scheduling objectives often prioritize minimizing costs. Therefore, optimizing resource allocation to reduce overall costs through rationally leveraging resource pricing differences across cloud platforms is crucial. Autonomous driving tasks represent a class of scenarios with extremely stringent requirements for real-time and reliability. Vehicles must process large amounts of sensor data in real time to make accurate driving decisions, such as acceleration, braking, and steering. Therefore, autonomous driving tasks require access to edge resources close to the vehicle, leveraging the low latency of the edge cloud to rapidly process data and minimize latency. Furthermore, redundant backups and other technical measures ensure high system reliability, ensuring safe and stable operation in a variety of complex road conditions and environments.

[0005] In summary, the diversity of complex cloud environments and the complexity of task types have posed unprecedented challenges to cloud task scheduling methods and systems. An innovative solution is urgently needed to achieve efficient resource utilization and precise task scheduling to meet the growing digital needs of various industries.

[0006] Faced with complex cloud environments and diverse task types, efficient task scheduling is crucial for achieving efficient resource utilization and meeting user needs. As cloud tasks continue to grow in complexity, traditional task scheduling approaches face numerous challenges. Existing cloud scheduling algorithms often focus on a single objective, making it difficult for single scheduling methods to fully account for the diverse nature of tasks and the complex resource environment. Furthermore, a few multi-objective scheduling algorithms seek to balance multiple objectives, failing to meet the unique requirements of diverse tasks. Furthermore, for tasks that incorporate multimodal information, existing scheduling methods are unable to fully leverage this information to make optimal scheduling decisions. Therefore, it is necessary to design a scheduling method that can accommodate diverse tasks in complex cloud environments. Summary of the Invention

[0007] (1) Technical issues to be resolved

[0008] In view of the above problems, the present invention provides a cloud task scheduling method, device, equipment and medium based on intelligent agent, which comprehensively considers the complex cloud environment and various types of tasks, and effectively realizes the comprehensive consideration and precise scheduling of multiple types of tasks through innovative intelligent agent architecture and multi-model fusion strategy, assisting users or cloud application providers to better schedule tasks according to demand, lowering the threshold for cloud resource use, improving the efficiency of cloud resource use, and promoting the promotion and application of cloud computing.

[0009] (2) Technical solution

[0010] In response to the above technical problems, embodiments of the present invention propose an agent-based cloud task scheduling method, apparatus, device, and medium.

[0011] According to a first aspect of the present invention, there is provided a cloud task scheduling method based on an intelligent agent, comprising: in response to a task request, obtaining task characteristics and user preferences; based on the task characteristics and user preferences, generating a task scheduling action plan using an intelligent agent, wherein the intelligent agent comprises a trained large model, a configuration module, a planning module, a memory module and a plurality of tools, the tools comprising a task description generation tool, a node information acquisition tool, a load fluctuation prediction tool, a node selection tool, a task scheduling scheme generation tool, and a task execution tool; according to the action plan, sequentially calling each tool using an intelligent agent to generate a task scheduling scheme, wherein, according to the action plan, sequentially calling each tool using an intelligent agent to generate a task scheduling scheme comprises: based on the task characteristics and user preferences, using an intelligent agent to call a task description generation tool to determine the task type, and performing an analysis of the task characteristics, user preferences and Task types are integrated and task descriptions are generated in a unified data format; the intelligent agent is used to call the node information acquisition tool to collect resource costs, resource usage information and link status information, and a network topology diagram is generated based on graph theory and visualization technology; the intelligent agent is used to call the load fluctuation prediction tool to predict the task load fluctuation based on the trained time series prediction model; the intelligent agent is used to call the node selection tool to select link nodes based on the network topology diagram and task load fluctuation; the intelligent agent is used to call the task scheduling plan generation tool to select the task scheduling method based on the task type and user preference, and generate a task scheduling plan based on the task description and the selected link nodes using the selected task scheduling method. Task characteristics include task resource requirements, task execution time requirements, task priority, task interruptibility and task dependencies. In some exemplary embodiments, the big model includes a big model suitable for natural language processing and multimodal information processing, and the method of training the big model includes: obtaining training data, the training data including data of various task characteristics; using the training data to train the big model to adjust the parameters of the big model to complete the training of the big model, wherein the big model can process multimodal input data, and the input data includes text, time series data, and image modality data.

[0012] In some exemplary embodiments, the timing prediction model includes a long short-term memory network, and the method for training the timing prediction model includes: obtaining historical task load data and marking the time points of key events, wherein the task load data includes traffic data and hardware status data, and the key events include traffic peaks and hardware failures; using the historical task load data to train the timing prediction model, and adjusting the timing prediction model so that the timing prediction model can predict task load fluctuations.

[0013] In some exemplary embodiments, the task scheduling method includes a reinforcement learning scheduling method, a heuristic algorithm scheduling method, a meta-heuristic algorithm scheduling method, and a rule-based scheduling algorithm, wherein the method for training the reinforcement learning scheduling method includes: setting a reward function based on the reinforcement learning scheduling method, the reward function is used to reflect the pros and cons of the scheduling strategy; designing a state space and an action space, the state space includes the resource requirements of the task, whether it can be interrupted, and the resource status of the current cloud environment, and the action space includes different scheduling decisions; based on the reward function, the state space, and the action space, using the reinforcement learning algorithm, the reinforcement learning scheduling method is trained in a simulated cloud task environment so that the reinforcement learning scheduling method learns the optimal scheduling strategy; and / or, the method for training the heuristic algorithm scheduling method includes: based on the task characteristics The method of training a genetic algorithm model comprises: defining a chromosome encoding method and encoding a scheduling scheme as a chromosome; setting a fitness function and evaluating the pros and cons of a scheduling scheme corresponding to each chromosome; continuously evolving chromosomes through genetic operations such as selection, crossover, and mutation to obtain a better scheduling scheme, thereby training a genetic algorithm model suitable for cloud task scheduling; and / or the method of training a rule-based scheduling algorithm comprises: formulating scheduling rules based on domain knowledge and scheduling requirements; applying the scheduling rules in actual task scenarios, and continuously adjusting and improving the scheduling rules based on feedback.

[0014] In some exemplary embodiments, the nodes in the network topology diagram represent cloud resource nodes, and the edges represent the links between the nodes; the network topology diagram includes the bandwidth, delay, and packet loss rate of the links; a dynamic update mechanism is adopted to ensure that the network topology diagram reflects the actual status of the cloud environment in real time.

[0015] In some exemplary embodiments, selecting link nodes based on the network topology and task load fluctuations includes: analyzing the bandwidth and delay information of each link based on the network topology using a trained large model; identifying key information of the link, including reliability and transmission efficiency, based on the bandwidth and delay information; obtaining initial link nodes based on resource costs and task load fluctuations; and further screening the initial link nodes based on task characteristics to obtain the final link nodes.

[0016] In some exemplary embodiments, it also includes: allocating the task scheduling plan to the selected cloud node for execution; continuously monitoring the execution status of the task, wherein, when an abnormal situation occurs during the task execution, taking corresponding response measures, including: recording abnormal information; selecting a corresponding processing method according to the severity and type of the abnormality; and feeding back the abnormal information to the task scheduling agent to adjust and optimize the task scheduling plan.

[0017] According to the second aspect of the present invention, there is provided a cloud task scheduling device based on an intelligent agent, comprising: an acquisition module for acquiring task characteristics and user preferences in response to a task request; a planning module for generating a task scheduling action plan using an intelligent agent based on the task characteristics and user preferences, wherein the intelligent agent comprises a trained large model, a configuration module, a planning module, a memory module and a plurality of tools, the tools comprising a task description generation tool, a node information acquisition tool, a load fluctuation prediction tool, a node selection tool, a task scheduling scheme generation tool and a task execution tool; a generation module for sequentially calling various tools using an intelligent agent according to an action plan to generate a task scheduling scheme, wherein sequentially calling various tools using an intelligent agent according to an action plan to generate a task scheduling scheme comprises: based on the task characteristics and user preferences, using an intelligent agent to call a task description generation tool to determine the task type, and The intelligent agent is used to call the node information acquisition tool to collect resource costs, resource usage information and link status information, and generate a network topology diagram based on graph theory and visualization technology; the intelligent agent is used to call the load fluctuation prediction tool to predict the task load fluctuation based on the trained time series prediction model; the intelligent agent is used to call the node selection tool to select link nodes based on the network topology diagram and task load fluctuation; the intelligent agent is used to call the task scheduling plan generation tool to select the task scheduling method based on the task type and user preferences, and generate a task scheduling plan based on the task description and the selected link nodes using the selected task scheduling method. The task characteristics include the task resource requirements, task execution time requirements, task priority, task interruptibility and task dependencies.

[0018] According to a third aspect of the present invention, an electronic device is provided, comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the above method.

[0019] The fourth aspect of the present invention further provides a computer-readable storage medium having a computer program or instructions stored thereon, which implements the steps of the above method when the computer program or instructions are executed by a processor.

[0020] (3) Beneficial effects

[0021] It can be seen from the above technical solutions that the agent-based cloud task scheduling method, apparatus, device, and medium provided by the embodiments of the present invention have at least the following beneficial effects:

[0022] (1) By fine-tuning the large model to construct a task scheduling agent, the flexible decision-making of the agent and the collaborative work between various tools can realize the automation of the entire process from data collection, task scheduling to task execution, reduce manual intervention and improve scheduling efficiency.

[0023] (2) Based on the big model, the system automatically selects the appropriate scheduling method according to the task type, so that it can adapt to the task rules in different fields, improve the versatility and flexibility of the scheduling system, and can cope with diverse and complex task scenarios. At the same time, it has strong compatibility with complex cloud scenarios such as heterogeneous clouds, distributed clouds, federated clouds, hybrid clouds, and edge clouds. It can flexibly integrate the resource pools of different cloud service providers to adapt to the complex IT architecture needs of enterprises.

[0024] (3) It supports multiple modal inputs such as text, time series data, and images, and can fully capture task characteristics and environmental information, providing richer contextual information for scheduling decisions, thereby improving the comprehensiveness and accuracy of decisions.

[0025] (4) Combining large models, time series prediction models, reinforcement learning scheduling methods, heuristic scheduling methods, meta-heuristic scheduling methods and rule-based scheduling methods to achieve complementary advantages of different scheduling strategies, adapt to the needs of different types of tasks, and ensure the diversity and efficiency of scheduling decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The above contents and other objects, features and advantages of the present invention will become more apparent through the following description of the embodiments of the present invention with reference to the accompanying drawings, in which:

[0027] Figure 1 A flowchart of an agent-based cloud task scheduling method according to an embodiment of the present invention is schematically shown.

[0028] Figure 2The following schematically shows a flow chart of a task execution and monitoring method according to an embodiment of the present invention.

[0029] Figure 3 An agent-based cloud task scheduling device according to an embodiment of the present invention is schematically shown.

[0030] Figure 4 A block diagram of an electronic device according to an agent-based cloud task scheduling method according to an embodiment of the present invention is schematically shown. DETAILED DESCRIPTION

[0031] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present invention. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of embodiments of the present invention. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concept of the present invention.

[0032] The terms used herein are only for describing specific embodiments and are not intended to limit the present invention. The terms "comprise," "include," etc. used herein indicate the presence of features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0033] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0034] When expressions such as "at least one of A, B, and C, etc." are used, they should generally be interpreted in accordance with the meaning commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0035] Figure 1 A flowchart of an agent-based cloud task scheduling method according to an embodiment of the present invention is schematically shown.

[0036] like Figure 1 As shown, an agent-based cloud task scheduling method according to an embodiment of the present invention includes steps S110 to S130.

[0037] In step S110 , in response to the task request, task characteristics and user preferences are obtained.

[0038] After the task is submitted, detailed task characteristics are extracted from it, including the task's resource requirements (such as the number of CPU cores, memory size, storage capacity, network bandwidth requirements, etc.), the task's execution time requirements (such as deadline, estimated execution time, etc.), the task's priority, the task's interruptibility, the task's dependencies, etc.

[0039] Embodiments of the present invention can provide a variety of user preference setting methods, such as a graphical interface, command line input, and configuration file upload. Users can set preferences based on their needs, such as processing tasks as quickly as possible, processing tasks at the lowest cost, ensuring high reliability of task execution, and prioritizing the use of specific types of cloud resources.

[0040] In step S120, based on the task characteristics and user preferences, an intelligent agent is used to generate a task scheduling action plan, wherein the intelligent agent includes a trained large model, a configuration module, a planning module, a memory module and a variety of tools, including a task description generation tool, a node information acquisition tool, a load fluctuation prediction tool, a node selection tool, a task scheduling plan generation tool, and a task execution tool.

[0041] When a task arrives, the agent plans its action plan in real time and, based on the plan, calls the appropriate tools as needed to complete task scheduling and execution. During this process, the agent dynamically adjusts its action plan to ensure task completion. If a node becomes unavailable during task execution, it will reselect an available node and schedule and execute the task.

[0042] An intelligent agent is a computer program based on a large language model that has the ability to plan, think, remember, and use tools, and can autonomously complete given tasks. It has autonomy, adaptability, and interaction capabilities, and can perceive the environment, make autonomous decisions, and execute actions to achieve specific goals. It has the following characteristics: (1) Autonomy: The intelligent agent can control its own behavior based on its own internal state and perception of the environment without direct external intervention. (2) Perception ability: It can perceive information about the surrounding environment through sensors or data interfaces. (3) Decision-making ability: Based on the perceived environmental information and its own goals, the intelligent agent can use complex algorithms or logical systems to make reasonable decisions. (4) Execution ability: The intelligent agent can convert decisions into actual actions, thereby affecting the environment or its own state.

[0043] In an embodiment of the present invention, the agent is composed of the following parts:

[0044] The trained big model, as the core foundation of the intelligent agent, is mainly responsible for analyzing and processing data, decision-making and planning, generating output results, etc., and uses its powerful language understanding and generation capabilities to handle various tasks. Big models include big models suitable for natural language processing and multimodal information processing. The method of training big models includes: obtaining training data, which includes data on various task characteristics, such as collecting medical task data, emphasizing high reliability requirements and related indicator data; using training data to train the big model to adjust the parameters of the big model and complete the training of the big model so that it can better adapt to the rules of various field tasks during the learning process, thereby supporting the subsequent processing of multiple tasks. Among them, the big model can process multimodal input data, including text, time series data, and image modality data, to improve the comprehensiveness of scheduling decisions.

[0045] The configuration module is used to define the role, profession, skills, behavior pattern, etc. of the intelligent agent, helping the intelligent agent to exhibit specific behaviors and capabilities in different scenarios.

[0046] The planning module is used to break down complex tasks into executable small tasks, select and formulate efficient workflows to achieve goals, and guide the agent's action process through reasoning and planning.

[0047] The memory module is a knowledge base used to store various types of data required by the intelligent agent. Its content comes from the information and data generated during the interaction between the intelligent agent and the environment, or from an external knowledge base. It involves the acquisition, storage, management and retrieval of information, etc., helping the intelligent agent to use historical information to make better decisions.

[0048] The tool module is used to provide the agent with more skills than the large model itself, thereby expanding the agent's skill base and improving its capabilities. It can receive input parameters, execute corresponding functions, and return results. The tools include: task description generation tool, node information acquisition tool, load fluctuation prediction tool, node selection tool, task scheduling solution generation tool, and task execution tool.

[0049] In step S130, according to the action plan, the intelligent agent calls each tool in sequence to generate a task scheduling plan.

[0050] In some exemplary embodiments, step S130 includes steps S131 to S135 .

[0051] In step S131, based on the task characteristics and user preferences, the intelligent agent calls the task description generation tool to determine the task type, integrates the task characteristics, user preferences and task type, and generates a task description in a unified data format.

[0052] The agent invokes the task description generation tool on demand, leveraging a specially designed information collection interface to efficiently and accurately obtain detailed information related to the task. It also comprehensively collects user preferences for task handling. Based on established data standards and specifications, the agent systematically integrates the acquired task information and user preferences to generate a unified and standardized task description, which serves as an important input for subsequent task scheduling decision-making processes. For example, it parses task submission information to determine resource requirements and interruptibility.

[0053] The task description generation tool is designed with a standardized and flexible task information collection interface that supports multiple data formats and transmission protocols. This interface can be connected to task submission systems from different sources, such as internal enterprise business systems and third-party task publishing platforms.

[0054] After collecting the task submission information, detailed task characteristics are extracted from it, including the task's resource requirements (such as the number of CPU cores, memory size, storage capacity, network bandwidth requirements, etc.), the task's execution time requirements (such as deadline, expected execution time, etc.), the task's priority, the task's interruptibility, the task's dependencies, etc.

[0055] Provides a variety of user preference settings, such as a graphical interface, command line input, and configuration file upload. Users can set preferences based on their needs, such as processing tasks as quickly as possible, processing tasks at the lowest cost, ensuring high reliability of task execution, and prioritizing the use of specific types of cloud resources.

[0056] For example, based on task characteristics and user preferences, determine the task type, such as whether it is a computationally intensive task, a data transmission task, or other types of tasks.

[0057] Unified data formats include JSON and XML. Task descriptions clearly distinguish between task characteristics and user preference fields to facilitate subsequent processing and analysis.

[0058] In step S132, the intelligent agent calls the node information acquisition tool to collect resource costs, resource usage information and link status information, and generates a network topology diagram based on graph theory and visualization technology.

[0059] Collect resource cost and usage information, such as CPU utilization, memory usage, and remaining storage resources, as well as link status information, such as network bandwidth, latency, packet loss rate, and jitter. Adopt appropriate communication methods for different cloud platforms, such as using the Representational State Transfer Application Programming Interface (RESTful API) for data exchange with public clouds and using internal network protocols for efficient communication with local clouds.

[0060] The massive amount of collected information is cleaned and preprocessed to remove duplicate, erroneous, and invalid data. Based on this processed information, graph theory and visualization techniques are used to generate an intuitive network topology diagram. In this diagram, nodes represent cloud resource nodes, and edges represent links between nodes. Key attributes such as link bandwidth, latency, and packet loss rate are also annotated. A dynamic update mechanism is implemented to ensure that the network topology diagram reflects the actual status of the cloud environment in real time.

[0061] In step S133, the intelligent agent is used to call the load fluctuation prediction tool to predict the task load fluctuation based on the trained time series prediction model.

[0062] The system passes relevant information about the current task, such as the task type and historical execution data, to the time series prediction model. Based on this information and its own trained model parameters, the time series prediction model predicts task load fluctuations, such as whether traffic peaks will occur in the future or the possibility of hardware failures.

[0063] In some exemplary embodiments, the timing prediction model includes a long short-term memory network, and the method for training the timing prediction model includes: obtaining historical task load data and marking the time points of key events, wherein the task load data includes traffic data and hardware status data, and the key events include traffic peaks and hardware failures; using the historical task load data to train the timing prediction model, and adjusting the timing prediction model so that the timing prediction model can predict task load fluctuations.

[0064] In step S134, the intelligent agent calls the node selection tool to select the link node based on the network topology diagram and task load fluctuation.

[0065] In some exemplary embodiments, step S134 includes steps S1341 to S1344.

[0066] In step S1341, based on the network topology diagram, the bandwidth and delay information of each link are analyzed using the trained large model.

[0067] In step S1342 , based on the bandwidth and delay information, key information of the link is identified, including reliability and transmission efficiency.

[0068] In step S1343, the initial link nodes are obtained based on resource costs and task load fluctuations. If a traffic peak is predicted, nodes with high bandwidth and low latency links are prioritized to cope with the possible high load. If a high risk of hardware failure is predicted, nodes that may be affected are avoided.

[0069] In step S1344, the initial link nodes are further screened based on the characteristics of the task to obtain the final link nodes. For example, for tasks with high resource requirements, nodes with abundant resources are selected; for tasks that cannot be interrupted, nodes with high reliability are selected.

[0070] In step S135, the intelligent agent calls the task scheduling solution generation tool, selects the task scheduling method based on the task type and user preference, and generates the task scheduling solution based on the task description and the selected link node using the selected task scheduling method.

[0071] Task scheduling methods can be tailored to different types of tasks, including parallel tasks and workflow tasks. Task scheduling methods can employ various scheduling methods, such as reinforcement learning, heuristic algorithms, metaheuristic algorithms, and rule-based scheduling algorithms, to suit different types of tasks. These methods include reinforcement learning scheduling, heuristic algorithm scheduling, metaheuristic algorithm scheduling, and rule-based scheduling algorithms.

[0072] Optionally, the method for training the reinforcement learning scheduling method includes: setting a reward function based on the reinforcement learning scheduling method, for example, giving high rewards to scheduling strategies that can complete tasks quickly and have high resource utilization, and the reward function is used to reflect the pros and cons of the scheduling strategy; designing a state space and an action space, the state space includes the resource requirements of the task, whether it can be interrupted, and the resource status of the current cloud environment, and the action space includes different scheduling decisions, such as selecting different cloud nodes, allocating different amounts of resources, etc.; based on the reward function, state space, and action space, using a reinforcement learning algorithm, such as a deep Q network algorithm, to train the reinforcement learning scheduling method in a simulated cloud task environment, so that the reinforcement learning scheduling method learns the optimal scheduling strategy.

[0073] Optionally, the method of training the heuristic algorithm scheduling method includes: setting heuristic rules based on task characteristics and cloud environment characteristics, such as allocating resources to urgent tasks with small resource requirements based on the urgency of the tasks and resource requirements; based on actual task data, running the heuristic algorithm using the heuristic algorithm scheduling method to optimize the parameters and rules of the heuristic algorithm.

[0074] Optionally, the meta-heuristic algorithm scheduling method includes a genetic algorithm model, and the training method of the genetic algorithm model includes: defining a chromosome encoding method and encoding the scheduling plan as a chromosome; setting a fitness function and evaluating the pros and cons of the scheduling plan corresponding to each chromosome; through genetic operations such as selection, crossover, and mutation, continuously evolving the chromosomes to obtain a better scheduling plan, thereby training a genetic algorithm model suitable for cloud task scheduling.

[0075] Optionally, methods for training rule-based scheduling algorithms include: formulating scheduling rules based on domain knowledge and scheduling requirements, such as prioritizing the allocation of reliable cloud node resources for non-interruptible tasks; applying the scheduling rules in actual task scenarios, and continuously adjusting and improving the scheduling rules based on feedback.

[0076] The task description, available nodes, and related information are passed to the corresponding model. For example, for computationally intensive tasks where the user prefers the fastest processing speed, a scheduling method based on reinforcement learning and optimized for computational resources is used; for data transmission tasks, a heuristic algorithm scheduling method that considers network link conditions is used.

[0077] After receiving input information, a scheduling method performs task planning based on its own algorithms and parameters, generating a task scheduling plan. For example, a reinforcement learning scheduling method selects appropriate actions from its learned optimal strategy based on the current state and generates a specific scheduling plan, including information such as which cloud nodes the tasks are assigned to, how much resources are allocated to each node, and the order in which tasks are executed. Heuristic scheduling methods sort tasks and allocate resources based on designed heuristic rules to generate a scheduling plan. Metaheuristic scheduling methods generate new scheduling plans through methods such as genetic operations, evaluate their pros and cons, and select the optimal one. Rule-based scheduling algorithms generate scheduling plans based on established scheduling rules.

[0078] exist Figure 1 The method shown in the figure can also include using an agent to call a task execution tool to execute and monitor the task. The execution and monitoring method includes steps S210 to S220. Figure 2 .

[0079] In step S210 , the task scheduling solution is assigned to the selected cloud node for execution.

[0080] After receiving the scheduling plan generated by the task scheduling plan generation tool, the agent calls the task execution tool to assign tasks to the selected cloud nodes and start the task execution process in an orderly manner. During task assignment, it interacts with the cloud node's task management system and transmits task-related information (such as task code, input data, configuration parameters, etc.) to the target node using a standardized interface protocol.

[0081] In step S220 , the execution status of the task is continuously monitored.

[0082] During task execution, a real-time monitoring mechanism closely monitors task progress and resource usage. Once an abnormality is detected during task execution, such as resource shortages or link interruptions, feedback is quickly provided to the agent to ensure the continuity and stability of task execution.

[0083] Specifically, during task execution, the system continuously monitors the task's execution status. By communicating with the cloud node's monitoring interface, it obtains real-time information on task progress (e.g., completed workload, remaining workload, etc.) and resource usage (e.g., CPU utilization, memory usage, storage read / write rates, etc.). Visualization technology is used to intuitively display task execution status, facilitating real-time monitoring and management by administrators.

[0084] When anomalies occur during task execution, such as insufficient resources, link interruptions, or task failures, timely response measures are implemented. Exception information, including the time, type, and scope of impact, is recorded. Based on the severity and type of the anomaly, appropriate handling methods are selected, such as rescheduling the task, adjusting resource allocation, or retrying the task. Furthermore, this information is fed back to the task scheduling agent to adjust and optimize the action plan.

[0085] After the task is completed, the results are verified and confirmed. The task output is checked to see if it meets expectations and compared with the task input requirements and objectives. The task results are fed back to the relevant users and systems, and the task status information and resource usage records are updated.

[0086] Figure 3 An agent-based cloud task scheduling device according to an embodiment of the present invention is schematically shown.

[0087] like Figure 3 As shown, the agent-based cloud task scheduling device 800 of this embodiment includes an acquisition module 810 , a planning module 820 and a generation module 830 .

[0088] The acquisition module 810 is configured to acquire task characteristics and user preferences in response to a task request.

[0089] The planning module 820 is used to generate a task scheduling action plan based on task characteristics and user preferences using an intelligent agent. The intelligent agent includes a trained large model, a configuration module, a planning module, a memory module, and a variety of tools. The tools include a task description generation tool, a node information acquisition tool, a load fluctuation prediction tool, a node selection tool, a task scheduling plan generation tool, and a task execution tool.

[0090] The generating module 830 is configured to generate a task scheduling scheme by sequentially calling tools by using the intelligent agent according to the action plan, and the generating a task scheduling scheme by sequentially calling tools by using the intelligent agent according to the action plan includes: calling a task description generating tool by using the intelligent agent to determine a task type based on a task characteristic and a user preference, integrating the task characteristic, the user preference, and the task type, and generating a task description in a unified data format; calling a node information obtaining tool by using the intelligent agent to collect resource cost, resource usage information, and link condition information, and generating a network topology graph based on graph theory and visualization technology; calling a load fluctuation prediction tool by using the intelligent agent, and predicting a task load fluctuation condition based on a trained time series prediction model; calling a node selection tool by using the intelligent agent, and selecting a link node based on the network topology graph and the task load fluctuation condition; and calling a task scheduling scheme generating tool by using the intelligent agent, selecting a task scheduling method based on the task type and the user preference, and generating a task scheduling scheme by using the selected task scheduling method based on the task description and the selected link node, wherein the task characteristic includes a resource requirement of the task, an execution time requirement of the task, a priority of the task, interruptability of the task, and a dependency relationship of the task. According to an embodiment of the present application, any one or more of the obtaining module 810, the planning module 820, and the generating module 830 can be combined in one module, or any one of the modules can be split into multiple modules. Alternatively, at least part of the function of one or more of the modules can be combined with at least part of the function of another module, and implemented in one module. According to an embodiment of the present application, at least one of the obtaining module 810, the planning module 820, and the generating module 830 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on board, a system on package, an application specific integrated circuit (ASIC), or any other reasonable way of integrating or packaging a circuit, a hardware or firmware, or implemented in any one of software, hardware, and firmware or in a proper combination of any one or more of them. Alternatively, at least one of the obtaining module 810, the planning module 820, and the generating module 830 can be at least partially implemented as a computer program module that can perform corresponding functions when the computer program module is run.

[0091] Figure 4 A block diagram of an electronic device according to an embodiment of the present application is schematically shown.

[0092] As Figure 4As shown, the electronic device 900 according to an embodiment of the present application includes a processor 901 which can perform various appropriate actions and processes in accordance with a program stored in a read only memory (ROM) 902 or a program loaded from a storage section 908 into a random access memory (RAM) 903. The processor 901 can include, for example, a general purpose microprocessor (e.g., a CPU), an instruction set processor, and / or a related chip set, and / or a special purpose microprocessor (e.g., an application specific integrated circuit (ASIC)), and so on. The processor 901 can also include an on-board memory for cache use. The processor 901 can include a single processing unit or multiple processing units to perform the various actions of the method processes according to embodiments of the present application.

[0093] In the RAM 903, various programs and data required for the operation of the electronic device 900 are stored. The processor 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. The processor 901 performs various operations of the method processes according to embodiments of the present application by executing the programs in the ROM 902 and / or the RAM 903. Note that the programs can also be stored in one or more memories other than the ROM 902 and the RAM 903. The processor 901 can also perform various operations of the method processes according to embodiments of the present application by executing the programs stored in the one or more memories.

[0094] According to an embodiment of the present application, the electronic device 900 can also include an input / output (I / O) interface 905 which is also connected to the bus 904. The electronic device 900 can also include one or more of the following components connected to the input / output (I / O) interface 905: an input section 906 including a keyboard, a mouse, etc.; an output section 907 including a display such as a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 908 including a hard disk, etc.; and a communication section 909 including a network interface card such as a LAN card, a modem, etc. The communication section 909 performs communication processing via a network such as the Internet. A drive 910 is also connected to the input / output (I / O) interface 905 as necessary. A removable medium 911 such as a magnetic disk, a magneto-optical disk, a semiconductor memory, etc. is attached to the drive 910 as necessary, so that a computer program read out from the removable medium 911 is installed in the storage section 908 as necessary.

[0095] The present invention also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments, or may exist independently and not incorporated into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of the present invention.

[0096] According to an embodiment of the present invention, a computer-readable storage medium may be a non-volatile computer-readable storage medium, and may include, for example, but not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to an embodiment of the present invention, a computer-readable storage medium may include the ROM 902 and / or RAM 903 described above, and / or one or more memories other than ROM 902 and RAM 903.

[0097] It will be understood by those skilled in the art that the features described in the various embodiments of the present invention may be combined and / or coupled in various ways, even if such combinations or couplings are not explicitly described in the present invention. In particular, the features described in the various embodiments of the present invention may be combined and / or coupled in various ways without departing from the spirit and teachings of the present invention. All such combinations and / or couplings fall within the scope of the present invention.

Claims

1. A cloud task scheduling method based on intelligent agent, characterized in that: include: Responding to a task request, obtaining task characteristics and user preferences; Based on the task characteristics and the user preferences, an intelligent agent is used to generate a task scheduling action plan, wherein the intelligent agent includes a trained large model, a configuration module, a planning module, a memory module, and a variety of tools, including a task description generation tool, a node information acquisition tool, a load fluctuation prediction tool, a node selection tool, a task scheduling solution generation tool, and a task execution tool; According to the action plan, the intelligent agent calls each tool in turn to generate a task scheduling plan. The step of using the agent to sequentially call various tools according to the action plan to generate a task scheduling plan includes: Based on the task characteristics and the user preferences, the agent calls a task description generation tool to determine the task type, integrates the task characteristics, the user preferences and the task type, and generates a task description in a unified data format; Utilize the agent to call the node information acquisition tool to collect resource costs, resource usage information, and link status information, and generate a network topology diagram based on graph theory and visualization technology; Using the intelligent agent to call the load fluctuation prediction tool, based on the trained time series prediction model, predict the task load fluctuation; Using the agent to call a node selection tool, select a link node based on the network topology diagram and the task load fluctuation; Using the agent to call a task scheduling solution generation tool, based on the task type and user preference, select a task scheduling method, based on the task description and the selected link node, generate a task scheduling solution using the selected task scheduling method, The task characteristics include the resource requirements of the task, the execution time requirements of the task, the priority of the task, the interruptibility of the task, and the dependency of the task.

2. The method according to claim 1, characterized in that The large model includes a large model suitable for natural language processing and multimodal information processing, and the method for training the large model includes: Acquiring training data, wherein the training data includes data of various task characteristics; Using the training data to train the large model, so as to adjust the parameters of the large model and complete the training of the large model, The large model is capable of processing multimodal input data, including text, time series data, and image modality data.

3. The method according to claim 1, characterized in that The time series prediction model includes a long short-term memory network, and the method for training the time series prediction model includes: Obtain historical task load data and mark the time points of key events, wherein the task load data includes traffic data and hardware status data, and the key events include traffic peaks and hardware failures; The time series prediction model is trained using the historical task load data, and the time series prediction model is adjusted so that the time series prediction model can predict task load fluctuations.

4. The method according to claim 1, wherein The task scheduling method includes reinforcement learning scheduling method, heuristic algorithm scheduling method, meta-heuristic algorithm scheduling method and rule-based scheduling algorithm, The method of training the reinforcement learning scheduling method includes: Based on the reinforcement learning scheduling method, a reward function is set, wherein the reward function is used to reflect the quality of the scheduling strategy; Designing a state space and an action space. The state space includes the task's resource requirements, whether it can be interrupted, and the current resource status of the cloud environment. The action space includes different scheduling decisions. Based on the reward function, the state space, and the action space, using a reinforcement learning algorithm, the reinforcement learning scheduling method is trained in a simulated cloud task environment, so that the reinforcement learning scheduling method learns an optimal scheduling strategy; and / or, The method of training the heuristic algorithm scheduling method includes: Setting heuristic rules based on the characteristics of the task and the characteristics of the cloud environment; Based on actual task data, the heuristic algorithm is run using the heuristic algorithm scheduling method to optimize parameters and rules of the heuristic algorithm; and / or, The meta-heuristic algorithm scheduling method includes a genetic algorithm model, and the training method of the genetic algorithm model includes: Define the chromosome encoding method and encode the scheduling plan into chromosomes; Set the fitness function to evaluate the pros and cons of the scheduling scheme corresponding to each chromosome; Through genetic operations such as selection, crossover, and mutation, chromosomes are continuously evolved to obtain better scheduling solutions, thereby training a genetic algorithm model suitable for cloud task scheduling; and / or, The method of training the rule-based scheduling algorithm includes: Formulate scheduling rules based on domain knowledge and scheduling requirements; The scheduling rules are applied in actual task scenarios and are continuously adjusted and improved based on feedback.

5. The method according to claim 1, wherein The nodes in the network topology diagram represent cloud resource nodes, and the edges represent links between nodes; The network topology diagram includes the bandwidth, delay and packet loss rate of the link; A dynamic update mechanism is adopted to ensure that the network topology map reflects the actual status of the cloud environment in real time.

6. The method according to claim 1, characterized in that The selecting a link node based on the network topology diagram and the task load fluctuation includes: Based on the network topology diagram, the bandwidth and delay information of each link are analyzed using the trained large model; Based on the bandwidth and the delay information, identifying key information of the link, including reliability and transmission efficiency; Based on the resource cost and the task load fluctuation, obtaining an initial link node; Based on the task characteristics, the initial link nodes are further screened to obtain the final link nodes.

7. The method according to claim 1, characterized in that Also includes: Allocating the task scheduling solution to the selected cloud node for execution; Continuously monitor the execution status of tasks, and take corresponding countermeasures when abnormal situations occur during task execution, including: recording abnormal information; selecting corresponding processing methods based on the severity and type of the abnormality; and feeding back the abnormal information to the intelligent agent to adjust and optimize the task scheduling plan.

8. A cloud task scheduling device based on an intelligent agent, characterized in that: The device comprises: an acquisition module, for acquiring task characteristics and user preferences in response to a task request; A planning module, configured to generate a task scheduling action plan using an agent based on the task characteristics and user preferences, wherein the agent includes a trained large model, a configuration module, a planning module, a memory module, and a variety of tools, including a task description generation tool, a node information acquisition tool, a load fluctuation prediction tool, a node selection tool, a task scheduling solution generation tool, and a task execution tool; The generation module is used to call various tools in sequence according to the action plan using the intelligent agent to generate a task scheduling plan. The step of using the agent to sequentially call various tools according to the action plan to generate a task scheduling plan includes: Based on the task characteristics and the user preferences, the agent calls a task description generation tool to determine the task type, integrates the task characteristics, the user preferences and the task type, and generates a task description in a unified data format; Utilize the agent to call the node information acquisition tool to collect resource costs, resource usage information, and link status information, and generate a network topology diagram based on graph theory and visualization technology; Using the intelligent agent to call the load fluctuation prediction tool, based on the trained time series prediction model, predict the task load fluctuation; Using the agent to call a node selection tool, select a link node based on the network topology diagram and the task load fluctuation; Using the agent to call a task scheduling solution generation tool, based on the task type and user preference, select a task scheduling method, based on the task description and the selected link node, generate a task scheduling solution using the selected task scheduling method, The task characteristics include the resource requirements of the task, the execution time requirements of the task, the priority of the task, the interruptibility of the task, and the dependency of the task.

9. An electronic device, characterized in that: include: one or more processors; a memory for storing one or more computer programs, The one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program or instruction stored thereon, characterized in that: When the computer program or instruction is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images