Server-free intelligent resource scheduling system based on large language model

Abstract

The invention discloses a server-free intelligent resource scheduling system based on a large language model, which guides a resource scheduler through the large language model and realizes efficient scheduling and resource utilization optimization applied to a distributed system. According to the technology, the load demand, the system state and the resource distribution of the application are analyzed in real time and dynamically adjusted, so that the scheduler can flexibly adapt to a complex and changeable working load environment, and efficient utilization of computing resources and high availability of the system are ensured. The invention aims to improve the intelligent level of resource scheduling under a server-free architecture, reduce the operation cost of the system and improve the stability and performance of the system.

Description

Technical Field

[0001] The present invention relates to the field of cloud computing, and in particular to a serverless intelligent resource scheduling system based on a large language model. Background Art

[0002] As many organizations around the world continue to seek effective ways to reduce infrastructure pressure and hope to improve business performance through cost-effective information technology solutions, the serverless architecture market is experiencing rapid growth. In 2023, the global serverless architecture market size is US$9.42 billion, and the compound annual growth rate is expected to be 28.2% from 2024 to 2030, and the market size is expected to reach US$50.09 billion by 2030. Serverless computing (Serverless Computing), referred to as serverless, is a cloud computing architecture in which cloud service providers (such as Amazon Cloud Services) dynamically allocate computing resources based on demand and are responsible for server management. Serverless only runs application code on a per-request basis and automatically scales computing resources up and down based on the number of requests. With serverless, customers only pay for the resources used when the application is running, effectively reducing costs.

[0003] With the rapid development of serverless technology, the management and scheduling of computing resources have become increasingly complex. In traditional serverless computing architectures, resource scheduling usually relies on static rules or schedulers based on algorithmic models, which often cannot respond to dynamically changing workloads and resource requirements in real time. Especially today when containerized applications and microservice architectures are widely used, how to efficiently utilize computing resources in distributed systems to ensure high availability and good performance of applications has become an urgent problem to be solved.

[0004] Therefore, a serverless intelligent resource scheduling system based on a large language model is desired. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed. The embodiment of this application provides a serverless intelligent resource scheduling system based on a large language model, which realizes the intelligence and real-time performance of scheduling decisions, effectively copes with complex distributed system environments, improves the accuracy and efficiency of resource scheduling, and ensures that application loads run on appropriate nodes.

[0006] According to one aspect of the present application, a server-free intelligent resource scheduling system based on a large language model is provided, which includes: a task management module, which is used to receive and parse user task requests, and convert the natural language input by the user into structured data that can be processed by the system; a resource monitoring module, which is used to monitor the resource status of each computing node in the system in real time; a large language model analysis engine module, which is used to analyze task requirements and resource status and generate a scheduling plan; a scheduler module, which is used to assign tasks to specific computing nodes for execution according to the scheduling plan generated by the large language model analysis engine; a deployment node module, which is used to deploy tasks to actual computing nodes for execution, to ensure that tasks can be started and completed as expected.

[0007] Preferably, the task management module is used for: the task manager module obtains the task request submitted by the user, and the task management module is responsible for parsing the request and extracting parameter information of the task's computing requirements, memory usage, and storage requirements.

[0008] Preferably, the task management module includes: transmitting the parameter information to a large language model analysis engine module.

[0009] Preferably, the resource monitoring module is responsible for collecting and updating the resource status of all nodes in the system in real time, and the resource status of all nodes includes: CPU, memory, storage and network bandwidth.

[0010] Preferably, the resource monitoring module includes: regularly reporting the resource utilization and health status of each node to the large language model analysis engine module.

[0011] Preferably, the large language model analysis engine module includes: the large language model analysis engine is the core module of the system, which quickly adapts to environmental changes through semantic analysis of task requirements and node status, through context understanding and reasoning, and dynamically adjusts scheduling strategies according to resources in the system.

[0012] Preferably, the large language model analysis engine module includes: a task request parsing and allocation unit, which is used to accept parameter information from the task management module, use context understanding and reasoning capabilities to understand the execution requirements of the task, and generate a resource scheduling plan; a resource monitoring and dynamic adjustment unit, which is used to accept resource information from the resource monitoring module and dynamically adjust the resource allocation strategy.

[0013] Preferably, the scheduler module is used to determine the best execution position of the task according to the scheduling scheme generated by the large language model analysis engine module.

[0014] Preferably, the deployment node module is used to: receive task allocation instructions from the scheduler, and actually deploy specific tasks to designated computing nodes.

[0015] According to another aspect of the present application, a serverless intelligent resource scheduling system based on a large language model is provided, which includes: S1, reading computer resource information; S2, monitoring all node resource information; S3, the large language model combines task request resource information and node resource information to generate a scheduling strategy; S4: according to the scheduling strategy, the scheduler deploys the task to the node.

[0016] The present invention is based on a serverless resource scheduler architecture of a large language model, which significantly improves resource utilization efficiency and system flexibility through intelligent resource scheduling and dynamic management. The present invention uses a large language model to analyze and predict resource requirements in real time, and can automatically adjust resource allocation. This mechanism is different from the traditional static scheduling method, which is often unable to dynamically adjust according to the actual load, resulting in waste or shortage of resources. Traditional resource scheduling schemes rely on manual intervention or require a lot of training costs. The present invention uses a large language model to make automatic decisions, reducing dependence on manual intervention and requiring no additional training costs.

[0017] The present invention combines the natural language processing capabilities of a large language model to process user requests in real time and provide feedback. This makes the system more user-friendly and efficient in user interaction, while traditional serverless schedulers often lack this flexibility, resulting in a poor user experience. Traditional solutions are based on static rules or predictive models, which are not scalable enough in different environments. In the face of new environments, a new set of rules may have to be formulated or a new model may have to be trained. In contrast, the present invention is based on a large language model, and thanks to the large language model's powerful contextual understanding capabilities and rich knowledge level, it can achieve fast and simple serverless expansion to meet actual needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0019] Figure 1 A block diagram of a serverless intelligent resource scheduling system based on a large language model according to an embodiment of the present application

[0020] Figure 2 This is an architecture diagram of a serverless intelligent resource scheduling system based on a large language model according to an embodiment of the present application.

[0021] Figure 3This is an architectural diagram of an intelligent parsing and allocation mechanism for task requests in a serverless intelligent resource scheduling system based on a large language model according to an embodiment of the present application.

[0022] Figure 4 This is an architecture diagram of a real-time resource monitoring and dynamic adjustment mechanism in a serverless intelligent resource scheduling system based on a large language model according to an embodiment of the present application.

[0023] Figure 5 This is a flowchart of a server-free intelligent resource scheduling method based on a large language model according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] Various exemplary embodiments, features and aspects of the present application will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0025] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0026] In addition, in order to better illustrate the present application, numerous specific details are provided in the following specific embodiments. It should be understood by those skilled in the art that the present application can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present application.

[0027] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0028] In order to achieve efficient resource scheduling under the serverless architecture, existing technologies mainly solve the resource management problem in the cloud computing environment from two aspects: resource allocation and task scheduling. They mainly include: optimization of resource usage and intelligent resource scheduling.

[0029] Resource optimization refers to ensuring that computing resources (such as CPU, memory, network, etc.) can be fully utilized through intelligent allocation algorithms and scheduling strategies before task execution. Since resource scheduling is not only related to the operating efficiency of tasks, but also directly affects the overall performance and cost control of the system, resource optimization is particularly important in serverless architecture. It mainly includes resource isolation and utilization of idle resources. Resource isolation technology is to avoid mutual interference between different tasks or applications when sharing resources and to ensure the stability and security of resources. Typical technologies include containerization technology (such as Docker), which ensures that the resource usage of each application is independent and will not be affected by other applications by encapsulating applications in independent containers. Through isolation, the system can effectively prevent the high-load operation of an application from occupying too many resources, thereby affecting the normal operation of other applications. Idle resource utilization in serverless architecture reduces resource waste by making full use of temporarily unused resources in the system. Freyr is a new resource manager for serverless platforms that maximizes resource efficiency by dynamically collecting idle resources.

[0030] Intelligent resource scheduling ensures that computing tasks can be efficiently allocated to the most appropriate computing nodes through real-time analysis of current tasks and dynamic adjustment of scheduling algorithms, thereby maximizing system performance and task execution efficiency. Intelligent resource scheduling is mainly divided into rule-based scheduling and scheduling based on machine learning models. Rule-based scheduling refers to the allocation and scheduling of resources through predefined rules and conditions. For example, the Kubernetes scheduler determines which node to deploy a container on by preset resource requests, priorities, and labels. This type of scheduling method is simple and easy to use, and performs well in certain resource demand scenarios. However, when dealing with complex and changing workloads, rule scheduling has poor flexibility and is difficult to cope with dynamically changing environments. Intelligent scheduling systems based on machine learning models can automatically learn the best scheduling strategy based on current resource conditions and task requirements by collecting and analyzing historical data, and optimize it in future task scheduling. For example, Aquatope uses a set of scalable and proven Bayesian machine learning models to create preheated containers before function calls, minimizing resource costs. However, scheduling systems based on machine learning rely on a large amount of training data and computing resources. In the initial deployment phase, they often require a long learning cycle and are difficult to achieve rapid resource response. In addition, the accuracy of machine learning models is also limited by the quality and quantity of data. Once the environment changes significantly, the model's predictive ability may decline.

[0031] The above methods greatly improve the efficiency of serverless resource scheduling, but the following problems still exist: Limitations of traditional static rule scheduling. The existing static rule scheduling method lacks sufficient flexibility, especially in dynamic and changing workload environments, and cannot be adjusted in real time. Traditional scheduling methods often have difficulty coping with complex distributed systems and microservice architectures, resulting in uneven resource utilization and even performance bottlenecks. Therefore, a flexible and dynamic scheduling system is an important guarantee for improving the efficiency of serverless resource scheduling.

[0032] The learning cycle of machine learning-based scheduling systems is long. Existing scheduling systems based on machine learning rely on a large amount of historical data for training, and a long learning cycle is required during the initial deployment to optimize the scheduling strategy. At the same time, when the environment changes drastically, the response speed of such scheduling systems is slow, which may cause resource allocation delays. In addition, the existing machine learning scheduling systems have high requirements for data quality and quantity. If the training data is insufficient or the data quality is poor, it will directly affect the scheduling accuracy of the model, resulting in unreasonable scheduling and waste of resources. Therefore, a scheduling system that can shorten the learning cycle and make intelligent resource allocation decisions even when the data is limited or the data quality is not high is particularly critical.

[0033] Intelligent scheduling systems are slow to respond to complex environments. Current intelligent scheduling systems often fail to adjust scheduling strategies in a timely manner when faced with the diversity, complexity, and high-frequency changes in distributed environments. This lag may lead to inefficient resource utilization, degraded application performance, and even affect service availability. Therefore, a scheduler that is good at processing complex multi-dimensional information, can analyze system status and task requirements in real time, and quickly make optimal scheduling decisions can ensure efficient resource allocation and high system availability.

[0034] With the development of serverless architecture, more and more organizations are seeking to optimize business performance and reduce infrastructure costs through efficient cloud computing architecture. However, existing serverless computing architectures face multiple challenges in resource scheduling and management. Traditional scheduling systems usually rely on static rules or machine learning-based algorithms. These methods have significant defects in dealing with complex and dynamically changing workloads and are difficult to effectively respond to changes in resource requirements in distributed systems. It can be seen that designing a stable and efficient scheduler is an important link to ensure the smooth operation of serverless resource scheduling. The purpose of the present invention is to provide a serverless intelligent scheduling method based on a large language model in response to the deficiencies in the above-mentioned prior art. By using the natural language processing and context understanding capabilities of the large language model, the scheduler can analyze the system status and task requirements in real time, and quickly make the optimal scheduling decision without the need for a large amount of training data. Compared with traditional scheduling methods based on rules or machine learning, the large language model scheduler can adapt to complex dynamic environments more flexibly, improve resource utilization, shorten the learning cycle, and effectively respond to the challenges brought about by insufficient data or environmental changes, thereby improving the overall performance and high availability of the system.

[0035] In ordinary resource scheduling, task requests need to be pre-defined templates. The system classifies tasks according to these rules, and then selects appropriate nodes based on parameters such as task type, priority, and resource requirements. Some systems also schedule based on historical data or preset workload patterns. For example, resource allocation is determined based on load forecasts and historical task behavior. The disadvantages of the above-mentioned ordinary resource scheduling methods are that they are static and inflexible, and lack the ability to adapt to complex and dynamically changing task requirements and resource status in real time. Large language models can instantly understand new requests and update scheduling plans without the need to manually adjust rules or retrain the system.

[0036] The present invention is further described in detail below with reference to the accompanying drawings and specific implementation methods:

[0037] Figure 1 FIG. 1 is a block diagram of a serverless intelligent resource scheduling system based on a large language model according to an embodiment of the present application. Figure 1 As shown, the serverless intelligent resource scheduling system based on the large language model is composed of multiple subsystems, mainly including a task management module, a resource monitoring module, a deployment node module, a large language model analysis engine module and a scheduler module.

[0038] Figure 2 FIG. 1 is an architecture diagram of a serverless intelligent resource scheduling system based on a large language model according to an embodiment of the present application. Figure 2As shown in the figure, these modules work together to ensure that the system can dynamically and efficiently schedule tasks. The task management module is used to receive and parse the user's task requests and convert the natural language input by the user into structured data that the system can process. The resource monitoring module is used to monitor the resource status of each computing node in the system in real time. The large language model analysis engine module is used to analyze task requirements and resource status and generate a scheduling plan. The scheduler module is used to assign tasks to specific computing nodes for execution according to the scheduling plan generated by the large language model analysis engine. The deployment node module is used to deploy tasks to actual computing nodes for execution, ensuring that tasks can be started and completed as expected.

[0039] Figure 3 FIG. 1 is an architecture diagram of an intelligent parsing and allocation mechanism for task requests in a serverless intelligent resource scheduling system based on a large language model according to an embodiment of the present application. Figure 3 As shown, in an embodiment of the present application, the task management module is used for: the task manager module obtains the task request submitted by the user, and the task management module is responsible for parsing the request and extracting the parameter information of the computing requirement, memory usage, and storage requirement of the task. The parameter information is passed to the large language model analysis engine module. It should be understood that the task manager module obtains the task request submitted by the user, and the task management module is responsible for parsing the request and extracting the computing requirement, memory usage, storage requirement and other information of the task. Subsequently, this information will be passed to the large language model analysis engine, and the large language model uses context understanding and reasoning capabilities to understand the execution requirements of the task, generate a resource scheduling plan, ensure that the task can be assigned to a node that meets the resource requirements, and ensure the rationality of resource allocation and the efficiency of task execution. This process can not only improve resource utilization, but also reduce the overall cost of system operation.

[0040] Specifically, the task management module also essentially utilizes the large language model technology. The user will submit the task requirements (for example, the task needs to be run on a machine with a 2-core CPU and 4GB of memory) to the task management module. Through the prompt word engineering technology, the task management module will convert the task requirements from natural language into JSON format ({cpu:2,memory:4g}). Now, the natural language task requirements are converted into structured information that can be recognized by computers. Semantic analysis involves the mechanism of the large language model. Simply put, the large language model can understand the relationship between words through a large amount of corpus training, and the output results are what it considers reasonable and relevant.

[0041] Figure 4 FIG. 1 is an architecture diagram of a real-time resource monitoring and dynamic adjustment mechanism in a serverless intelligent resource scheduling system based on a large language model according to an embodiment of the present application. Figure 4As shown, in an embodiment of the present application, the resource monitoring module is responsible for collecting and updating the resource status of all nodes in the system in real time, and all nodes include: CPU, memory, storage and network bandwidth. Regularly report the resource utilization and health status of each node to the large language model analysis engine module. It should be understood that the resource monitoring module of the system is responsible for collecting and updating the resource status of all nodes in the system in real time, including CPU, memory, storage, network bandwidth, etc. By regularly reporting the resource utilization and health status of each node, the system can dynamically adjust the resource allocation strategy. When a node is short of resources, the large language model can regenerate the scheduling plan and migrate the task to a more suitable node to avoid single-point resource exhaustion and system performance degradation. This real-time adjustment mechanism ensures the high availability and stability of the system.

[0042] Specifically, after the user submits the task requirements, the system will automatically collect the resource status of all running nodes (node ​​1: 1 core CPU, 2GB memory; node 2: 2 core CPU, 4GB memory), and organize the status of all nodes into JSON format.

[0043] In the embodiment of the present application, the large language model analysis engine module includes: the large language model analysis engine is the core module of the system, which quickly adapts to environmental changes through semantic analysis of task requirements and node status, through context understanding and reasoning, and dynamically adjusts the scheduling strategy according to the resources in the system. Specifically, it includes: a task request parsing and allocation unit, which is used to accept parameter information from the task management module, use context understanding and reasoning capabilities, understand the execution requirements of the task, and generate a resource scheduling plan; a resource monitoring and dynamic adjustment unit, which is used to accept resource information from the resource monitoring module and dynamically adjust the resource allocation strategy. It should be understood that the large language model analysis engine is the core module of the system, which generates a scheduling plan in real time through semantic analysis of task requirements and node status. Compared with traditional scheduling systems based on rules or machine learning, the large language model of the present invention can quickly adapt to environmental changes through rich context understanding and reasoning, and dynamically adjust the scheduling strategy according to the resources in the system. The scheduler efficiently allocates tasks to appropriate nodes based on the suggestions provided by the large language model. This scheduling method effectively avoids resource waste and improves the overall performance of the system.

[0044] Correspondingly, the above task requirements and resource information will be sent to the large language model (GPT or Wenxinyiyan) together. Through the prompt word engineering technology, the large language model will be required to answer the node name (node ​​2) that is most suitable for task operation. After obtaining the output result of the large language model, the system will notify the scheduler module to bind the task to node 2, and the deployment node module will virtualize and deploy the task to node 2 when the task starts running. In fact, through the prompt word engineering technology, the large language model can understand more complex task requirements than the above, and guide the large language model to output a suitable task running node. The large language model analysis engine is that it can deeply understand the user's task requirements based on natural language processing technology, and combine it with the system's resource status to generate a resource scheduling plan. Traditional resource scheduling systems usually rely on hard-coded rules or simple matching based on limited resource status, while the large language model can dynamically extract deep information from task requests and resource status through semantic analysis capabilities.

[0045] Specifically, the present application implements serverless dynamic resource scheduling through a large language model, analyzes load changes in real time, and automatically optimizes resource allocation. Compared with existing machine learning-based models, this method not only does not rely on historical data, but also understands natural language input, thereby more flexibly adapting to complex user needs. The present invention allows for rapid resource reallocation during serverless resource use to ensure that task requirements are met in a timely manner. Traditional machine learning-based reallocation strategies are often limited to known patterns and are difficult to cope with sudden changes, while the present invention can make adjustments more quickly and effectively through real-time feedback and decision-making mechanisms. The present invention uses the intelligent parsing capabilities of a large language model to quickly understand user requests and perform efficient serverless task scheduling. Compared with traditional machine learning models, which usually require additional feature engineering and manual intervention, the present invention can directly process complex natural languages ​​and simplify the process.

[0046] Accordingly, the serverless resource scheduling and reallocation decisions of the present application all rely on the large language model. This centralized decision-making mechanism can achieve more efficient resource management in a multi-tasking environment. Compared with traditional decentralized decision-making, the present invention ensures the coordination and consistency of various decisions. The present application integrates a real-time feedback system based on a large language model, which can instantly optimize the serverless scheduling strategy according to user feedback and system status. This mechanism is different from the traditional offline update method based on machine learning, which enables the system to quickly adapt to changes in user needs. The serverless intelligent interaction design of the present application relies on a large language model, allowing users to operate through natural language. This highly flexible interaction method contrasts with the traditional model that requires complex command input, which greatly improves the user experience and task parsing efficiency.

[0047] In an embodiment of the present application, the scheduler module is used to: determine the optimal execution location of the task according to the scheduling scheme generated by the large language model analysis engine module. It should be understood that the scheduler module is the component responsible for the final decision on which computing node the task should be executed on. It determines the optimal execution location of the task based on the scheduling scheme provided by the large language model analysis engine module. The scheduler module is the component responsible for the final decision on which computing node the task should be executed on. It determines the optimal execution location of the task based on the scheduling scheme provided by the large language model analysis engine module. The scheduler module is the component responsible for the final decision on which computing node the task should be executed on. It determines the optimal execution location of the task based on the scheduling scheme provided by the large language model analysis engine module.

[0048] In an embodiment of the present application, the deployment node module is used to: receive the task allocation instruction from the scheduler, and actually deploy the specific task to the specified computing node. It should be understood that the deployment node module is responsible for transferring the task from the scheduling stage to the actual execution stage. It is the bridge from the task being scheduled to the start of operation. Although the scheduler module determines the best execution position of the task, the deployment node module is required to actually execute this decision and assign the task to the specified computing node. Although the scheduler module determines the best execution position of the task, the deployment node module is required to actually execute this decision and assign the task to the specified computing node.

[0049] Accordingly, a specific implementation example scheduling process is provided:

[0050] First, the task management module is responsible for receiving and parsing the user's task request, converting the natural language input by the user into structured data that the system can process. User task requests are usually described in natural language, such as: "Start a data analysis task, requiring the use of at least 4 CPU cores and 16GB of memory." The task parsing process includes the user submitting a natural language task request through the front-end interface or API. Through the large language model technology, the task management module extracts key information and generates structured data. (For example: Task type: data analysis; Resource requirements: 4 CPU cores, 16GB of memory; Time requirement: completed within 2 hours) Specifically, the extracted information is encapsulated as a task structure (Task, Json format) and passed to subsequent modules.

[0051] Secondly, the resource monitoring module is responsible for collecting and updating the resource status of each computing node in the system in real time, such as the usage of CPU, memory, disk, GPU, etc. This information is crucial for task scheduling decisions. The node status includes the current resource usage of each computing node, including CPU usage, memory usage, storage space, network bandwidth, etc. Resource idleness includes whether each node has idle resources available for scheduling and how much idle resources are available. Node health status includes whether the computing node is running normally and whether there is an abnormal state. Data sources include real-time collection of monitoring data through monitoring tools or APIs within the system. For example, use tools such as Prometheus and Zabbix to obtain resource usage. Output real-time node resource status data for subsequent scheduling decisions. The data can be a list containing node ID and resource usage.

[0052] Next, the large language model analysis engine module generates a scheduling plan based on the task requirements extracted by the task management module and the resource status information provided by the resource monitoring module, combined with business logic. The task structure (Task) includes information such as task type, resource requirements, time requirements, priority, etc. Node resource status data includes the resource status of each computing node, such as CPU, memory, storage, GPU, etc. Understand the execution requirements of the task based on the task type, resource requirements (number of CPU cores, memory, storage) and time limit (such as completion within 2 hours). Combined with real-time resource status data (such as idle CPU cores and memory), determine which nodes can meet the resource requirements of the task. If multiple nodes meet the requirements, the model also needs to select nodes based on factors such as node load and historical task execution. Generate a scheduling plan based on the priority, resource requirements and system resource status of the task. The plan includes: selecting a suitable computing node for task execution, ensuring that the resource allocation of the node will not be overloaded, and at the same time meeting the resource requirements of the task. Output scheduling plan data: node ID of the task assignment.

[0053] Next, the scheduler module is responsible for allocating tasks to specific computing nodes for execution according to the scheduling scheme generated by the large language model analysis engine. The task allocation results obtained from the large language model analysis engine. Detailed resource information of the currently available nodes in the system. According to the scheduling scheme, the task is allocated to the specified node. For example, if the task requires the use of 4 CPU cores and 16GB of memory, the scheduler selects a node that meets the resource requirements. Output: Task allocation information: including task ID, assigned computing node ID, task resource requirements, and start time.

[0054] Finally, the deployment node module is responsible for deploying the task to the actual computing node for execution, ensuring that the task can be started and completed as expected. The task allocation plan output by the scheduler module. Detailed information of the node to ensure that the task can run smoothly on the node. According to the specific requirements of the task, create an execution environment (such as starting a virtual machine, container, etc.) on the target computing node and deploy the task. Ensure that the computing resources used by the task (such as CPU, memory, storage, etc.) have been correctly configured. Submit the task to the computing node and start executing the task. Output: After successfully deploying and starting the task, the system will provide the execution status of the task (such as "running", "completed" or "failed").

[0055] In summary, the present invention proposes a serverless intelligent resource scheduling method based on a large language model, aiming to solve the resource scheduling complexity problem in the existing serverless architecture. Traditional static rule schedulers and scheduling systems based on machine learning often show shortcomings such as insufficient flexibility, long learning cycles, and strong data dependence when facing dynamically changing workloads. They are unable to allocate resources in a timely and accurate manner, resulting in resource waste and system performance degradation. By using the powerful semantic understanding and reasoning capabilities of a large language model, the present invention realizes the intelligence and real-time nature of scheduling decisions, effectively copes with complex distributed system environments, improves the accuracy and efficiency of resource scheduling, and ensures that application loads run on appropriate nodes.

[0056] Figure 5 FIG. 1 is a flow chart of a server-free intelligent resource scheduling method based on a large language model according to an embodiment of the present application. Figure 5 As shown, the server-free intelligent resource scheduling method based on the large language model includes: S1, reading computer resource information; S2, monitoring all node resource information; S3, the large language model combines task request resource information and node resource information to generate a scheduling strategy; S4: according to the scheduling strategy, the scheduler deploys the task to the node.

[0057] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A serverless intelligent resource scheduling system based on a large language model, characterized in that: include: The task management module is used to receive and parse the user's task request and convert the natural language input by the user into structured data that can be processed by the system; Resource monitoring module, used to monitor the resource status of each computing node in the system in real time; Large language model analysis engine module, used to analyze task requirements and resource status and generate scheduling plans; The scheduler module is used to assign tasks to specific computing nodes for execution according to the scheduling plan generated by the large language model analysis engine; The deployment node module is used to deploy tasks to actual computing nodes for execution, ensuring that tasks can be started and completed as expected.

2. The serverless intelligent resource scheduling system based on a large language model according to claim 1 is characterized in that: The task management module is used for: the task manager module obtains the task request submitted by the user, and the task management module is responsible for parsing the request and extracting parameter information of the task's computing requirements, memory usage, and storage requirements.

3. The serverless intelligent resource scheduling system based on a large language model according to claim 2 is characterized in that: The task management module includes: transmitting the parameter information to the large language model analysis engine module.

4. The serverless intelligent resource scheduling system based on a large language model according to claim 3 is characterized in that: The resource monitoring module is responsible for collecting and updating the resource status of all nodes in the system in real time, including: CPU, memory, storage and network bandwidth.

5. The serverless intelligent resource scheduling system based on a large language model according to claim 4 is characterized in that: The resource monitoring module includes: regularly reporting the resource utilization and health status of each node to the large language model analysis engine module.

6. The serverless intelligent resource scheduling system based on a large language model according to claim 5, characterized in that: The large language model analysis engine module includes: the large language model analysis engine is the core module of the system. It quickly adapts to environmental changes through semantic analysis of task requirements and node status, context understanding and reasoning, and dynamically adjusts scheduling strategies according to resources in the system.

7. The serverless intelligent resource scheduling system based on a large language model according to claim 6, characterized in that: The large language model analysis engine module includes: The task request parsing and allocation unit is used to receive parameter information from the task management module, use context understanding and reasoning capabilities to understand the execution requirements of the task, and generate a resource scheduling plan; The resource monitoring and dynamic adjustment unit is used to receive resource information from the resource monitoring module and dynamically adjust the resource allocation strategy.

8. The serverless intelligent resource scheduling system based on a large language model according to claim 7, characterized in that: The scheduler module is used to determine the best execution position of the task according to the scheduling plan generated by the large language model analysis engine module.

9. The serverless intelligent resource scheduling system based on a large language model according to claim 8, characterized in that: The deployment node module is used to receive task allocation instructions from the scheduler and actually deploy specific tasks to designated computing nodes.

10. The serverless intelligent resource scheduling method based on a large language model according to claim 9, comprising: S1, read computer resource information; S2, monitors all node resource information; S3, the large language model combines the task request resource information and node resource information to generate a scheduling strategy; S4, according to the scheduling strategy, the scheduler deploys tasks to nodes.

Images