Water conservancy model resource scheduling method and device based on cloud native, and medium

Abstract

The invention discloses a water conservancy model resource scheduling method and device based on cloud native, and a medium, and the method comprises the steps: analyzing a water conservancy model task request, and obtaining a plurality of task tags of a water conservancy model; determining a task response coefficient of each task label according to a pre-constructed mapping relation table; performing priority calculation on the task response coefficient of each task label to obtain a response priority of the water conservancy model task; outputting the water conservancy model task to a task queue; in a resource scheduling period, constructing a global resource map of a cloud edge end; according to the task response priority and the global resource map, distributing part of the high-real-time tasks in the task queue to edge nodes, so as to run the high-real-time tasks through containers deployed by the edge nodes; and running the remaining tasks in the task queue through a container deployed in the cloud. And the resource scheduling response timeliness of the water conservancy model and the cloud edge-end cooperation efficiency are improved.

Description

Technical Field

[0001] The present application relates to the field of water conservancy management technology, and in particular to a cloud-native-based water conservancy model resource scheduling method, equipment, and medium. Background Art

[0002] With the deepening of smart water conservancy construction, water conservancy models are playing a core role in flood forecasting, water resource scheduling, disaster warning, and other scenarios. Traditional water conservancy model tasks usually rely on centralized cloud computing platforms for processing.

[0003] However, water conservancy modeling tasks are complex and diverse (e.g., hydrodynamic calculations, water quality simulations, risk assessments, etc.). Existing systems lack fine-grained task parsing mechanisms, making it impossible to distinguish between high-real-time tasks (e.g., rainstorm and flood peak warnings) and offline analysis tasks. All tasks are uniformly submitted to the cloud for processing, resulting in queue delays for critical tasks due to resource competition, making it difficult to meet the minute-by-minute timeliness requirements of flood season emergency response.

[0004] Heterogeneous cloud-edge resources (high-performance servers in the cloud, lightweight nodes on the edge, and terminal sensor devices) have not yet formed collaborative scheduling capabilities, and the cloud-based centralized processing mode cannot effectively utilize the near-field computing capabilities of edge nodes.

[0005] Furthermore, existing resource allocation strategies are mostly based on static rules and cannot be dynamically adjusted based on the real-time priority of tasks. Especially in the case of sudden water events, traditional scheduling systems struggle to seize resources in a timely manner to execute high-priority tasks, causing key model calculations to lag behind the evolution of the disaster.

[0006] In summary, the water conservancy model task response is delayed and the efficiency of edge-cloud heterogeneous resource collaborative scheduling is low. Summary of the Invention

[0007] The embodiments of the present application provide a cloud-native-based water conservancy model resource scheduling method, device, and medium, which are used to solve the problems of delayed response of water conservancy model tasks and low efficiency of edge-cloud heterogeneous resource collaborative scheduling.

[0008] The embodiments of this application adopt the following technical solutions: On the one hand, an embodiment of the present application provides a cloud-native based water conservancy model resource scheduling method, the method including: parsing the water conservancy model task request to obtain multiple task tags of the water conservancy model; determining the task response coefficient of each task tag according to a pre-built mapping relationship table; performing priority calculation on the task response coefficient of each task tag to obtain the response priority of the water conservancy model task; outputting the water conservancy model task to a task queue; constructing a global resource map of the cloud edge within the resource scheduling cycle; distributing some high real-time tasks in the task queue to the edge node according to the task response priority and the global resource map, so as to run the high real-time tasks through the container deployed on the edge node; and running the remaining tasks in the task queue through the container deployed in the cloud.

[0009] In one example, the task tag includes task type, resource requirement, and data timeliness. The priority calculation of the task response coefficient of each task tag is performed to obtain the response priority of the water conservancy model task, specifically including: calling the priority weight combination of the water conservancy model task; and weighted summing the task response coefficient of each task tag according to the priority weight combination to obtain the response priority coefficient.

[0010] In one example, the method also includes: collecting cloud resource status and edge resource status; the resource status includes node location, computing power type, remaining computing power, and network delay; modeling the cloud resource status and the edge resource status to obtain a global resource map.

[0011] In one example, according to the task response priority and the global resource map, some high real-time tasks in the task queue are distributed to the edge nodes, specifically including: extracting real-time lightweight tasks whose task response priority exceeds the preset priority from the task queue; determining the target real-time lightweight tasks that can be carried by different edge nodes respectively according to the edge node resource status in the global resource map and the load resources of the real-time lightweight tasks; and distributing the target real-time lightweight tasks to the corresponding edge nodes.

[0012] In one example, the method further includes: determining the amount of water conservancy data to be extracted from the cloud message queue based on the task queue; inputting the amount of water conservancy data into a long-short-term memory model to predict the load resource peak in the future; comparing the load resource peak with the global resource map to determine the container instance expansion level; when the container instance expansion level is higher than the preset level, expanding the cloud resources.

[0013] In one example, the load peak is compared with the global resource map to determine the container instance expansion level, specifically including: calculating the resource capacity of the cloud edge according to the global resource map; calculating the ratio between the load resource peak and the resource capacity; mapping the ratio in a preset expansion level mapping table to obtain the container instance expansion level.

[0014] In one example, the method also includes: deploying a lightweight container on the edge node; the lightweight container integrates a data acquisition program, a filtering model, a compression model and an anomaly detection model; collecting water conservancy data according to the data acquisition program; performing anomaly detection on the water conservancy data according to the lightweight anomaly detection model to obtain preprocessed water conservancy data; filtering the preprocessed water conservancy data according to the filtering model to obtain denoised water conservancy data; compressing the denoised water conservancy data according to the compression model to obtain a compressed water conservancy data packet; adding an emergency tag to the compressed water conservancy data packet according to the urgency level of the water conservancy data; sending the compressed water conservancy data packet carrying the emergency tag to the cloud message queue; when the emergency tag meets the emergency warning conditions, performing hidden danger detection on the compressed water conservancy data packet through the hidden danger detection model in the cloud container.

[0015] In one example, the method also includes: when an abnormal task operation log is captured, rebuilding a container at the buffer node and migrating the abnormal water conservancy model task to the reconstructed container; when the abnormal task operation log is related to the resource occupancy rate, determining the compensation coefficient of the abnormal water conservancy model task according to the resource occupancy rate; the compensation coefficient is greater than 1; and compensating the resource demand weight according to the compensation coefficient.

[0016] On the other hand, an embodiment of the present application provides a cloud-native water model resource scheduling device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute any one of the above-mentioned cloud-native water model resource scheduling methods.

[0017] On the other hand, an embodiment of the present application provides a non-volatile computer storage medium for cloud-native water conservancy model resource scheduling, which stores computer-executable instructions, and the computer-executable instructions can execute any of the above-mentioned cloud-native water conservancy model resource scheduling methods.

[0018] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects: Based on the cloud-native architecture, on the basis of deploying containers on both edge nodes and the cloud, the cloud-native scheduling platform can distribute tasks to edge nodes for high-real-time tasks through task classification, global resource map and dynamic pre-allocation mechanism, which can not only save cloud resources but also reduce response delays. This heterogeneous resource orchestration strategy can greatly improve the resource allocation success rate and model training efficiency of centralized emergency tasks (such as heavy rain simulation), especially meeting the timeliness requirements of disaster emergency scenarios and improving the efficiency of cloud-edge collaboration. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solution of the present application, some embodiments of the present application will be described in detail below with reference to the accompanying drawings, in which: Figure 1 A flowchart of a cloud-native-based water conservancy model resource scheduling method provided in an embodiment of the present application; Figure 2 A schematic structural diagram of a cloud-native water conservancy model resource scheduling device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0020] To make the objectives, technical solutions, and advantages of this application more clear, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0021] Some embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0022] Figure 1 This is a flow chart of a cloud-native-based water resource scheduling method provided in an embodiment of the present application. Certain input parameters or intermediate results in this process allow for manual intervention and adjustment to help improve accuracy.

[0023] The implementation of the analysis method involved in the embodiment of the present application can be a terminal device or a server, and the present application does not impose any special restrictions on this. For the convenience of understanding and description, the following embodiments are described in detail using a cloud native scheduling server as an example.

[0024] It should be noted that the server can be a single device or a system composed of multiple devices, that is, a distributed server, and this application does not make any specific restrictions on this.

[0025] It should be noted that cloud native refers to a cloud computing paradigm based on containerization (such as Docker), microservice architecture and dynamic orchestration (such as Kubernetes), which supports the elastic expansion and high availability of water conservancy systems.

[0026] GPU: Graphics Processing Unit, dedicated to parallel computing, is used in this invention to accelerate high-performance computing tasks of large water conservancy models (such as flood simulation).

[0027] LSTM (Long Short-Term Memory) is a time series prediction model used to predict resource requirements for future water conservancy tasks. This involves the process of task decomposition, resource scheduling, and collaborative optimization. Core orchestration models include centralized, decentralized, hierarchical, and federated orchestration models.

[0028] MQTT: A lightweight IoT communication protocol used for data transmission between edge nodes and the cloud.

[0029] PPO (Proximal Policy Optimization): Proximal Policy Optimization algorithm, a reinforcement learning method, is used to train the decision model of the elastic scheduling engine.

[0030] Elastic Scaling: Dynamically adjust the number of container instances (scaling) based on real-time load, such as automatically scaling GPU nodes when there is a surge in heavy rain forecasting tasks.

[0031] Resource profiling: Model the real-time status of cloud-edge resources (computing power, network, storage) to form a global visual resource map.

[0032] Figure 1 The process in includes the following steps: S101: Parse the water conservancy model task request to obtain multiple task tags of the water conservancy model.

[0033] Task tags can include task type, resource requirements, data timeliness, etc. For example, a water conservancy model task is a rainstorm flood diffusion simulation task, the task type is emergency flood simulation, the resource requirements are water level data and weather radar data, and the data timeliness is 2 hours (i.e., the results must be output within 2 hours).

[0034] It should be noted that the task type is related to the task content and the urgency of the task.

[0035] S102: Determine the task response coefficient of each task label according to the pre-built mapping relationship table.

[0036] It should be noted that the mapping relationship table includes the correspondence between different water conservancy model tasks and the content range of each task label, as well as the correspondence between different task label content ranges and task response coefficients.

[0037] That is, in the mapping relationship table, the water conservancy model task is searched to obtain the matching water conservancy model task. Then, in the matching water conservancy model task record, the content of each task tag is matched to obtain the corresponding task response coefficient.

[0038] S103: Calculate the priority of the task response coefficient of each task tag to obtain the response priority of the water conservancy model task.

[0039] In some embodiments of the present application, the priority calculation process is as follows: The priority weight combination for calling the water conservancy model task.

[0040] The task response coefficient of each task label is weighted and summed according to the priority weight combination to obtain the response priority coefficient.

[0041] It should be noted that each water conservancy model task has its own priority weight combination, and the priority weight combination can be dynamically optimized.

[0042] For example, the priority calculation formula is as follows:

[0043] in, is the response priority coefficient, is the task response coefficient of the task type, is the task response coefficient of resource demand, is the task response coefficient of data timeliness, is the weight of the task type, is the weight of resource demand, is the weight of data timeliness.

[0044] It should be noted that 、 and It is an adjustable parameter that can be dynamically optimized based on the execution results of historical water conservancy model tasks.

[0045] S104: Output the water conservancy model task to a task queue.

[0046] S105: During the resource scheduling cycle, a global resource map of the cloud and edge is constructed.

[0047] It should be noted that the resource scheduling cycle refers to the regular reconstruction of the global resource map of the cloud edge and the resource scheduling of tasks in the task queue.

[0048] In some embodiments of the present application, the process of constructing a global resource map is as follows: First, the cloud resource status and edge resource status are collected. The resource status includes node location, computing power type, remaining computing power, and network latency.

[0049] For example, computing power types include CPU general computing power, GPU accelerated computing power, etc.

[0050] Then, the cloud resource status and edge resource status are modeled to obtain a global resource map.

[0051] For example, Prometheus can be used to collect the CPU / GPU utilization, memory usage, network bandwidth, and remaining computing power of edge nodes (such as hydrological station servers) of the cloud Kubernetes cluster to generate a global resource status map (JSON format) including node location, computing power type, current load, and network latency.

[0052] S106: Distribute some high real-time tasks in the task queue to edge nodes according to the task response priority and the global resource map, so as to run the high real-time tasks through containers deployed on the edge nodes.

[0053] In some embodiments of the present application, based on the concept of heterogeneous resource orchestration, the distribution process is as follows: First, real-time lightweight tasks whose task response priority exceeds the preset priority are extracted from the task queue.

[0054] Then, according to the edge node resource status in the global resource map and the load resources of the real-time lightweight task, the target real-time lightweight tasks that different edge nodes can carry are determined.

[0055] It should be noted that when determining the edge nodes to which the target real-time lightweight tasks are distributed, the main consideration is whether the computing power type of the edge nodes matches the load resource type, and the remaining computing power that can be used is higher than the load resource.

[0056] Finally, the target real-time lightweight tasks are distributed to the corresponding edge nodes.

[0057] For example, for high-performance real-time tasks (such as real-time analysis of hydrological data), the scheduling platform distributes these tasks to edge nodes and binds model training tasks to cloud-based GPU node tags. A streamlined image (Alpine Linux base image + model inference framework) is built for edge nodes, reducing the image size to less than 50MB, ensuring rapid deployment in low-bandwidth environments.

[0058] S107: Run the remaining tasks in the task queue through the container deployed in the cloud.

[0059] In some embodiments of the present application, the intelligent agent is trained based on the PPO algorithm, the state space includes resource profiles and task queues, and the action space is the container instance expansion and contraction instructions.

[0060] The LSTM model can be used to predict task load trends within the future (for example, 1 hour). If the predicted peak value exceeds 80% of the current resource capacity, capacity expansion is triggered in advance.

[0061] Based on this, the amount of water conservancy data to be extracted from the cloud message queue is determined according to the task queue.

[0062] The water conservancy data volume is input into the long short-term memory model to predict the load resource peak in the future.

[0063] Compare the peak load resource value with the global resource map to determine the container instance scaling level.

[0064] When the container instance expansion level exceeds the preset level, the cloud resources are expanded.

[0065] The process of comparing the load resource peak with the global resource map is as follows: First, based on the global resource map, the resource capacity of the cloud-edge is calculated, that is, the remaining computing power capacity.

[0066] Then, the ratio between the peak load resource and the resource capacity is calculated.

[0067] In the preset expansion level mapping table, the comparison value is mapped to obtain the container instance expansion level.

[0068] It should be noted that the smaller the ratio, the lower the container instance scalability level. This is because the smaller the ratio, the more abundant the resource capacity is relative to the peak load resource.

[0069] It should be noted that after the hydraulic model task is completed, the expansion container can be recycled.

[0070] In some embodiments of the present application, since existing water conservancy edge scenarios (such as reservoirs and sluice stations) are often limited by network bandwidth fluctuations, the massive sensor data transmitted back to the cloud further aggravates the response delay. Therefore, the present application denoises the sensor data, and the data in the cloud message queue is uploaded through the edge node.

[0071] First, a lightweight container is deployed on the edge node, which integrates the data collection program, filtering model, compression model, and anomaly detection model.

[0072] Collect water conservancy data according to the data collection procedures.

[0073] Anomaly detection is performed on water conservancy data based on the lightweight anomaly detection model to obtain preprocessed water conservancy data.

[0074] The pre-processed water conservancy data is filtered according to the filtering model to obtain denoised water conservancy data.

[0075] The denoised water conservancy data is compressed according to the compression model to obtain a compressed water conservancy data packet.

[0076] According to the urgency level of the water conservancy data, an emergency tag is added to the compressed water conservancy data packet.

[0077] Send the compressed water conservancy data packet carrying the emergency tag to the cloud message queue.

[0078] When the emergency tag meets the emergency warning conditions, the hidden danger detection model in the cloud container is used to detect hidden dangers in the compressed water conservancy data packet.

[0079] In other words, for edge data collection: deploy lightweight containerized agents (based on Docker) at each hydrological station, integrate data filtering and compression algorithms, and collect sensor data such as water level, rainfall, and flow rate in real time.

[0080] For preprocessing: Run lightweight anomaly detection models (such as TinyML) on edge nodes to filter out noisy data and reduce cloud computing load.

[0081] For data transmission: water conservancy data is sent to a cloud message queue (such as Kafka) via the MQTT protocol, and the data packet carries metadata tags (such as emergency level: water level exceeds the warning line).

[0082] In some embodiments of the present application, the optimization process of the water conservancy model weight combination is as follows: When the abnormal task operation log is captured, the container is rebuilt on the buffer node and the abnormal water conservancy model task is migrated to the rebuilt container.

[0083] When the task operation abnormality log is related to the resource utilization rate, the compensation coefficient of the abnormal water conservancy model task is determined based on the resource utilization rate; the compensation coefficient is greater than 1. The higher the resource utilization rate, the higher the compensation coefficient.

[0084] The resource demand weight is compensated according to the compensation coefficient, wherein the compensation coefficient is multiplied by the resource demand weight to obtain the compensated resource demand weight.

[0085] That is to say, the actual resources occupied during task operation are relatively higher, that is, the resource demand weight of the water conservancy model task is increased.

[0086] For example, real-time monitoring of task execution status (such as latency, resource utilization, and error rate) and extraction of abnormal events (such as "GPU node OOM error") through log analysis can capture task anomalies.

[0087] It should be noted that the reinforcement learning PPO model can be retrained regularly based on the task operation log information, and the weight parameter value in the priority formula can be dynamically adjusted.

[0088] It should be noted that although the embodiments of this application are based on Figure 1 Steps S101 to S107 are described in sequence, but this does not mean that steps S101 to S107 must be performed in a strict order. Figure 1 The order shown in FIG1 is to introduce and explain steps S101 to S107 in order to facilitate those skilled in the art to understand the technical solutions of the embodiments of the present application. In other words, in the embodiments of the present application, the order of steps S101 to S107 can be appropriately adjusted according to actual needs.

[0089] In the prior art, there are the following problems: In summary, the system includes a data perception layer, a cloud-native scheduling module, and a dynamic feedback module. The task parsing module prioritizes water conservancy model tasks, and the resource profiling module monitors the cloud-edge resource status in real time. A heterogeneous resource orchestrator is used to coordinate the scheduling of edge nodes and cloud GPU clusters. Furthermore, an improved elastic scheduling engine (integrating reinforcement learning and time series prediction algorithms) is used to generate container instance scaling policies. The core components include task hierarchical pre-allocation, a lightweight edge-cloud resource scheduling mechanism, and an elastic scaling algorithm. The specific solutions are as follows: 1. Based on the cloud-native architecture, with containers deployed on both edge nodes and the cloud, the cloud-native scheduling platform distributes high-real-time tasks to edge nodes through task classification, a global resource map, and a dynamic pre-allocation mechanism. This not only saves cloud resources but also reduces response latency. This heterogeneous resource orchestration strategy significantly improves the resource allocation success rate and model training efficiency for centralized emergency tasks (such as heavy rain simulation), especially meeting the timeliness requirements of disaster emergency scenarios.

[0090] However, in existing technologies, the existing cloud platform resource scheduling strategies are not optimized for the task characteristics of large water conservancy models (such as strong data dependency and phased computing). The general container scaling mechanism is prone to cause resource competition or idleness, and it is impossible to dynamically adjust resource allocation according to task priority.

[0091] Water conservancy business involves multiple types of computing resources such as edge nodes (hydrological stations, sensors) and cloud-based GPU clusters. Existing technologies lack a unified dynamic orchestration strategy, resulting in the inability to dispatch high-real-time tasks to nearby edge devices. Offline training tasks occupy edge resources, resulting in low overall collaborative efficiency.

[0092] Traditional methods fail to establish real-time correlation between water conservancy task characteristics (such as data timeliness and model calculation complexity) and resource status (heterogeneous computing power utilization and network latency), resulting in delayed scheduling decisions and an inability to meet the needs of rapid and flexible response in disaster emergency scenarios.

[0093] 2. The expansion method of the existing technology is to manually add servers. This application automatically expands and shrinks container instances based on a multi-objective elastic scheduling algorithm to expand cloud resources, improve cloud resource utilization, and reduce idle resource waste through a dynamic recycling mechanism. That is, through a hybrid algorithm of time series prediction and reinforcement learning, the system can predict resource demand in advance (such as peak load in the rainy season), dynamically adjust the expansion strategy, and avoid the lag of the traditional threshold trigger mechanism.

[0094] In existing technologies, traditional resource scheduling systems use static rules or simple threshold trigger mechanisms, which cannot adapt to the sudden and highly concurrent computing needs in water conservancy business scenarios (such as flood prediction and disaster simulation), resulting in delayed or redundant resource supply and difficulty in balancing real-time response and resource costs.

[0095] 3. When a task fails, it can automatically rebuild the container and has self-healing capabilities, eliminating the need for manual intervention or expensive backup servers. Existing technologies require manual fault detection and DNS switching.

[0096] 4. Compatible with multiple types of water conservancy models such as hydrology, meteorology, and ecology, and supports cross-model task dependency scheduling (such as completing weather forecasts before triggering flood simulations).

[0097] Based on the same idea, some embodiments of the present application also provide devices and non-volatile computer storage media corresponding to the above methods.

[0098] Figure 2 A schematic diagram of the structure of a cloud-native water conservancy model resource scheduling device provided in an embodiment of the present application includes: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute any one of the above-mentioned cloud-native-based water conservancy model resource scheduling methods.

[0099] Some embodiments of the present application provide a non-volatile computer storage medium for cloud-native water model resource scheduling, which stores computer-executable instructions. The computer-executable instructions can execute any of the above-mentioned cloud-native water model resource scheduling methods.

[0100] The various embodiments in this application are described in a progressive manner. Similar portions between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences between the other embodiments. In particular, the device and medium embodiments are generally similar to the method embodiments, so their descriptions are relatively simple. For relevant portions, refer to the descriptions of the method embodiments.

[0101] The devices and media provided in the embodiments of the present application correspond one-to-one to the methods. Therefore, the devices and media also have similar beneficial technical effects to their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0102] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0103] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0104] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0105] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0106] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0107] Memory may include non-permanent storage in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0108] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can be implemented using any method or technology for information storage. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change RAM (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media, such as modulated data signals and carrier waves.

[0109] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0110] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modifications, equivalent replacements, improvements, etc. made within the technical principles of the present application should fall within the scope of protection of the present application.

Claims

1. A cloud-native water conservancy model resource scheduling method, characterized in that: The method comprises: Parse the water conservancy model task request to obtain multiple task tags of the water conservancy model; Determine the task response coefficient of each task label based on the pre-built mapping relationship table; Calculate the priority of the task response coefficient of each task label to obtain the response priority of the water conservancy model task; Outputting the water conservancy model task to a task queue; During the resource scheduling cycle, a global resource map of the cloud, edge, and end is constructed; Distribute some of the high-real-time tasks in the task queue to edge nodes based on task response priorities and the global resource map, so as to run the high-real-time tasks through containers deployed on the edge nodes; Run the remaining tasks in the task queue through containers deployed in the cloud.

2. The method according to claim 1, characterized in that The task tag includes the task type, resource requirement, and data timeliness. The priority calculation of the task response coefficient of each task tag is performed to obtain the response priority of the water conservancy model task, specifically including: Calling the priority weight combination of the water conservancy model task; The task response coefficient of each task label is weighted and summed according to the priority weight combination to obtain the response priority coefficient.

3. The method according to claim 1, characterized in that The method further comprises: Collect cloud resource status and edge resource status; resource status includes node location, computing power type, remaining computing power, and network latency; The cloud resource status and the edge resource status are modeled to obtain a global resource map.

4. The method according to claim 3, characterized in that The method of distributing some high-real-time tasks in the task queue to edge nodes according to the task response priority and the global resource map specifically includes: Extracting, from the task queue, real-time lightweight tasks whose task response priority exceeds a preset priority; Determine the target real-time lightweight tasks that different edge nodes can carry based on the edge node resource status in the global resource map and the load resources of the real-time lightweight tasks; Distribute the target real-time lightweight tasks to the corresponding edge nodes.

5. The method according to claim 1, wherein The method further comprises: Determining the amount of water conservancy data to be extracted from the cloud message queue based on the task queue; Inputting the water conservancy data into a long short-term memory model to predict the peak load resource in the future; Comparing the load resource peak with the global resource map to determine the container instance expansion level; When the container instance expansion level exceeds the preset level, the cloud resources are expanded.

6. The method according to claim 5, characterized in that Comparing the load peak with the global resource map to determine the container instance expansion level specifically includes: Calculate the resource capacity of the cloud edge based on the global resource map; Calculating a ratio between the load resource peak and the resource capacity; The ratio is mapped in a preset expansion level mapping table to obtain the container instance expansion level.

7. The method according to claim 1, characterized in that The method further comprises: The edge node deploys a lightweight container; the lightweight container integrates a data collection program, a filtering model, a compression model, and an anomaly detection model; Collect water conservancy data according to the data collection procedure; Performing anomaly detection on the water conservancy data according to a lightweight anomaly detection model to obtain preprocessed water conservancy data; Filtering the pre-processed water conservancy data according to the filtering model to obtain denoised water conservancy data; compressing the denoised water conservancy data according to a compression model to obtain a compressed water conservancy data packet; adding an emergency tag to the compressed water conservancy data packet according to the emergency level of the water conservancy data; Send the compressed water conservancy data packet with the emergency tag to the cloud message queue; When the emergency tag meets the emergency warning conditions, the hidden danger detection model in the cloud container is used to detect hidden dangers in the compressed water conservancy data packet.

8. The method according to claim 1, characterized in that The method further comprises: When an abnormal task operation log is captured, the container is rebuilt on the buffer node and the abnormal water conservancy model task is migrated to the rebuilt container; When the task operation abnormality log is related to the resource occupancy rate, a compensation coefficient of the abnormal water conservancy model task is determined according to the resource occupancy rate; the compensation coefficient is greater than 1; The resource demand weight is compensated according to the compensation coefficient.

9. A cloud-native water conservancy model resource scheduling device, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the cloud-native water conservancy model resource scheduling method described in any one of claims 1-8.

10. A non-volatile computer storage medium for cloud-native water resource model scheduling, storing computer-executable instructions, characterized in that: The computer-executable instructions can execute the cloud-native-based water conservancy model resource scheduling method described in any one of claims 1 to 8.

Images