Adaptive plug-in scheduling and dynamic optimization evolution method and system

Abstract

The invention relates to the technical field of plug-in resource management, and discloses an adaptive plug-in scheduling and dynamic optimization evolution method and system, and the method comprises the steps: obtaining a plug-in task generated by an edge calculation module based on a target device real-time monitoring index data flow; monitoring running state data of the plug-in system in real time and preprocessing the running state data; carrying out resource allocation on the plug-in task by utilizing a reinforcement learning algorithm based on the preprocessed data, and generating a plug-in task resource allocation result; performing plug-in resource allocation on the plug-in task and the plug-in task resource allocation result based on a virtualized resource pool mode to generate a plug-in resource allocation result; and generating a plug-in code based on a plug-in resource allocation result by utilizing a model-driven development and automatic code generation technology, and running a feedback result to optimize and evolve a plug-in. Through an intelligent and adaptive mechanism, the real-time response capability, the performance, the expandability and the sustainable development capability of the plug-in are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of plug-in resource management, and in particular to a method and system for adaptive plug-in scheduling and dynamic optimization evolution. Background Art

[0002] With the development of industrial automation and the Internet of Things (IoT), dynamic plug-in mechanisms have become a key technology in industrial data collection and edge computing. However, while existing dynamic plug-in mechanisms support dynamic loading and unloading of plug-ins, they often face performance bottlenecks or resource waste in complex industrial environments. Furthermore, edge computing and dynamic plug-in mechanisms often function independently, failing to achieve effective collaborative optimization. This results in inefficient data processing and scalability. Traditional plug-in development methods also suffer from long development cycles, prone to errors, and difficulty adapting quickly to changing requirements. Summary of the Invention

[0003] In view of this, the present invention provides an adaptive plug-in scheduling and dynamic optimization evolution method and system to solve the problems of existing plug-ins having poor efficiency in data processing, long development cycle, and difficulty in quickly adapting to changing needs.

[0004] In a first aspect, the present invention provides a method for adaptive plug-in scheduling and dynamic optimization evolution, the method comprising:

[0005] Obtain the plug-in task generated by the edge computing module based on the target device's real-time monitoring indicator data stream;

[0006] Monitor the running status data of the plug-in system in real time and pre-process the running status data;

[0007] Based on the pre-processed running status data, the reinforcement learning algorithm is used to allocate resources for the plug-in tasks and generate the plug-in task resource allocation results;

[0008] Perform plug-in resource allocation on the plug-in tasks and the plug-in task resource allocation results based on the virtualized resource pool method to generate a plug-in resource allocation result;

[0009] Based on the plug-in resource allocation results, the plug-in code is generated using model-driven development and automatic code generation technology, and the plug-in is optimized and evolved according to the plug-in code running feedback results.

[0010] The adaptive plug-in scheduling and dynamic optimization evolution method provided by the embodiment of the present invention obtains the plug-in tasks generated by the edge computing module based on the real-time monitoring indicator data stream of the target device. The entire system can closely match the actual operating status of the target device to trigger the corresponding plug-in tasks. The reinforcement learning algorithm can learn the optimal resource allocation strategy based on the constantly changing state of the plug-in system and past experience, which can not only ensure that important plug-in tasks obtain sufficient resources to run efficiently, but also avoid idle or over-allocation of resources, thereby improving the resource utilization of the entire plug-in system, optimizing system performance, and ensuring stable and efficient operation of the system in a complex and changing environment; the virtualized resource pool abstracts and integrates physical resources, and can flexibly and on-demand allocate the required resources to different plug-ins according to the plug-in tasks and the generated resource allocation results; with the help of model-driven development and automatic code generation technology, plug-in code is generated based on the plug-in resource allocation results, and the plug-in is optimized and evolved according to the feedback results of the plug-in code operation, realizing the self-update and improvement of the plug-in system. This adaptive plug-in scheduling and dynamic optimization evolution method works synergistically across multiple links, from task generation, system status monitoring, resource allocation, resource deployment to plug-in optimization evolution, enabling it to better cope with complex and changeable actual application scenarios, and significantly improving the plug-in's real-time response capabilities, performance, scalability, and sustainable development capabilities.

[0011] In an optional embodiment, the real-time monitoring of the plug-in system's operating status data and preprocessing the operating status data include:

[0012] Real-time monitoring of the plug-in system's operating status data, including: each plug-in's CPU usage, memory usage data, I / O latency data, number of pending tasks, task priority, network bandwidth usage, and battery consumption rate;

[0013] Using a preset dynamic filtering algorithm to remove noise from the operating status data to obtain pre-processed operating status data;

[0014] When the pre-processed operating status data exceeds the corresponding preset threshold, an abnormal alarm signal is generated.

[0015] By monitoring various data such as the CPU usage and memory usage of each plug-in, the embodiment of the present invention can fully understand the operating status of the plug-in system from multiple dimensions, thereby serving as the basis for subsequent resource allocation and system optimization. The pre-processed operating status data is more reliable, making the analysis and decision-making based on this data more accurate. When the pre-processed operating status data exceeds the corresponding preset threshold, an abnormal alarm signal is generated, which can promptly detect problems that may occur in the plug-in system. By promptly detecting and handling abnormal situations, the plug-in system can maintain a more reliable and stable operating state.

[0016] In an optional embodiment, allocating resources to plug-in tasks using a reinforcement learning algorithm based on the pre-processed running status data to generate a plug-in task resource allocation result includes:

[0017] Obtain the number of pending tasks, task priority, current network bandwidth usage, CPU usage, and memory usage data of each plug-in after preprocessing, which is used to describe the state space of the reinforcement learning algorithm;

[0018] Set the set of actions that the plug-in takes, including: adjusting the plug-in's CPU usage, memory usage data, dynamically changing the priority of pending tasks, and limiting or increasing network bandwidth;

[0019] Set reward rules, including: the shorter the time required to complete the task, the higher the reward; the higher the resource utilization rate, the higher the reward; the higher the completion rate of priority tasks, the higher the reward;

[0020] Initialize the state space and train the preset reinforcement learning algorithm. During training, dynamically adjust the learning rate to generate the optimal plug-in task resource allocation result.

[0021] The embodiment of the present invention obtains the dynamically changing running status number of the plug-in system in real time and updates the state space according to the preprocessed data, so that the algorithm can always keep up with system changes and adjust the resource allocation strategy in time to ensure that resources can be reasonably allocated under different conditions such as system load and task urgency, so that the plug-in system can run efficiently.

[0022] In an optional implementation, performing plug-in task resource allocation on the plug-in task and the plug-in resource allocation result based on the virtualized resource pool to generate the plug-in task resource allocation result includes:

[0023] Merge all physical computing resources into a unified resource pool and divide the physical resources into multiple virtual resource units;

[0024] Allocating virtual resource units to the edge computing modules and plug-ins that apply for resources based on the priority of the tasks;

[0025] Input the plug-in tasks generated by the real-time edge computing module into the preset load forecasting model to obtain the load forecast results for the future preset time period, and generate the resource allocation results of the plug-in system based on the load forecast results;

[0026] The resource allocation results of the plug-in system generated based on the load prediction results are compared with the resource monitoring data of the real-time plug-in system. The allocation strategy is adjusted according to the comparison results. When a plug-in fails, its corresponding tasks are transferred to other plug-ins for processing.

[0027] The embodiment of the present invention merges all physical computing resources into a unified resource pool, realizes the centralized management and abstract integration of hardware resources, and allocates virtual resource units to the edge computing modules and plug-ins applying for resources based on the priority of the task, so that resource allocation can be carried out closely around the importance and urgency of the task. The plug-in tasks generated by the real-time edge computing module are input into the preset load prediction model to obtain the load prediction results within the future preset time period, and based on this, the resource allocation results of the plug-in system are generated, which provides the system with forward-looking resource planning capabilities, and can reasonably arrange resources according to future load conditions, avoiding over-configuration or under-configuration of resources. Through real-time comparison and adjustment, the system can timely discover these changes and quickly reallocate resources to ensure that resources are always in a reasonable allocation state and maintain stable and efficient operation of the system. When a plug-in fails, its corresponding tasks are transferred to other plug-ins for processing, thereby enhancing the fault tolerance of the system.

[0028] In an optional implementation, the training process of the preset load prediction model includes:

[0029] Obtain historical load data, resource usage patterns, task types, time information, and external factors affecting resource consumption, wherein the historical load data includes: CPU usage, memory usage, I / O latency data, and battery consumption rate;

[0030] Cut historical load data into time windows, use a sliding window approach to generate training samples, and extract feature data from the training samples, including timestamps, task types, and external factors that affect resource consumption.

[0031] The feature data is input into the LSTM network for training, and the trained model is used as the load prediction model. Its output results include: CPU usage, memory usage, I / O delay data, and battery consumption rate within a preset time period in the future.

[0032] The embodiment of the present invention obtains historical load data covering CPU usage, memory occupancy, I / O delay data, battery consumption rate, resource usage pattern, task type, time information and external factors affecting resource consumption, so that the data used to train the model is extremely comprehensive. The feature data is input into the LSTM for training, which fully utilizes the advantage of the LSTM network in processing time series data, so that it can deeply learn the resource consumption patterns of the system under various complex situations, and can significantly improve the accuracy and reliability of load prediction, thereby helping the plug-in system to manage resources more scientifically and enhancing the plug-in system's ability to cope with complex and changing task scenarios.

[0033] In an optional embodiment, plug-in code is generated based on the plug-in resource allocation result using model-driven development and automatic code generation technology, and the plug-in is optimized and evolved according to the plug-in code operation feedback result, including:

[0034] Generate plug-in code based on the plug-in resource allocation results using model-driven development and automatic code generation technology, and automatically compile, test, and deploy it;

[0035] When detecting plug-in version updates and function improvement requirements, use continuous integration and continuous delivery tools to perform plug-in updates, automated testing, and regression testing, and automatically optimize test cases based on the test results;

[0036] When the plug-in code running feedback result is monitored, the running data is analyzed based on the feedback result to optimize the plug-in resource allocation.

[0037] The embodiment of the present invention utilizes model-driven development and automatic code generation technology based on the plug-in resource allocation result to generate plug-in code and automatically compile, test and deploy it, which greatly shortens the plug-in development cycle. The automatic compilation and testing process can timely discover syntax errors, logical loopholes and incompatibility with the system environment in the code, and correct them at an early stage to avoid problems being carried over into subsequent actual operation. In addition, the continuous optimization mechanism based on feedback enables the entire plug-in system to continuously self-adjust and evolve according to the actual operation situation, ensuring that the plug-in system always maintains good adaptability and competitiveness during long-term operation, and promoting the development of the entire system in a more efficient and stable direction.

[0038] In a second aspect, the present invention provides an adaptive plug-in scheduling and dynamic optimization evolution system, comprising:

[0039] The plug-in task acquisition module is used to obtain the plug-in tasks generated by the edge computing module based on the real-time monitoring indicator data stream of the target device;

[0040] The plug-in task resource allocation module is used to allocate resources for plug-in tasks based on the pre-processed running status data using the reinforcement learning algorithm and generate plug-in task resource allocation results;

[0041] Real-time operation data monitoring module, used to monitor the operation status data of the plug-in in real time and pre-process the operation status data;

[0042] A plug-in resource allocation module is used to allocate plug-in task resources based on the virtualized resource pool method and generate a plug-in resource allocation result.

[0043] The plug-in automatic generation and optimization module is used to generate plug-in code based on the plug-in resource allocation results using model-driven development and automatic code generation technology, and optimize and evolve the plug-in according to the plug-in code operation feedback results.

[0044] In a third aspect, the present invention provides a computer device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to thereby execute the adaptive plug-in scheduling and dynamic optimization evolution method of the above-mentioned first aspect or any corresponding embodiment thereof.

[0045] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the adaptive plug-in scheduling and dynamic optimization evolution method of the above-mentioned first aspect or any corresponding embodiment thereof.

[0046] In a fifth aspect, the present invention provides a computer program product comprising computer instructions, the computer instructions being used to enable a computer to execute the adaptive plug-in scheduling and dynamic optimization evolution method of the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0048] Figure 1 1 is a flow chart of a method for adaptive plug-in scheduling and dynamic optimization evolution according to an embodiment of the present invention;

[0049] Figure 2 is a flowchart of another adaptive plug-in scheduling and dynamic optimization evolution method according to an embodiment of the present invention;

[0050] Figure 3 is a structural block diagram of an adaptive plug-in scheduling and dynamic optimization evolution system according to an embodiment of the present invention;

[0051] Figure 4 Schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0052] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of the present invention.

[0053] In order to achieve more efficient and intelligent industrial data processing and improve the scalability of the plug-in system, an embodiment of the present invention provides an embodiment of an adaptive plug-in scheduling and dynamic optimization evolution method. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0054] This embodiment provides an adaptive plug-in scheduling and dynamic optimization evolution method. Figure 1 FIG. 1 is a flow chart of a method for adaptive plug-in scheduling and dynamic optimization evolution according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0055] S101, obtaining a plug-in task generated by the edge computing module based on the real-time monitoring indicator data stream of the target device.

[0056] Specifically, in the industrial Internet of Things scenario, the edge computing module can collect sensor data in real time. Changes in this data will trigger the generation of plug-in tasks, which can ensure that the plug-in system can perceive changes in device status in real time, respond quickly, and quickly process tasks closely related to device status, ensuring that plug-in tasks are accurately matched to the current needs of the device, thereby effectively improving the timeliness of response to changes in the device operating environment, and avoiding problems such as reduced equipment operating efficiency or failure to handle faults in a timely manner due to unreasonable task arrangements.

[0057] S102: monitor the running status data of the plug-in system in real time and pre-process the running status data.

[0058] Specifically, the embodiment of the present invention performs data analysis and anomaly detection by collecting key performance indicators such as CPU usage, memory usage data, I / O delay data, number of pending tasks, task priority, network bandwidth usage, battery consumption rate, etc. of each plug-in.

[0059] Among them, by monitoring the CPU usage, memory occupancy, etc. of each plug-in, resource allocation can be dynamically adjusted according to the actual needs of the plug-in; by analyzing the network bandwidth usage, network bottlenecks can be discovered and optimized; analysis of I / O delay data helps to optimize memory management and I / O operations, thereby improving the overall operating efficiency of the system; task priority, for example, real-time monitoring is more important than batch log analysis. The data source is that when the system is deployed and initialized, the administrator or developer will pre-define the priority of each task based on the nature of the task, business needs and equipment capabilities. The system configuration file (such as configuration file, database table, environment variables, etc.) contains task priority information. The priority may be classified according to the type of task, for example: high-priority tasks are critical tasks (such as fault handling, real-time monitoring, data collection); medium-priority tasks include scheduled tasks and maintenance tasks; low-priority tasks do not include background data processing, logging and other tasks that do not affect the real-time performance of the system.

[0060] For battery-powered devices, battery drain rate is a key metric. By analyzing the relationship between battery drain rate and other performance indicators, we can optimize plugin execution strategies to reduce battery consumption. For example, if we find that a plugin significantly increases battery drain during execution, we can adjust its execution frequency or optimize its algorithm to reduce unnecessary energy consumption and thus extend the device's battery life.

[0061] By real-time monitoring of the plug-in system's operating status data and preprocessing it, the preprocessed operating status data becomes more reliable, making analysis and decision-making based on this data more accurate. When the preprocessed operating status data exceeds the corresponding preset threshold, an abnormal alarm signal is generated, which can promptly detect possible problems in the plug-in system. By promptly detecting and handling abnormal situations, the plug-in system can maintain a more reliable and stable operating state.

[0062] S103 , allocating resources to the plug-in task using a reinforcement learning algorithm based on the pre-processed running status data, and generating a plug-in task resource allocation result.

[0063] The embodiment of the present invention utilizes a reinforcement learning algorithm to allocate resources for plug-in tasks based on pre-processed operating status data, and can dynamically adjust resources according to the actual situation of the system. The reinforcement learning algorithm can learn the optimal resource allocation strategy to adapt to different plug-in task requirements and system operating status, which helps to improve resource utilization and reduce resource waste. For example, there is a multi-task image recognition plug-in system, including image recognition tasks of different resolutions. When it is detected that the CPU resources are tight, the reinforcement learning algorithm can, based on historical experience and current operating status data, give priority to allocating resources to low-resolution, high-real-time image recognition plug-in tasks, and suspend or reduce resource allocation for high-resolution, time-insensitive tasks, thereby ensuring the overall performance of the system.

[0064] S104: Perform plug-in resource allocation on the plug-in task and the plug-in task resource allocation result based on a virtualized resource pool to generate a plug-in resource allocation result.

[0065] The present invention allocates plug-in resources based on a virtualized resource pool, providing a flexible resource allocation mechanism. Virtualization technology can abstract physical resources (multiple servers or cloud resources) into a virtual resource pool, which can be flexibly allocated according to plug-in tasks and resource allocation results. This enables the plug-in system to better adapt to the dynamic changes of plug-in tasks. For example, a plug-in system in a cloud computing environment can virtually divide the server's CPU, memory, storage and other resources through a virtualized resource pool. When a new plug-in task requires a large amount of memory resources, sufficient virtual memory can be dynamically allocated from the resource pool to the plug-in without affecting the normal operation of other plug-ins. It also facilitates the recycling and reallocation of resources according to the life cycle of the plug-in.

[0066] S105 , generating plug-in code using model-driven development and automatic code generation technology based on the plug-in resource allocation result, and optimizing and evolving the plug-in according to the plug-in code running feedback result.

[0067] The embodiments of the present invention greatly reduce the workload and time cost of manual code writing through model-driven development and automatic code generation technology; when it is necessary to add new functions to the plug-in, modify existing functions or adjust resource requirements, it is only necessary to modify the model accordingly and then regenerate the code, without having to start from scratch for large-scale code writing and debugging work, which greatly shortens the development cycle and enables the plug-in to meet new business needs in a timely manner. By analyzing the plug-in operation feedback data and the feedback-based optimization evolution process, the performance of the plug-in can be adjusted in real time to always adapt to changes in the system's operating environment and business needs. For example, when the system load increases, by optimizing the plug-in code and resource allocation strategy, it is ensured that the performance of the plug-in will not show a significant decline, maintaining the overall stability and reliability of the system.

[0068] The adaptive plug-in scheduling and dynamic optimization evolution method provided by the embodiment of the present invention works synergistically across multiple links, including plug-in task generation, system status monitoring, resource allocation, resource deployment, and plug-in optimization evolution. Through this intelligent and adaptive mechanism, it can better cope with complex and changeable actual application scenarios, significantly improving the real-time response capability, performance, scalability, and sustainable development capabilities of the plug-in.

[0069] The embodiment of the present invention also provides another adaptive plug-in scheduling and dynamic optimization evolution method, such as Figure 2 As shown, the following steps are included:

[0070] S201, obtaining a plug-in task generated by the edge computing module based on the real-time monitoring indicator data stream of the target device.

[0071] In one embodiment, the target device is an industrial device, and the edge computing module collects real-time data, including temperature, pressure, vibration, and other data, and analyzes and generates analysis results. The data analysis process includes:

[0072] A1. Preprocessing: Use a preset filtering algorithm to remove noise from the monitoring data and then standardize the data;

[0073] A2. Feature extraction and trend analysis: Extract key features of monitoring data, such as the average, maximum, and minimum values of temperature, as well as fluctuations in pressure and vibration;

[0074] A3. Output analysis results: The processed and analyzed data, such as trend prediction results, abnormal status, etc., are sent to the plug-in system as the basis for plug-in task execution.

[0075] Furthermore, the plug-in system performs specific tasks based on the analysis results output by the edge computing module, such as real-time warnings, adjusting equipment parameters, and activating backup equipment. For example, if the edge computing module predicts that the equipment temperature will exceed the set threshold within the next 5 minutes, the plug-in system will immediately activate the warning mechanism, notify the operator, and generate an alarm. For example, if the temperature is about to exceed the standard, the plug-in system will activate the plug-in to adjust the fan speed or water cooling system to prevent overheating by increasing cooling. For the backup equipment activation task, if an equipment failure is predicted, the plug-in system will automatically switch to the backup equipment to ensure that the production line will not be shut down.

[0076] S202: monitor the running status data of the plug-in system in real time and pre-process the running status data.

[0077] The embodiment of the present invention performs data analysis on key performance indicators such as CPU usage, memory usage data, I / O delay data, number of pending tasks, task priority, network bandwidth usage, battery consumption rate, etc. of each plug-in. The process is to use a dynamic filtering algorithm to remove noise and retain valuable signals. For example, the calculation can be performed using a median filter method, which performs window processing on the data and replaces the current data point with the median of the data in the window.

[0078] In a specific embodiment, the process of filtering out noise from the real-time acquired operating status data using a median filter includes the following steps:

[0079] B1, set the window size; the larger the window, the worse the details; however, if the window is too small, the effect is not good when the density is high. By comprehensively comparing the details and effects, the embodiment of the present invention finally determines the window size to be 5.

[0080] B2, traverses each type of data stream in turn, and forms a window with each data point and its surrounding 4 data points (a total of 5 data points).

[0081] B3 sorts the data in the window, selects the middle value to replace the current data point, and repeats the above process until all data points have been processed.

[0082] Assume that the following CPU usage data (unit: %) is collected in real time: [65, 68, 72, 95, 70, 68, 69, 80, 200, 70, 71]. When using a window of size 5 for median filtering, the data is processed as follows:

[0083] The median of the first window [65, 68, 72, 95, 70] is 70; the median of the second window [68, 72, 95, 70, 68] is 70, the median of the third window [72, 95, 70, 68, 69] is 70, the median of the fourth window [95, 70, 68, 69, 80] is 70, and the median of the fifth window [70, 68, 69, 80, 200] is 70. Subsequent data processing is similar. After processing, the denoised data is [70, 70, 70, 70, 70, 69, 70, 70, 70, 71].

[0084] Furthermore, an embodiment of the present invention adopts a threshold-based anomaly detection method to set corresponding threshold ranges for different data types. For example, a CPU usage rate exceeding 70% is considered abnormal. When a data point exceeds the threshold, it is determined to be abnormal. Therefore, after the above-mentioned denoising process, the threshold is set to 70%. The last data point 71 of the denoised data has exceeded the threshold, which will trigger an abnormality alarm. It can timely discover possible problems in the plug-in system, thereby enabling the plug-in system to maintain a more reliable and stable operating state by timely discovering and handling abnormal situations.

[0085] S203 , allocating resources to the plug-in task using a reinforcement learning algorithm based on the pre-processed running status data, and generating a plug-in task resource allocation result.

[0086] Specifically, step S203 is executed, which specifically includes the following steps:

[0087] S2031, obtaining the pre-processed data of the number of pending tasks, task priority, current network bandwidth usage, CPU usage, and memory usage of each plug-in for use in the state space description of the reinforcement learning algorithm.

[0088] These data in the state space of the embodiment of the present invention cover multiple key aspects of plug-in tasks and system resources. For example, the number of pending tasks reflects the workload, the task priority clarifies the importance and sequence of tasks, and the network bandwidth utilization rate, CPU utilization rate and memory usage data reflect the current resource consumption level.

[0089] S2032, setting a set of actions to be taken by the plug-in, including: adjusting the CPU usage and memory usage data of the plug-in, dynamically changing the priority order of pending tasks, and limiting or increasing network bandwidth.

[0090] The embodiment of the present invention can flexibly adjust the allocation of resources according to the actual needs of the task through diversified action selection, realize on-demand allocation of resources, avoid resource waste, and ensure that each plug-in task can obtain appropriate resource guarantees, thereby improving the execution efficiency of the task.

[0091] S2033, setting reward rules, including: the less time required to complete the task, the higher the reward; the higher the resource utilization rate, the higher the reward; the higher the priority task completion rate, the higher the reward.

[0092] The embodiment of the present invention uses the principle that the shorter the time required to complete a task, the higher the reward as an incentive guide, prompting the reinforcement learning algorithm to continuously seek for a better resource allocation strategy, so that the plug-in task can be completed as soon as possible, reducing the waiting time of the task in the system, and improving the response speed and task processing throughput of the entire system. In particular, for plug-in tasks with high real-time requirements, it can significantly improve the user experience and the business processing capabilities of the system; the rule that the higher the resource utilization rate, the higher the reward encourages the algorithm to fully consider how to maximize the role of limited resources when allocating resources, and avoid idle or over-allocation of resources; the rule that the higher the completion rate of priority tasks, the higher the reward, emphasizes the protection of important tasks, and guides the algorithm to give priority to meeting the resource requirements of high-priority tasks when allocating resources, ensuring the smooth development of key businesses.

[0093] S2034, initialize the state space and train the preset reinforcement learning algorithm, and dynamically adjust the learning rate during training to generate the optimal plug-in task resource allocation result.

[0094] In a specific example scenario, the monitoring plug-in needs to process 1GB / second of data flow, with real-time monitoring priority 1 and log analysis priority 3. The network bandwidth is limited: the total bandwidth is only 10MB / second. The resource allocation process for executing the plug-in task using the reinforcement learning algorithm is as follows:

[0095] 1) Initialization state: The monitoring plug-in queue has 100 tasks and the current allocated bandwidth is 5 MB / s; the logging plug-in queue has 300 tasks and the current allocated bandwidth is 3 MB / s; the remaining system bandwidth is 2 MB / s.

[0096] 2) Rewards for Evaluation Actions: The monitoring plugin's bandwidth allocation was increased to 7MB / s, with rewards increasing due to its high task priority. The logging plugin's bandwidth allocation was reduced to 1MB / s, with rewards increasing due to reasonable resource redistribution.

[0097] 3) Dynamically adjust the learning rate: If the real-time monitoring plug-in task queue fluctuates significantly (for example, the number of tasks increases by 50% within 10 seconds), the learning rate is increased from 0.1 to 0.3 to accelerate the model to learn the optimal scheduling strategy.

[0098] 4) Select the optimal strategy: Set the monitoring plug-in's priority to the highest and allocate 7 MB / s of bandwidth to it. Reduce the bandwidth allocated to the log analysis plug-in and lower its priority.

[0099] Through the above analysis of the resource allocation results of the plug-in task, we can get the following:

[0100] ① Resource allocation optimization: The monitoring plug-in task completion rate increased by 30%, and the delay was reduced to less than 1 second.

[0101] ②Task priority guarantee: Critical tasks are scheduled according to priority and are not interfered with by secondary tasks.

[0102] ③ Improved bandwidth utilization: Utilization increased from 70% to 95%, with no bottleneck.

[0103] Traditional methods use static scheduling strategies that cannot cope with bandwidth fluctuations, resulting in delays in critical tasks. The embodiment of the present invention introduces an adaptive learning rate, which enables the reinforcement learning model to quickly adapt to dynamic environments and makes resource allocation more efficient.

[0104] S204: Perform plug-in resource allocation on the plug-in task and the plug-in task resource allocation result based on the virtualized resource pool method to generate a plug-in resource allocation result.

[0105] Specifically, executing step S204 includes the following steps:

[0106] S2041: Merge all physical computing resources into a unified resource pool, and divide the physical resources into multiple virtual resource units.

[0107] The embodiment of the present invention merges all physical computing resources (multiple servers or cloud resources) into a unified resource pool. Virtual resources can be created using a virtual machine hypervisor (such as VMware, KVM) or container technology (Docker, Kubernetes) and divided into multiple virtual resource units. Each virtual resource unit can be assigned to different tasks as needed. The resources (CPU, memory, storage) of each virtual machine or container are divided from the shared resource pool, but they are isolated from each other, so that each module can run efficiently without interfering with other modules. Centralized management and abstract integration of hardware resources are achieved based on the virtualization resource pool method. After integration in this way, there is no need to pay attention to the resource status of each specific device, but to make overall arrangements from the perspective of the overall resource pool, which greatly simplifies the complexity of resource management.

[0108] S2042, allocate virtual resource units to the edge computing modules and plug-ins that apply for resources based on the priority of the tasks.

[0109] Specifically, the resource usage of each module is monitored in real time, including indicators such as CPU usage, memory usage, I / O data, and network bandwidth. The load of each module is regularly collected and updated by a monitoring component to form real-time data. When the load of a module reaches the preset threshold, a resource request is sent to the resource pool management module, requesting the allocation of more virtual CPUs or memory. When the module load is reduced (for example, the task is completed or the load fluctuation is reduced), the unused resources are automatically reclaimed and allocated to modules that need more resources. For example, the plug-in system may need more memory to process complex calculations at certain times. The system will dynamically adjust resource allocation and release resources that are temporarily not needed by the edge computing module. When allocating resources, the task priority of different modules is taken into account. For example, the real-time data processing task of the edge computing module may be higher than certain low-priority background tasks, and resources are allocated to modules with higher priority first.

[0110] In the embodiment of the present invention, through resource pooling technology, the edge computing module and the plug-in system share computing resources (such as CPU, memory, storage, etc.), and dynamically adjust resource allocation according to the actual load to achieve load balancing. Resource pooling is achieved through virtualization technology to ensure that different modules can flexibly access shared resources.

[0111] S2043, input the plug-in task generated by the real-time edge computing module into the preset load prediction model, obtain the load prediction result within the future preset time period, and generate the resource allocation result of the plug-in system based on the load prediction result.

[0112] The training process of the preset load prediction model in the implementation of the present invention includes:

[0113] C1. Obtain historical load data, resource usage patterns, task types, time information, and external factors that affect resource consumption. Historical load data includes: CPU usage, memory usage, I / O latency data, and battery consumption rate. Resource usage patterns are the resource usage of each plug-in at different times, including peaks and fluctuations. The type of task currently performed by the plug-in may include real-time monitoring, data processing, log analysis, etc. Time information includes timestamps or time windows to capture temporal changes in load. For example, time periods such as hours, days, or weeks may be used. External factors include ambient temperature, network bandwidth, device health status, and other factors that may affect resource consumption.

[0114] C2: Cut historical load data into time windows and use a sliding window approach to generate training samples. Feature data from these training samples is extracted, including timestamps, task types, and external factors that affect resource consumption. For example, using the sliding window approach to generate training samples, we can use data from the past 10 time points to predict the load at the 11th time point.

[0115] C3 inputs the feature data into the LSTM network for training, resulting in a trained model that serves as the load forecasting model. Its output includes CPU usage, memory usage, I / O latency, and battery consumption for a preset time period. The LSTM model uses a backpropagation algorithm to adjust network weights to minimize prediction error, resulting in a trained model.

[0116] S2044, compare the resource allocation result of the plug-in system generated based on the load prediction result with the resource monitoring data of the real-time plug-in system, adjust the allocation strategy according to the comparison result, and when a plug-in fails, transfer its corresponding task to other plug-ins for processing.

[0117] Specifically, the resource usage during actual operation may deviate from the prediction, such as the sudden appearance of a large number of abnormal plug-in task requests, or the premature completion of some plug-in tasks, which releases resources. Through real-time comparison and adjustment, the system can promptly detect these changes and quickly reallocate resources to ensure that resources are always in a reasonable allocation state and maintain stable and efficient operation of the system. If there is a large deviation between the prediction and the actual load, the parameters of the prediction model will be further adjusted to further optimize the resource allocation strategy. When a plug-in fails, the task will be automatically transferred to other available plug-ins for processing to ensure the high availability and stability of the system. The fault-tolerant mechanism uses distributed redundancy technology to ensure that the plug-in system can still maintain stable operation in the event of multiple points of failure.

[0118] S205 , generating plug-in code using model-driven development and automatic code generation technology based on the plug-in resource allocation result, and optimizing and evolving the plug-in according to the plug-in code running feedback result.

[0119] Specifically, executing step S205 includes the following steps:

[0120] S2051, based on the plug-in resource allocation results, uses model-driven development and automatic code generation technology to generate plug-in code and automatically compile, test and deploy it; the code generator used supports multiple languages and platforms to ensure the cross-platform compatibility of the plug-in, and the compilation process uses incremental compilation technology to further shorten the development cycle.

[0121] S2052: When plug-in version updates and function improvement requirements are detected, use continuous integration and continuous delivery tools to update the plug-in and perform automated testing and regression testing, and automatically optimize test cases based on the test results. For example, use CI / CD tools to automatically detect plug-in version updates and function improvement requirements, and combine automated testing and regression testing to ensure the stability and compatibility of the plug-in after each update. CI / CD tools integrate AI-driven test optimization modules that can automatically optimize test cases based on test results.

[0122] S2053, when the plug-in code running feedback results are monitored, the running data is analyzed based on the feedback results to optimize the plug-in resource allocation; for example, based on the performance, stability and user experience of the runtime user feedback, the plug-in is automatically optimized and evolved. By analyzing the running data and automatically adjusting the plug-in algorithm or resource allocation strategy, the performance of the plug-in can be ensured to be continuously improved.

[0123] In this embodiment, an adaptive plug-in scheduling and dynamic optimization evolution system is also provided. The system is used to implement the above-mentioned embodiments and preferred implementation methods. The details that have been described will not be repeated here. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the system described in the following embodiments is preferably implemented in software, implementation in hardware, or a combination of software and hardware, is also possible and conceivable.

[0124] This embodiment provides an adaptive plug-in scheduling and dynamic optimization evolution system, such as Figure 3 Shown, including:

[0125] The plug-in task acquisition module 301 is used to obtain the plug-in task generated by the edge computing module based on the real-time monitoring indicator data stream of the target device;

[0126] The plug-in task resource allocation module 302 is used to allocate resources to the plug-in tasks based on the pre-processed running status data using a reinforcement learning algorithm to generate a plug-in task resource allocation result;

[0127] The real-time operation data monitoring module 303 is used to monitor the operation status data of the plug-in in real time and pre-process the operation status data;

[0128] The plug-in resource allocation module 304 is used to allocate plug-in task resources based on the virtualized resource pool method and generate a plug-in resource allocation result.

[0129] The plug-in automatic generation and optimization module 305 is used to generate plug-in code based on the plug-in resource allocation result by using model-driven development and automatic code generation technology, and optimize and evolve the plug-in according to the plug-in code operation feedback result.

[0130] In some optional implementations, the plug-in task acquisition module 301 includes:

[0131] The running status data monitoring unit is used to monitor the running status data of the plug-in system in real time, including: CPU usage, memory usage data, I / O delay data, number of pending tasks, task priority, network bandwidth usage, and battery consumption rate of each plug-in;

[0132] A data preprocessing unit, configured to remove noise from the operating status data using a preset dynamic filtering algorithm to obtain preprocessed operating status data;

[0133] The abnormality alarm unit is used to generate an abnormality alarm signal when the pre-processed operating status data exceeds the corresponding preset threshold.

[0134] In some optional implementations, the plug-in task resource allocation module 302 includes:

[0135] The state space determination unit is used to obtain the number of pending tasks, task priority, current network bandwidth usage, CPU usage, and memory usage data of each plug-in after preprocessing, which is used for the state space description of the reinforcement learning algorithm;

[0136] An action set determination unit is used to set the action set taken by the plug-in, including: adjusting the CPU usage and memory usage data of the plug-in, dynamically changing the priority order of pending tasks, and limiting or increasing network bandwidth;

[0137] The reward rule setting unit is used to set reward rules, including: the shorter the time required to complete the task, the higher the reward; the higher the resource utilization rate, the higher the reward; the higher the priority task completion rate, the higher the reward;

[0138] The plug-in task resource allocation unit is used to initialize the state space and train the preset reinforcement learning algorithm. During training, it combines dynamic adjustment of the learning rate to generate the optimal plug-in task resource allocation result.

[0139] In some optional implementations, the plug-in resource allocation module 304 includes:

[0140] A resource pool construction unit is used to merge all physical computing resources into a unified resource pool and divide the physical resources into multiple virtual resource units;

[0141] A virtual resource allocation unit is used to allocate virtual resource units to edge computing modules and plug-ins that apply for resources based on task priority;

[0142] A load prediction unit is used to input the plug-in tasks generated by the real-time edge computing module into a preset load prediction model to obtain the load prediction results within a preset time period in the future, and generate resource allocation results for the plug-in system based on the load prediction results;

[0143] The resource adjustment unit is used to compare the resource allocation results of the plug-in system generated based on the load prediction results with the resource monitoring data of the real-time plug-in system, adjust the allocation strategy according to the comparison results, and transfer the corresponding tasks to other plug-ins for processing when a plug-in fails.

[0144] In some optional implementations, the training process of the preset load prediction model in the load prediction unit includes:

[0145] Obtain historical load data, resource usage patterns, task types, time information, and external factors affecting resource consumption, wherein the historical load data includes: CPU usage, memory usage, I / O latency data, and battery consumption rate;

[0146] Cut historical load data into time windows, use a sliding window approach to generate training samples, and extract feature data from the training samples, including timestamps, task types, and external factors that affect resource consumption.

[0147] The feature data is input into the LSTM network for training, and the trained model is used as the load prediction model. Its output results include: CPU usage, memory usage, I / O delay data, and battery consumption rate within a preset time period in the future.

[0148] In some optional implementations, the plug-in automatic generation and optimization module 305 includes:

[0149] A plug-in code automatic generation unit is used to generate plug-in code based on the plug-in resource allocation results using model-driven development and automatic code generation technology, and automatically compile, test and deploy the plug-in code;

[0150] The plug-in code automatic update unit, when detecting the need for plug-in version updates and function improvements, uses continuous integration and continuous delivery tools to perform plug-in updates, automated testing, and regression testing, and automatically optimizes test cases based on the test results;

[0151] The optimization evolution unit is configured to analyze the operation data based on the feedback result of the plug-in code operation when monitoring the feedback result of the plug-in code operation to optimize the plug-in resource allocation.

[0152] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0153] The adaptive plug-in scheduling and dynamic optimization evolution system in this embodiment is presented in the form of functional units, where the units refer to ASIC (Application Specific Integrated Circuit) circuits, processors and memories that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0154] The embodiment of the present invention also provides a computer device having the above Figure 3 The adaptive plug-in scheduling and dynamic optimization evolution system shown.

[0155] See also Figure 4 , Figure 4 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 4 As shown, the computer device includes: one or more processors 10, memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components utilize different buses to communicate with each other and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in the memory or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Equally, multiple computer devices can be connected, and each device provides part of the necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 4 A processor 10 is taken as an example.

[0156] The processor 10 may be a central processing unit, a network processor, or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic, or any combination thereof.

[0157] The memory 20 stores instructions that can be executed by at least one processor 10, so as to enable at least one processor 10 to execute the method shown in the above embodiment.

[0158] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created based on the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0159] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0160] The computer device further includes a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0161] The embodiment of the present invention also provides a computer-readable storage medium. The above-mentioned method according to the embodiment of the present invention can be implemented in hardware, firmware, or implemented as a computer code that can be recorded in a storage medium, or implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor central control system or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method shown in the above embodiment is implemented.

[0162] A portion of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the form in which the computer program instruction exists in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc. Accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium that can be accessed by the computer.

[0163] Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention. Such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for adaptive plug-in scheduling and dynamic optimization evolution, characterized in that: include: Obtain the plug-in task generated by the edge computing module based on the target device's real-time monitoring indicator data stream; Monitor the running status data of the plug-in system in real time and pre-process the running status data; Based on the pre-processed running status data, the reinforcement learning algorithm is used to allocate resources for the plug-in tasks and generate the plug-in task resource allocation results; Perform plug-in resource allocation on the plug-in tasks and the plug-in task resource allocation results based on the virtualized resource pool method to generate a plug-in resource allocation result; Based on the plug-in resource allocation results, the plug-in code is generated using model-driven development and automatic code generation technology, and the plug-in is optimized and evolved according to the plug-in code running feedback results.

2. The method according to claim 1, characterized in that The real-time monitoring of the plug-in system's operating status data and pre-processing of the operating status data include: Real-time monitoring of the plug-in system's operating status data, including: each plug-in's CPU usage, memory usage data, I / O latency data, number of pending tasks, task priority, network bandwidth usage, and battery consumption rate; Using a preset dynamic filtering algorithm to remove noise from the operating status data to obtain pre-processed operating status data; When the pre-processed operating status data exceeds the corresponding preset threshold, an abnormal alarm signal is generated.

3. The method according to claim 2, characterized in that The method of allocating resources to plug-in tasks based on the pre-processed running status data using a reinforcement learning algorithm to generate a plug-in task resource allocation result includes: Obtain the number of pending tasks, task priority, current network bandwidth usage, CPU usage, and memory usage data of each plug-in after preprocessing, which is used to describe the state space of the reinforcement learning algorithm; Set the set of actions that the plug-in takes, including: adjusting the plug-in's CPU usage, memory usage data, dynamically changing the priority of pending tasks, and limiting or increasing network bandwidth; Set reward rules, including: the shorter the time required to complete the task, the higher the reward; the higher the resource utilization rate, the higher the reward; the higher the completion rate of priority tasks, the higher the reward; Initialize the state space and train the preset reinforcement learning algorithm. During training, dynamically adjust the learning rate to generate the optimal plug-in task resource allocation result.

4. The method according to claim 1, wherein The method of performing plug-in task resource allocation on the plug-in task and the plug-in resource allocation result based on the virtualized resource pool to generate the plug-in task resource allocation result includes: Merge all physical computing resources into a unified resource pool and divide the physical resources into multiple virtual resource units; Allocate virtual resource units to edge computing modules and plug-ins that request resources based on task priority; Input the plug-in tasks generated by the real-time edge computing module into the preset load forecasting model to obtain the load forecast results for the future preset time period, and generate the resource allocation results of the plug-in system based on the load forecast results; The resource allocation results of the plug-in system generated based on the load prediction results are compared with the resource monitoring data of the real-time plug-in system. The allocation strategy is adjusted according to the comparison results. When a plug-in fails, its corresponding tasks are transferred to other plug-ins for processing.

5. The method according to claim 4, characterized in that The training process of the preset load prediction model includes: Obtain historical load data, resource usage patterns, task types, time information, and external factors affecting resource consumption, wherein the historical load data includes: CPU usage, memory usage, I / O latency data, and battery consumption rate; Cut historical load data into time windows, use a sliding window approach to generate training samples, and extract feature data from the training samples, including timestamps, task types, and external factors that affect resource consumption. The feature data is input into the LSTM network for training, and the trained model is used as the load prediction model. Its output results include: CPU usage, memory usage, I / O delay data, and battery consumption rate within a preset time period in the future.

6. The method according to claim 1, characterized in that Generate plug-in code using model-driven development and automatic code generation technology based on plug-in resource allocation results, and optimize and evolve the plug-in based on plug-in code operation feedback results, including: Generate plug-in code based on the plug-in resource allocation results using model-driven development and automatic code generation technology, and automatically compile, test, and deploy it; When detecting plug-in version updates and function improvement requirements, use continuous integration and continuous delivery tools to perform plug-in updates, automated testing, and regression testing, and automatically optimize test cases based on the test results; When the plug-in code running feedback result is monitored, the running data is analyzed based on the feedback result to optimize the plug-in resource allocation.

7. An adaptive plug-in scheduling and dynamic optimization evolution system, characterized in that: include: The plug-in task acquisition module is used to obtain the plug-in tasks generated by the edge computing module based on the real-time monitoring indicator data stream of the target device; The plug-in task resource allocation module is used to allocate resources for plug-in tasks based on the pre-processed running status data using the reinforcement learning algorithm and generate plug-in task resource allocation results; Real-time operation data monitoring module, used to monitor the operation status data of the plug-in in real time and pre-process the operation status data; A plug-in resource allocation module is used to allocate plug-in task resources based on the virtualized resource pool method and generate a plug-in resource allocation result. The plug-in automatic generation and optimization module is used to generate plug-in code based on the plug-in resource allocation results using model-driven development and automatic code generation technology, and optimize and evolve the plug-in according to the plug-in code operation feedback results.

8. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the adaptive plug-in scheduling and dynamic optimization evolution method according to any one of claims 1 to 6 by executing the computer instructions.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the adaptive plug-in scheduling and dynamic optimization evolution method according to any one of claims 1 to 6.

10. A computer program product, characterized in that The method comprises computer instructions, wherein the computer instructions are used to enable a computer to execute the adaptive plug-in scheduling and dynamic optimization evolution method according to any one of claims 1 to 6.

Images