Intelligent resource scheduling method and system based on edge computing

Abstract

The invention discloses an intelligent resource scheduling method and system based on edge computing, and relates to the field of data processing, and the method comprises the steps: obtaining a real-time load evaluation value of an edge computing node; obtaining a high-load diagnosis result of the edge computing node; if the high load diagnosis result is high load, adjusting the energy expenditure adjustment parameter based on the priority of the real-time processing task; if the high load diagnosis result is low load, the energy expenditure adjustment parameter is adjusted based on the real-time load evaluation value. According to the invention, by combining the load evaluation of the edge computing node, the task priority adjustment, the energy management and the cloud cooperation, the resources are monitored in real time and the resource allocation is optimized, so that the efficient and stable operation of the system is ensured, the resource waste and overload conditions are avoided, and the performance, reliability and energy efficiency in the edge computing environment are improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and specifically to an intelligent resource scheduling method and system based on edge computing. Background Art

[0002] In today's era, with the rapid development of information technology and the advancement of digital transformation, especially in the application scenarios of the Internet of Things (IoT), artificial intelligence (AI), big data, 5G communication, etc., the computing requirements and data volume have increased explosively. When facing these requirements, the traditional centralized cloud computing architecture has exposed many limitations, which makes the intelligent resource scheduling method based on edge computing an inevitable development trend. In order to address challenges such as low latency, bandwidth pressure, real-time decision-making, heterogeneous resource management, and data privacy protection. Through intelligent resource scheduling methods, computing tasks can be efficiently allocated, resource utilization can be optimized, and the intelligence and response speed of the system can be improved, thereby supporting the efficient operation of various modern applications (such as intelligent manufacturing, autonomous driving, smart cities, etc.) in the edge computing environment.

[0003] The prior art, such as the invention patent with the publication number of CN114296898B, is an edge computing task scheduling method based on ARPSO, which is characterized by including: initializing a task scheduling cluster, forming a scheduling cluster for the subtasks of task scheduling and the virtual machines executing the subtasks through an initial random scheduling strategy; initializing a particle swarm; optimizing the particle swarm using the ARPSO algorithm formula, with the shortest time required for all tasks as the fitness value; updating and iterating the particle swarm using the fitness value formula, and obtaining the optimal scheduling cluster according to the global optimal position.

[0004] The prior art, such as the invention patent with the publication number of CN111597025B, is an edge computing scheduling algorithm and system. The edge computing scheduling algorithm includes the following steps: obtaining X computing tasks and determining the priority of each computing task; allocating the X computing tasks to queues with different priority levels according to the priority levels to form Y ready queues with different priority levels; allowing the execution time Ty of queues with different priority levels. If a task is not completed within Ty, it is paused and placed at the end of the ready queue of this priority level to queue up again. When each task is executed, the optimal n edge computing nodes are selected.

[0005] Based on the above technical solutions, it can be seen that the prior art often focuses on adopting fixed data processing mechanisms in the field of edge computing data processing. However, in actual applications, the load changes are often dynamic, and task backlogs may occur due to fixed initialization strategies or priority allocations. In addition, due to the high computing power requirements of edge computing, there is usually a large demand for energy. However, during idle periods, the energy consumption needs to be controlled to avoid waste of resources. Summary of the Invention

[0006] In view of the deficiencies of the prior art, the present invention provides an intelligent resource scheduling method and system based on edge computing. To achieve the above objectives, the present invention is implemented through the following technical solutions: An intelligent resource scheduling method based on edge computing includes: Obtain the real-time computing task volume of the edge computing node, and comprehensively analyze it with the real-time energy expenditure index of the edge computing node to obtain the real-time load evaluation value of the edge computing node.

[0007] Based on the performance parameters of the edge computing node, obtain the high-load diagnosis threshold of the edge computing node, and based on the real-time load evaluation value of the edge computing node, obtain the high-load diagnosis result of the edge computing node.

[0008] If the high-load diagnosis result of the edge computing node is high load, obtain the priority of the real-time processing tasks of the edge computing node, and adjust the energy expenditure adjustment parameter of the edge computing node based on the priority of the real-time processing tasks.

[0009] If the high-load diagnosis result of the edge computing node is low load, adjust the energy expenditure adjustment parameter of the edge computing node based on the real-time load evaluation value of the edge computing node.

[0010] As a preferred technical solution, obtaining the real-time computing task volume of the edge computing node, and comprehensively analyzing it with the real-time energy expenditure index of the edge computing node to obtain the real-time load evaluation value of the edge computing node, the specific process is as follows: Run a monitoring script on the edge computing node to obtain the real-time computing task volume of the edge computing node.

[0011] The real-time energy expenditure index of the edge computing node, the specific obtaining process includes: Obtain the hardware sensor data of the edge computing node in real time, including processor sensor data, graphics card sensor data, memory sensor data, and disk sensor data.

[0012] Obtain the processor occupancy data of the edge computing node in real time, including CPU usage rate, kernel state CPU occupancy ratio, and I / O waiting time.

[0013] Based on the hardware sensor data and processor occupancy data of the edge computing node, analyze and process to obtain the real-time energy expenditure index of the edge computing node.

[0014] Based on the real-time computing task volume and real-time energy expenditure index of the edge computing node, comprehensively analyze to obtain the real-time load evaluation value of the edge computing node.

[0015] As a preferred technical solution, based on the performance parameters of the edge computing node, the high-load diagnosis threshold of the edge computing node is obtained. The specific process is as follows: The performance parameters of the edge computing node include the maximum data volume that can be carried, the maximum processing speed, and the maximum OPS.

[0016] Based on the performance parameters of the edge computing node, the load performance evaluation value of the edge computing node is analyzed and processed. Based on the load performance evaluation value of the edge computing node, a mapping match is performed with the high-load diagnosis threshold corresponding to the pre-stored load performance evaluation value in the local built-in database of the edge computing node to obtain the high-load diagnosis threshold of the edge computing node.

[0017] The high-load diagnosis threshold of the edge computing node is used to determine whether the real-time load of the edge computing node is in a high-load state.

[0018] As a preferred technical solution, based on the real-time load evaluation value of the edge computing node, the high-load diagnosis result of the edge computing node is obtained, specifically including: The real-time load evaluation value of the edge computing node is compared with the high-load diagnosis threshold, and the high-load diagnosis result of the edge computing node is obtained based on the comparison result.

[0019] If the real-time load evaluation value of the edge computing node is greater than or equal to the high-load diagnosis threshold, the high-load diagnosis result of the edge computing node is high load.

[0020] If the real-time load evaluation value of the edge computing node is less than the high-load diagnosis threshold, the high-load diagnosis result of the edge computing node is low load.

[0021] As a preferred technical solution, if the high-load diagnosis result of the edge computing node is high load, the priority of the real-time processing task of the edge computing node is obtained, specifically including: Feature analysis is performed on the real-time processing task of the edge computing node. It is judged by scanning the proportion of unstructured data in the real-time processing task of the edge computing node. If the proportion of unstructured data in the real-time processing task exceeds the pre-set unstructured data proportion threshold of the edge computing node, then the difference is calculated to obtain the unstructured data proportion difference, and a mapping match is performed with the required performance factor corresponding to the pre-stored unstructured data proportion difference in the built-in database of the edge computing node to obtain the required performance factor of the edge computing node.

[0022] The maximum delay processing time of the real-time processing task is extracted and compared with the preset delay critical threshold. If the maximum delay processing time of the real-time processing task is less than or equal to the delay critical threshold, then the difference is calculated to obtain the delay processing time difference, and a mapping match is performed with the task delay factor corresponding to the pre-stored delay processing time difference in the built-in database of the edge computing node to obtain the task delay factor of the edge computing node.

[0023] Extract the target processing result of the real-time processing task. If the number of target senders of the target processing result of the real-time processing task is greater than the preset sender number threshold, perform a difference operation to obtain the target sender number difference, and map and match it with the task function factor corresponding to the target sender number difference pre-stored in the built-in database of the edge computing node to obtain the task function factor of the edge computing node.

[0024] Integrate the required performance factor, task latency factor, and task function factor of the edge computing node, and perform coupled correction processing to obtain the priority score of the real-time processing task of the edge computing node.

[0025] Based on the priority score of the real-time processing task of the edge computing node, map and match it with the priority corresponding to the interval where each priority score is pre-stored in the built-in database of the edge computing node to obtain the priority of the real-time processing task of the edge computing node.

[0026] As an optimal technical solution, adjust the energy expenditure adjustment parameter of the edge computing node based on the priority of the real-time processing task, specifically including: The priority of the real-time processing task includes the first priority, the second priority, and the third priority.

[0027] If the priority of the real-time processing task is the first priority, analyze and process based on the load performance evaluation value of the edge computing node to generate an overload suppression factor for the edge computing node. The overload suppression factor is used to prevent the edge computing node from being overloaded when performing the real-time processing task.

[0028] After introducing the overload suppression factor of the edge computing node to correct the energy expenditure adjustment parameter preset for the first priority, apply the corrected first energy expenditure adjustment parameter, and at the same time extract a part of the data volume of the real-time processing task to the cloud processor for auxiliary processing. The energy expenditure adjustment parameter includes the CPU operating frequency, GPU operating frequency, and processor supply voltage.

[0029] If the priority of the real-time processing task is the second priority, generate a real-time task processing acceleration index based on the basic parameters of the real-time processing task, correct the energy expenditure adjustment parameter preset for the second priority, and apply the corrected second energy expenditure adjustment parameter.

[0030] If the priority of the real-time processing task is the third priority, apply the energy expenditure adjustment parameter preset for the third priority. After application, calculate the adjusted real-time load evaluation value of the edge computing node again to obtain the high-load diagnosis result of the edge computing node. When the high-load diagnosis result of the edge computing node is still high load, subtract the adjusted real-time load evaluation value of the edge computing node from the high-load diagnosis threshold to obtain the high-load diagnosis difference. Map and match the high-load diagnosis difference with the energy expenditure adjustment parameter corresponding to the high-load diagnosis difference pre-stored in the local built-in database of the edge computing node to obtain the energy expenditure adjustment parameter of the edge computing node and apply it until the high-load diagnosis result of the edge computing node is low load.

[0031] As a preferred technical solution, extract a partial data volume of the real-time processing task to the cloud processor for auxiliary processing, specifically including: Based on the load performance evaluation value of the edge computing node, the maximum latency processing time of the real-time processing task, and the data volume of the real-time processing task, analyze and obtain the required scheduling data volume of the edge computing node.

[0032] Based on the performance parameters of the cloud processor and the real-time data volume being processed, analyze and obtain the maximum scheduling data volume of the edge computing node.

[0033] If the required scheduling data volume is greater than the maximum scheduling data volume, mark the excess data as a burst data stream and send it to other edge computing nodes for analysis and processing. The cloud processor receives a partial data volume corresponding to the required scheduling data volume of the edge computing node with the maximum scheduling data volume for auxiliary processing.

[0034] If the required scheduling data volume is less than or equal to the maximum scheduling data volume, extract data from the real-time processing task based on the required scheduling data volume of the edge computing node and send it to the cloud processor for auxiliary processing.

[0035] As a preferred technical solution, sending to other edge computing nodes specifically includes: Obtain the performance parameters of the transmission channel of the edge computing node, and combine the maximum latency processing time of the real-time processing task to obtain the farthest transmission distance of the edge computing node.

[0036] Taking the edge computing node as the center and the farthest transmission distance as the radius, divide to obtain the maximum transmission range of the edge computing node.

[0037] Within the maximum transmission range, select other edge computing nodes with the lowest real-time load evaluation value, and record this other edge computing node as the auxiliary edge computing node. After the auxiliary edge computing node receives the burst data stream, scan the burst data stream to obtain basic parameters, analyze and process them to obtain the basic feature values of the burst data stream, map and match the basic feature values of the burst data stream with the energy expenditure adjustment parameters corresponding to the pre-stored basic feature values to obtain the energy expenditure adjustment parameters of the auxiliary edge computing node, and apply them.

[0038] As a preferred technical solution, if the high-load diagnosis result of the edge computing node is low load, adjust the energy expenditure adjustment parameter of the edge computing node based on the real-time load evaluation value of the edge computing node, specifically including: If the high-load diagnosis result of the edge computing node is low load, subtract the real-time load evaluation value of the edge computing node from the high-load diagnosis threshold to obtain the low-load diagnosis difference value, map and match it with the energy expenditure adjustment parameter corresponding to the pre-stored low-load diagnosis difference value in the local built-in database of the edge computing node to obtain the energy expenditure adjustment parameter of the edge computing node, and apply it.

[0039] In addition, an intelligent resource scheduling system based on edge computing is also provided, including: A real-time load evaluation module, which is used to obtain the real-time computing task volume of the edge computing node, and comprehensively analyze it with the real-time energy expenditure index of the edge computing node to obtain the real-time load evaluation value of the edge computing node.

[0040] A high-load diagnosis module, which is used to obtain the high-load diagnosis threshold of the edge computing node based on the performance parameters of the edge computing node, and obtain the high-load diagnosis result of the edge computing node based on the real-time load evaluation value of the edge computing node.

[0041] A high-load energy expenditure adjustment module, which is used to, if the high-load diagnosis result of the edge computing node is high load, obtain the priority of the real-time processing task of the edge computing node, and adjust the energy expenditure adjustment parameter of the edge computing node based on the priority of the real-time processing task.

[0042] A low-load energy expenditure adjustment module, which is used to, if the high-load diagnosis result of the edge computing node is low load, adjust the energy expenditure adjustment parameter of the edge computing node based on the real-time load evaluation value of the edge computing node.

[0043] Compared with the prior art, the embodiments of the present invention at least have the following beneficial effects: (1)The present invention provides an intelligent resource scheduling method based on edge computing. Through comprehensive analysis of real-time computing task volume and energy expenditure index, it can evaluate the load status of edge computing nodes in real time. This real-time monitoring and evaluation can effectively capture the dynamic load changes of nodes, making resource scheduling more flexible and accurate. By combining with node performance parameters, the load status of nodes can be accurately judged, overload situations can be avoided, and thus the processing efficiency and stability of computing tasks can be improved.

[0044] (2)The present invention performs dynamic scheduling according to the real-time priority of tasks. The division of task priorities comprehensively considers various factors such as the proportion of unstructured data, task delay processing time, and target processing results, making the scheduling strategy more intelligent and dynamic, and capable of better meeting the processing requirements of different tasks. According to the task priorities and load evaluation, the system can dynamically adjust the energy expenditure parameters of nodes, thereby optimizing energy consumption, avoiding excessive consumption, and reducing costs. Especially when the load is low, the energy consumption can be reasonably reduced.

[0045] (3)The present invention introduces balancing the load of edge computing nodes by adjusting energy expenditure and task allocation. When the load of a certain node is too high, the system can dynamically adjust its energy parameters or migrate some tasks to other nodes to avoid overload of a single node and improve the overall resource utilization rate. When the load of edge nodes is high, the system transfers some data tasks to cloud processors for auxiliary processing, thereby alleviating the pressure on edge nodes. This cloud-edge collaboration method enables reasonable allocation of resources, avoids overload of edge computing nodes, and at the same time improves the overall processing capacity of the system. By introducing an overload suppression factor, the system can effectively control the energy consumption of edge computing nodes and avoid overload situations. This helps to extend the service life of devices and improve the reliability of the system.

[0046] Of course, it is not necessary for any product implementing the present invention to achieve all the above advantages simultaneously. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a schematic flowchart of the method of the present invention.

[0048] Figure 2 It is a schematic diagram of the system module of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0049] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0050] In the description of the present invention, it should be understood that the terms "opening", "upper", "lower", "thickness", "top", "middle", "length", "inner", "periphery", etc. indicating orientation or positional relationships are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the components or elements referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation on the present invention.

[0051] Please refer to Figure 1 As shown, an intelligent resource scheduling method based on edge computing provided by an embodiment of the present invention includes: Obtain the real-time computing task volume of the edge computing node, and conduct comprehensive analysis with the real-time energy expenditure index of the edge computing node to obtain the real-time load evaluation value of the edge computing node.

[0052] Obtain the real-time computing task volume of the edge computing node, and conduct comprehensive analysis with the real-time energy expenditure index of the edge computing node to obtain the real-time load evaluation value of the edge computing node. The specific process is as follows: Run a monitoring script on the edge computing node to obtain the real-time computing task volume of the edge computing node.

[0053] The real-time energy expenditure index of the edge computing node. The specific obtaining process includes: Obtain the hardware sensor data of the edge computing node in real time, including processor sensor data, graphics card sensor data, memory sensor data, and disk sensor data.

[0054] At the same time, extract the unit factor set of the hardware sensor data from the local built-in database of the edge computing node.

[0055] It should be noted that the processor sensor data includes the CPU core temperature, CPU core voltage, and CPU real-time power consumption.

[0056] The graphics card sensor data includes the GPU core temperature, GPU core voltage, and GPU real-time power consumption.

[0057] The memory sensor data includes the memory module temperature, memory real-time power consumption, and memory operating frequency.

[0058] The disk sensor data includes the disk temperature, disk power consumption, and disk IOPS.

[0059] It should be noted that for the memory operating frequency and disk IOPS, the memory operating frequency refers to the clock frequency of the memory module, usually in MHz (megahertz), indicating the number of clock cycles executed by the memory per second. The higher the frequency, the more data can be processed per unit time, and high-frequency memory can respond to requests from the CPU and other components faster.

[0060] Disk IOPS (Input / Output Operations Per Second) is a key metric for measuring the performance of storage devices, representing the number of read / write requests that a disk can handle per second. The higher the IOPS, the stronger the concurrent processing ability of the disk. Low IOPS may cause application lags, especially during database or large-scale file operations.

[0061] Obtain the processor occupancy data of edge computing nodes in real time, including CPU utilization rate, kernel-mode CPU occupancy ratio, and I / O wait time.

[0062] It should be noted that the CPU utilization rate refers to the actual percentage of CPU resources used by a computer system within a given time. The kernel-mode CPU occupancy ratio is the proportion of the time when the CPU executes in the kernel mode to the total CPU usage time. The kernel mode is the mode in which the operating system manages hardware and resources, handling system calls, hardware interactions, etc. The I / O wait time is the time when the CPU is in the idle state and is waiting due to waiting for I / O operations (such as disk reads, network communications, etc.) to complete.

[0063] Extract the CPU utilization rate verification value, kernel-mode CPU occupancy ratio verification value, and I / O wait time verification value from the local built-in database of the edge computing node.

[0064] Based on the hardware sensor data and processor occupancy data of the edge computing node, analyze and process to obtain the real-time energy expenditure index of the edge computing node, specifically including: ; ; ; ; ; ; Among them, is the real-time energy expenditure index, is the processor sensor index, is the graphics card sensor index, is the memory sensor index, is the disk sensor index, is the processor occupancy index, is the CPU core temperature, is the CPU core voltage, is the CPU real-time power consumption, is the GPU core temperature, is the GPU core voltage, is the GPU real-time power consumption, is the memory module temperature is the real-time power consumption of the memory, is the memory operating frequency, is the disk temperature, is the disk power consumption, is the disk IOPS, is the CPU usage rate, is the proportion of the kernel-mode CPU, is the I / O waiting time, is the temperature unit factor, is the voltage unit factor, is the real-time power consumption unit factor, is the memory operating frequency unit factor, is the disk IOPS unit factor, is the CPU usage rate verification value, is the proportion of the kernel-mode CPU verification value, is the I / O waiting time verification value, is the processor sensor index weight factor, is the graphics card sensor index weight factor, is the memory sensor index weight factor, is the disk sensor index weight factor, is the processor occupancy index weight factor.

[0065] It should be noted that the processor sensor index weight factor, the graphics card sensor index weight factor, the memory sensor index weight factor, the disk sensor index weight factor, and the processor occupancy index weight factor, their value ranges are all between 0 and 1, and satisfy , the processor sensor index weight factor is the influence factor of the processor sensor index pre-stored in the local built-in database of the edge computing node, indicating the degree of influence of the processor sensor index on the real-time energy expenditure index of the edge computing node; the graphics card sensor index weight factor is the influence factor of the graphics card sensor index pre-stored in the local built-in database of the edge computing node, indicating the degree of influence of the graphics card sensor index on the real-time energy expenditure index of the edge computing node; the memory sensor index weight factor is the influence factor of the memory sensor index pre-stored in the local built-in database of the edge computing node, indicating the degree of influence of the memory sensor index on the real-time energy expenditure index of the edge computing node; the disk sensor index weight factor is the influence factor of the disk sensor index pre-stored in the local built-in database of the edge computing node, indicating the degree of influence of the disk sensor index on the real-time energy expenditure index of the edge computing node. The processor occupancy index weight factor is an influence factor of the processor occupancy index pre-stored in the local built-in database of the edge computing node, indicating the degree of influence of the processor occupancy index on the real-time energy expenditure index of the edge computing node. When used, it is directly extracted from the local built-in database of the edge computing node. For example, the processor sensor index, graphics card sensor index, memory sensor index, disk sensor index and processor occupancy index of the edge computing node are input into the mapping set preset in the local built-in database of the edge computing node to obtain the processor sensor index weight factor, graphics card sensor index weight factor, memory sensor index weight factor, disk sensor index weight factor and processor occupancy index weight factor, and the corresponding mapping relationship is one-to-one.

[0066] It should also be noted that when calculating the real-time energy expenditure index of edge computing nodes, the parameters used have a certain correlation, including: temperature is directly related to processor power consumption. Temperature increase usually means that the CPU load increases, and power consumption increases accordingly, affecting the energy expenditure index. Voltage is also a key factor in power consumption. High voltage improves performance but increases energy consumption. The real-time power consumption of the CPU is a direct indicator of energy consumption. The higher the power consumption, the greater the energy expenditure. The temperature, voltage and real-time power consumption of the GPU are closely related to its load and power consumption. The higher the load, the greater the power consumption, which affects the energy expenditure index. The temperature, power consumption and operating frequency of the memory directly affect the system energy consumption. The higher the frequency, the greater the power consumption. The temperature, power consumption and IOPS of the disk are proportional to the load. The greater the load, the higher the power consumption, which in turn increases energy expenditure. CPU utilization, kernel CPU proportion and I / O waiting time affect system load and power consumption. When utilization and kernel CPU proportion are high, power consumption increases. Long I / O waiting time may lead to inefficient operation. Although the CPU is idle, the energy consumption is high, which all affect energy expenditure.

[0067] Based on the real-time computing task volume and real-time energy expenditure index of the edge computing node, a real-time load evaluation value of the edge computing node is comprehensively analyzed, specifically including: ; Among them, is the real-time load evaluation value of the edge computing node, is the real-time computing task volume of the edge computing node, is the computing task volume unit factor, is the real-time energy expenditure index, is the real-time computing task volume weight factor, is the real-time energy expenditure index weight factor.

[0068] It should be noted that the real-time computing task volume weight factor and the real-time energy expenditure index weight factor both have a value range between 0 and 1, and satisfy , the real-time computing task volume weight factor is the influencing factor of the real-time computing task volume pre-stored in the local built-in database of the edge computing node, indicating the degree of influence of the real-time computing task volume on the real-time load evaluation value of the edge computing node; the real-time energy expenditure index weight factor is the influencing factor of the real-time energy expenditure index pre-stored in the local built-in database of the edge computing node, indicating the degree of influence of the real-time energy expenditure index on the real-time load evaluation value of the edge computing node, and it is directly extracted from the local built-in database of the edge computing node when used. For example, the real-time computing task volume and real-time energy expenditure index of the edge computing node are input into the preset mapping set in the local built-in database of the edge computing node to obtain the real-time computing task volume weight factor and the real-time energy expenditure index weight factor, and their corresponding mapping relationship is one-to-one.

[0069] Based on the performance parameters of the edge computing node, a high-load diagnosis threshold of the edge computing node is obtained, and based on the real-time load evaluation value of the edge computing node, a high-load diagnosis result of the edge computing node is obtained.

[0070] Based on the performance parameters of the edge computing node, a high-load diagnosis threshold of the edge computing node is obtained, and the specific process is as follows: The performance parameters of the edge computing node include the maximum data carrying capacity, the maximum processing speed, and the maximum OPS.

[0071] It should be noted that the maximum OPS (Operations Per Second) refers to the maximum number of operations that the edge computing node can execute per unit time, indicating the highest processing capacity that the system can maintain under high load. The higher the maximum OPS, the stronger the computing ability of the edge computing node, and the faster it can process data and tasks.

[0072] Based on the performance parameters of the edge computing node, an analysis and processing is carried out to obtain the load performance evaluation value of the edge computing node, specifically including: Among them, is the load performance evaluation value of the edge computing node, is the maximum data volume that the edge computing node can bear, is the maximum processing speed of the edge computing node, is the maximum OPS of the edge computing node, is the unit factor of the data volume that can be borne, is the unit factor of the processing speed, is the unit factor of OPS.

[0073] It should be noted that the performance parameters of the edge computing node include the maximum data volume that can be borne, the maximum processing speed, and the maximum OPS. There is a certain correlation among these parameters. The maximum data volume that can be borne is usually restricted by the maximum processing speed. Because the faster the processing speed, the more data the node can process within a unit time. The higher the maximum OPS, the faster the node can complete each operation, thereby increasing the data volume that can be processed per second and indirectly affecting the maximum data volume that can be borne. If the processing speed of the computing node is very high, it can quickly process more data, thereby increasing its maximum data volume that can be borne. On the contrary, if the processing speed is low, even if the node has a large bandwidth, it is difficult to bear more data. The maximum processing speed determines the processing efficiency of each task, while the maximum OPS measures the number of operations that the node can execute. The two complement each other. The higher the maximum OPS, the faster the processing speed, and more operations can be completed in a shorter time, improving the processing ability. The maximum OPS directly affects the data transmission and access speed. A higher OPS means that the node can execute data operations faster, support a larger data throughput, and is therefore closely related to the maximum data volume that can be borne. Processing tasks often require a large number of data operations, especially in scenarios where data needs to be read and written frequently. A higher OPS can improve the processing speed, reduce the impact of the I / O bottleneck on the processing task, and thus improve the overall performance of the system.

[0074] Based on the load performance evaluation value of the edge computing node, a mapping and matching is carried out with the high-load diagnosis threshold corresponding to the pre-stored load performance evaluation value in the local built-in database of the edge computing node to obtain the high-load diagnosis threshold of the edge computing node.

[0075] The high-load diagnosis threshold of the edge computing node is used to judge whether the real-time load of the edge computing node is in a high-load state.

[0076] Based on the real-time load evaluation value of the edge computing node, a high-load diagnosis result of the edge computing node is obtained, specifically including: Compare the real-time load evaluation value of the edge computing node with the high-load diagnosis threshold, and obtain the high-load diagnosis result of the edge computing node based on the comparison result.

[0077] If the real-time load evaluation value of the edge computing node is greater than or equal to the high-load diagnosis threshold, the high-load diagnosis result of the edge computing node is high load.

[0078] If the real-time load evaluation value of the edge computing node is less than the high-load diagnosis threshold, the high-load diagnosis result of the edge computing node is low load.

[0079] If the high-load diagnosis result of the edge computing node is high load, obtain the priority of the real-time processing tasks of the edge computing node, and adjust the energy expenditure adjustment parameter of the edge computing node based on the priority of the real-time processing tasks.

[0080] If the high-load diagnosis result of the edge computing node is high load, obtain the priority of the real-time processing tasks of the edge computing node, specifically including: Conduct feature analysis on the real-time processing tasks of the edge computing node. Judge by scanning the proportion of unstructured data in the real-time processing tasks of the edge computing node. If the proportion of unstructured data in the real-time processing task exceeds the pre-set unstructured data proportion threshold of the edge computing node, then calculate the difference to obtain the unstructured data proportion difference, and map and match it with the required performance factor corresponding to the unstructured data proportion difference pre-stored in the built-in database of the edge computing node to obtain the required performance factor of the edge computing node.

[0081] It should be noted that unstructured data refers to data without a predefined data model or organizational form, usually including a large amount of text, images, audio, video, log files, etc.

[0082] Extract the maximum delay processing time of the real-time processing task, compare it with the preset delay critical threshold. If the maximum delay processing time of the real-time processing task is less than or equal to the delay critical threshold, then calculate the difference to obtain the delay processing time difference, and map and match it with the task delay factor corresponding to the delay processing time difference pre-stored in the built-in database of the edge computing node to obtain the task delay factor of the edge computing node.

[0083] Extract the target processing result of the real-time processing task. If the number of target senders of the target processing result of the real-time processing task is greater than the preset sender number threshold, then calculate the difference to obtain the target sender number difference, and map and match it with the task function factor corresponding to the target sender number difference pre-stored in the built-in database of the edge computing node to obtain the task function factor of the edge computing node.

[0084] Integrate the required performance factor, task delay factor, and task function factor of the edge computing node's real-time processing task, and obtain the priority score of the edge computing node's real-time processing task after coupling and correction, specifically including: ; Among them, is the priority score of the edge computing node's real-time processing task, is the required performance factor of the edge computing node's real-time processing task, is the task delay factor of the edge computing node's real-time processing task, is the task function factor of the edge computing node's real-time processing task, is the weight factor of the required performance factor, is the weight factor of the task delay factor, is the weight factor of the task function factor.

[0085] It should be noted that the value ranges of the weight factor of the required performance factor, the weight factor of the task delay factor, and the weight factor of the task function factor are all between 0 and 1, and satisfy , the weight factor of the required performance factor is the influence factor of the required performance factor pre-stored in the local built-in database of the edge computing node, indicating the degree of influence of the required performance factor on the priority score of the edge computing node's real-time processing task; the weight factor of the task delay factor is the influence factor of the task delay factor pre-stored in the local built-in database of the edge computing node, indicating the degree of influence of the task delay factor on the priority score of the edge computing node's real-time processing task; the weight factor of the task function factor is the influence factor of the task function factor pre-stored in the local built-in database of the edge computing node, indicating the degree of influence of the task function factor on the priority score of the edge computing node's real-time processing task. When used, it is directly extracted from the local built-in database of the edge computing node. For example, after inputting the real-time required performance factor, real-time task delay factor, and real-time task function factor of the edge computing node's real-time processing task into the preset mapping set in the local built-in database of the edge computing node, the weight factor of the required performance factor, the weight factor of the task delay factor, and the weight factor of the task function factor are obtained, and their corresponding mapping relationships are one-to-one.

[0086] Based on the priority score of the edge computing node's real-time processing task, perform mapping matching with the priorities corresponding to each priority score range pre-stored in the built-in database of the edge computing node to obtain the priority of the edge computing node's real-time processing task.

[0087] Adjust the energy expenditure adjustment parameter of the edge computing node based on the priority of the real-time processing task, specifically including: The priorities of the real-time processing tasks include the first priority, the second priority, and the third priority.

[0088] If the priority of the real-time processing task is the first priority, then based on the load performance evaluation value of the edge computing node, an overload suppression factor of the edge computing node is analyzed and generated. The overload suppression factor is used to prevent the edge computing node from being overloaded when performing real-time processing tasks, and specifically includes: Based on the load performance evaluation value of the edge computing node, map and match it with the overload suppression factor corresponding to the pre-stored load performance evaluation value in the local built-in database of the edge computing node to obtain the overload suppression factor of the edge computing node.

[0089] After introducing the overload suppression factor of the edge computing node to correct the energy expenditure adjustment parameter preset for the first priority, apply the corrected first energy expenditure adjustment parameter, and at the same time extract a partial amount of data of the real-time processing task to the cloud processor for auxiliary processing. The energy expenditure adjustment parameter includes the CPU operating frequency, GPU operating frequency, and processor supply voltage.

[0090] If the priority of the real-time processing task is the second priority, then generate a real-time task processing acceleration index based on the basic parameters of the real-time processing task, correct the energy expenditure adjustment parameter preset for the second priority, and apply the corrected second energy expenditure adjustment parameter.

[0091] Generating a real-time task processing acceleration index based on the basic parameters of the real-time processing task specifically includes: The basic parameters of the real-time processing task include the total amount of data, the number of data dimensions, and the proportion of unstructured data.

[0092] Extract basic verification parameters from the local built-in database of the edge computing node, including the total data verification amount, the number of data dimension verifications, and the proportion verification value of unstructured data.

[0093] Compare and process the basic parameters of the real-time processing task with the basic verification parameters to obtain the real-time task processing acceleration index, specifically including: ; Among them, is the basic characteristic value of the real-time processing task, is the total amount of data, is the number of data dimensions, is the proportion of unstructured data, is the total data verification amount, is the number of data dimension verifications, is the proportion verification value of unstructured data, is the total amount of data weight factor, is the number of data dimension weight factors, is the proportion weight factor of unstructured data.

[0094] It should be noted that the value ranges of the total data volume weight factor, the data dimension number weight factor, and the unstructured data proportion weight factor are all between 0 and 1, and satisfy , where the total data volume weight factor is the influencing factor of the total data volume pre-stored in the local built-in database of the edge computing node, indicating the degree to which the total data volume affects the basic characteristic value of the real-time processing task; the data dimension number weight factor is the influencing factor of the number of data dimensions pre-stored in the local built-in database of the edge computing node, indicating the degree to which the number of data dimensions affects the basic characteristic value of the real-time processing task; the unstructured data proportion weight factor is the influencing factor of the proportion of unstructured data pre-stored in the local built-in database of the edge computing node, indicating the degree to which the proportion of unstructured data affects the basic characteristic value of the real-time processing task. When in use, it is directly extracted from the local built-in database of the edge computing node. For example, after inputting the total data volume, the number of data dimensions, and the proportion of unstructured data of the real-time processing task into the preset mapping set in the local built-in database of the edge computing node, the total data volume weight factor, the data dimension number weight factor, and the unstructured data proportion weight factor are obtained, and their corresponding mapping relationships are one-to-one.

[0095] It should also be noted that the basic parameters of the real-time processing task include the total data volume, the number of data dimensions, and the proportion of unstructured data. There is a certain correlation between these parameters. Generally, the more the number of data dimensions, the greater the total data volume will be, because each data record contains more information and requires more storage space. The higher the proportion of unstructured data, the greater the total data volume usually is, because unstructured data is usually more complex and difficult to compress, occupying more storage space. The number of data dimensions is usually related to structured data, and the dimensions of unstructured data are not fixed, so there is no direct linear relationship between these two parameters. The higher the proportion of unstructured data, the fewer the number of data dimensions may be, but its complexity and content may affect the way of data processing and analysis.

[0096] Based on the basic characteristic value of the real-time processing task, it is mapped and matched with the real-time task processing acceleration index corresponding to the basic characteristic value preset in the local built-in database of the edge computing node to obtain the real-time task processing acceleration index.

[0097] If the priority of the real-time processing task is the third priority, then apply the energy expenditure adjustment parameter preset for the third priority. After application, calculate the adjusted real-time load evaluation value of the edge computing node again to obtain the high-load diagnosis result of the edge computing node. When the high-load diagnosis result of the edge computing node is still high load, then perform a difference operation on the adjusted real-time load evaluation value of the edge computing node and the high-load diagnosis threshold to obtain the high-load diagnosis difference. Map and match the high-load diagnosis difference with the energy expenditure adjustment parameter corresponding to the high-load diagnosis difference pre-stored in the local built-in database of the edge computing node to obtain the energy expenditure adjustment parameter of the edge computing node and apply it until the high-load diagnosis result of the edge computing node is low load.

[0098] In the embodiment of the present invention, applying the energy expenditure adjustment parameter of the edge computing node until the high-load diagnosis result of the edge computing node is low load aims to reduce the load condition of the edge computing node.

[0099] Extract a partial data volume of the real-time processing task to the cloud processor for auxiliary processing, specifically including: Based on the load performance evaluation value of the edge computing node, the maximum latency processing time of the real-time processing task, and the data volume of the real-time processing task, analyze and obtain the demand scheduling data volume of the edge computing node, specifically including: Based on the load performance evaluation value of the edge computing node, map and match it with the data processing volume per unit time corresponding to the load performance evaluation value pre-stored in the local built-in database of the edge computing node to obtain the data processing volume per unit time of the edge computing node. Conduct comprehensive analysis with the data volume of the real-time processing task to obtain the total processing time of the edge computing node, specifically: , where is the data volume of the real-time processing task, is the data processing volume per unit time of the edge computing node, is the total processing time of the edge computing node. Compare it with the maximum latency processing time of the real-time processing task. If the total processing time of the edge computing node is greater than or equal to the maximum latency processing time of the real-time processing task, then perform a difference operation to obtain the processing time difference, and map and match it with the demand scheduling data volume corresponding to the processing time difference pre-stored in the local built-in database of the edge computing node to obtain the demand scheduling data volume of the edge computing node.

[0100] In the embodiment of the present invention, if the total processing time of the edge computing node is less than the maximum latency processing time of the real-time processing task, then no auxiliary processing is performed.

[0101] Based on the performance parameters of the cloud processor and the real-time data volume being processed, analyze and obtain the maximum scheduling data volume of the edge computing node, specifically including: The performance parameters of the cloud processor include the maximum data carrying capacity, the maximum processing speed, and the maximum OPS of the cloud processor.

[0102] ; Among them, is the real-time load performance evaluation value of the cloud processor, is the maximum data carrying capacity of the cloud processor, is the real-time data volume being processed by the cloud processor, is the maximum processing speed of the cloud processor, is the maximum OPS of the cloud processor, is the unit factor of the data carrying capacity, is the unit factor of the processing speed, is the unit factor of OPS.

[0103] Based on the real-time load performance evaluation value of the cloud processor, a mapping match is made with the data processing volume per unit time corresponding to the pre-stored load performance evaluation value in the built-in database of the cloud processor, obtaining the data processing volume per unit time of the cloud processor, and through comprehensive analysis with the maximum latency processing time of the real-time processing task, the maximum scheduling data volume is obtained. Specifically, it includes multiplying the processing volume per unit time of the cloud processing by the maximum latency processing time of the real-time processing task to obtain the maximum scheduling data volume.

[0104] If the required scheduling data volume is greater than the maximum scheduling data volume, the excess data is marked as a burst data stream, the data volume of the burst data stream is obtained, and sent to other edge computing nodes for analysis and processing. The cloud processor receives and assists in processing part of the data corresponding to the required scheduling data volume of the edge computing node with the maximum scheduling data volume.

[0105] If the required scheduling data volume is less than or equal to the maximum scheduling data volume, data extraction is performed on the real-time processing task based on the required scheduling data volume of the edge computing node and sent to the cloud processor for auxiliary processing.

[0106] Sending to other edge computing nodes specifically includes: Obtaining the performance parameters of the transmission channel of the edge computing node, and combining with the maximum latency processing time of the real-time processing task, obtaining the farthest transmission distance of the edge computing node, specifically including: The performance parameters of the transmission channel of the edge computing node include link processing delay, queuing delay, and transmission rate.

[0107] Obtaining the signal transmission rate from the built-in database of the edge computing node. In the embodiments of the present invention, each edge computing node uses optical fiber communication, so the signal transmission rate is a fixed value, specifically infinitely close to the speed of light.

[0108] ; Among them, is the farthest transmission distance of the edge computing node, is the maximum latency processing time of the real-time processing task, is the link processing latency of the transmission channel, is the queuing latency of the transmission channel, is the signal transmission rate, is the data volume of the burst data stream, is the transmission rate of the transmission channel.

[0109] Taking the edge computing node as the center and the farthest transmission distance as the radius, the maximum transmission range of the edge computing node is divided.

[0110] Within the maximum transmission range, select other edge computing nodes with the lowest real-time load evaluation value, and record this other edge computing node as the auxiliary edge computing node. After the auxiliary edge computing node receives the burst data stream, scan the burst data stream to obtain basic parameters, and analyze and process to obtain the basic characteristic values of the burst data stream. The specific calculation method is the same as that of the basic characteristic values of the real-time processing task.

[0111] Based on the mapping and matching of the basic characteristic values of the burst data stream and the energy expenditure adjustment parameters corresponding to the pre-stored basic characteristic values, obtain the energy expenditure adjustment parameters of the auxiliary edge computing node and apply them.

[0112] If the high-load diagnosis result of the edge computing node is low load, adjust the energy expenditure adjustment parameters of the edge computing node based on the real-time load evaluation value of the edge computing node.

[0113] If the high-load diagnosis result of the edge computing node is low load, adjust the energy expenditure adjustment parameters of the edge computing node based on the real-time load evaluation value of the edge computing node, specifically including: If the high-load diagnosis result of the edge computing node is low load, subtract the real-time load evaluation value of the edge computing node from the high-load diagnosis threshold to obtain the low-load diagnosis difference value, and perform mapping and matching with the energy expenditure adjustment parameters corresponding to the low-load diagnosis difference value pre-stored in the local built-in database of the edge computing node to obtain the energy expenditure adjustment parameters of the edge computing node and apply them.

[0114] In the embodiment of the present invention, the purpose of applying the energy expenditure adjustment parameters to adjust the edge computing node with a low high-load diagnosis result is to reduce the energy consumption of the edge computing node.

[0115] As Figure 2 shown, in this embodiment, the present invention provides an intelligent resource scheduling system based on edge computing, including: A real-time load evaluation module, which is used to obtain the real-time computing task volume of an edge computing node, and conduct comprehensive analysis with the real-time energy expenditure index of the edge computing node to obtain the real-time load evaluation value of the edge computing node.

[0116] A high-load diagnosis module, which is used to obtain the high-load diagnosis threshold of the edge computing node based on the performance parameters of the edge computing node, and obtain the high-load diagnosis result of the edge computing node based on the real-time load evaluation value of the edge computing node.

[0117] A high-load energy expenditure adjustment module, which is used to, if the high-load diagnosis result of the edge computing node is high load, obtain the priority of the real-time processing tasks of the edge computing node, and adjust the energy expenditure adjustment parameters of the edge computing node based on the priority of the real-time processing tasks.

[0118] A low-load energy expenditure adjustment module, which is used to, if the high-load diagnosis result of the edge computing node is low load, adjust the energy expenditure adjustment parameters of the edge computing node based on the real-time load evaluation value of the edge computing node.

[0119] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0120] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor limit the invention to the specific embodiments. Obviously, many modifications and variations can be made according to the content of this specification. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and utilize the present invention well. As long as it does not deviate from the structure of the present invention or exceed the scope defined by the present invention, it should fall within the protection scope of the present invention.

Claims

1. An intelligent resource scheduling method based on edge computing, characterized in that Including: Obtain the real-time computing task volume of the edge computing node, and conduct comprehensive analysis with the real-time energy expenditure index of the edge computing node to obtain the real-time load evaluation value of the edge computing node; Based on the performance parameters of the edge computing node, obtain the high-load diagnosis threshold of the edge computing node. Based on the real-time load evaluation value of the edge computing node, obtain the high-load diagnosis result of the edge computing node; If the high-load diagnosis result of the edge computing node is high load, obtain the priority of the real-time processing tasks of the edge computing node, and adjust the energy expenditure adjustment parameter of the edge computing node based on the priority of the real-time processing tasks; If the high-load diagnosis result of the edge computing node is low load, adjust the energy expenditure adjustment parameter of the edge computing node based on the real-time load evaluation value of the edge computing node.

2. The intelligent resource scheduling method based on edge computing according to claim 1, wherein: The process of obtaining the real-time computing task volume of the edge computing node and conducting comprehensive analysis with the real-time energy expenditure index of the edge computing node to obtain the real-time load evaluation value of the edge computing node is as follows: Run a monitoring script on the edge computing node to obtain the real-time computing task volume of the edge computing node; The specific process of obtaining the real-time energy expenditure index of the edge computing node includes: Obtain the hardware sensor data of the edge computing node in real time, including processor sensor data, graphics card sensor data, memory sensor data, and disk sensor data; Obtain the processor occupancy data of the edge computing node in real time, including CPU usage rate, kernel state CPU occupancy ratio, and I / O waiting time; Based on the hardware sensor data and processor occupancy data of the edge computing node, analyze and process to obtain the real-time energy expenditure index of the edge computing node; Based on the real-time computing task volume and real-time energy expenditure index of the edge computing node, conduct comprehensive analysis to obtain the real-time load evaluation value of the edge computing node.

3. The intelligent resource scheduling method based on edge computing according to claim 1 is characterized in that: The process of obtaining the high-load diagnosis threshold of the edge computing node based on the performance parameters of the edge computing node is as follows: The performance parameters of the edge computing node include the maximum data volume that can be carried, the maximum processing speed, and the maximum OPS; Based on the performance parameters of the edge computing node, analyze and process to obtain the load performance evaluation value of the edge computing node. Based on the load performance evaluation value of the edge computing node, perform mapping and matching with the high-load diagnosis threshold corresponding to the pre-stored load performance evaluation value in the local built-in database of the edge computing node to obtain the high-load diagnosis threshold of the edge computing node; The high-load diagnosis threshold of the edge computing node is used to determine whether the real-time load of the edge computing node is in a high-load state.

4. The intelligent resource scheduling method based on edge computing according to claim 3, wherein: The process of obtaining the high-load diagnosis result of the edge computing node based on the real-time load evaluation value of the edge computing node specifically includes: Compare the real-time load evaluation value of the edge computing node with the high-load diagnosis threshold, and obtain the high-load diagnosis result of the edge computing node based on the comparison result; If the real-time load evaluation value of the edge computing node is greater than or equal to the high-load diagnosis threshold, the high-load diagnosis result of the edge computing node is high load; If the real-time load evaluation value of the edge computing node is less than the high-load diagnosis threshold, the high-load diagnosis result of the edge computing node is low load.

5. The intelligent resource scheduling method based on edge computing according to claim 4, characterized in that: If the high-load diagnosis result of the edge computing node is high load, obtain the priority of the real-time processing tasks of the edge computing node, specifically including: Conduct feature analysis on the real-time processing tasks of the edge computing node. Judge by scanning the proportion of unstructured data in the real-time processing tasks of the edge computing node. If the proportion of unstructured data in the real-time processing task exceeds the pre-set unstructured data proportion threshold of the edge computing node, perform a difference operation to obtain the unstructured data proportion difference, and perform mapping and matching with the required performance factor corresponding to the unstructured data proportion difference pre-stored in the built-in database of the edge computing node to obtain the required performance factor of the edge computing node; Extract the maximum delay processing time of the real-time processing task, and compare it with the preset delay critical threshold. If the maximum delay processing time of the real-time processing task is less than or equal to the delay critical threshold, perform a difference operation to obtain the delay processing time difference, and perform mapping and matching with the task delay factor corresponding to the delay processing time difference pre-stored in the built-in database of the edge computing node to obtain the task delay factor of the edge computing node; Extract the target processing result of the real-time processing task. If the number of target senders of the target processing result of the real-time processing task is greater than the preset sender number threshold, perform a difference operation to obtain the target sender number difference, and perform mapping and matching with the task function factor corresponding to the target sender number difference pre-stored in the built-in database of the edge computing node to obtain the task function factor of the edge computing node; Integrate the required performance factor, task delay factor and task function factor of the edge computing node, and obtain the priority score of the real-time processing tasks of the edge computing node after coupling and correction processing; Based on the priority score of the real-time processing tasks of the edge computing node, perform mapping and matching with the priorities corresponding to the intervals where each priority score is pre-stored in the built-in database of the edge computing node to obtain the priority of the real-time processing tasks of the edge computing node.

6. The intelligent resource scheduling method based on edge computing according to claim 1, characterized in that: The adjustment of the energy expenditure adjustment parameter of the edge computing node based on the priority of the real-time processing task specifically includes: The priorities of the real-time processing tasks include the first priority, the second priority and the third priority; If the priority of the real-time processing task is the first priority, analyze and generate the overload suppression factor of the edge computing node based on the load performance evaluation value of the edge computing node. The overload suppression factor is used to prevent the edge computing node from being overloaded when performing real-time processing tasks; After introducing the overload suppression factor of the edge computing node to correct the energy expenditure adjustment parameter preset for the first priority, apply the corrected first energy expenditure adjustment parameter, and at the same time extract a part of the data volume of the real-time processing task to the cloud processor for auxiliary processing. The energy expenditure adjustment parameters include the CPU operating frequency, GPU operating frequency and processor supply voltage; If the priority of the real-time processing task is the second priority, generate a real-time task processing acceleration index based on the basic parameters of the real-time processing task, correct the energy expenditure adjustment parameter preset for the second priority, and apply the corrected second energy expenditure adjustment parameter; If the priority of the real-time processing task is the third priority, then apply the energy expenditure adjustment parameter preset for the third priority. After application, calculate the adjusted real-time load evaluation value of the edge computing node again to obtain the high-load diagnosis result of the edge computing node. When the high-load diagnosis result of the edge computing node is still high load, then perform a difference operation between the adjusted real-time load evaluation value of the edge computing node and the high-load diagnosis threshold to obtain the high-load diagnosis difference value. Map and match the high-load diagnosis difference value with the energy expenditure adjustment parameter corresponding to the high-load diagnosis difference value stored in the local built-in database of the edge computing node to obtain the energy expenditure adjustment parameter of the edge computing node and apply it until the high-load diagnosis result of the edge computing node is low load.

7. The intelligent resource scheduling method based on edge computing according to claim 6, wherein: The extraction of a partial data volume of the real-time processing task to the cloud processor for auxiliary processing specifically includes: Based on the load performance evaluation value of the edge computing node, the maximum latency processing time of the real-time processing task, and the data volume of the real-time processing task, analyze and obtain the required scheduling data volume of the edge computing node; Based on the performance parameters of the cloud processor and the real-time data volume being processed, analyze and obtain the maximum scheduling data volume of the edge computing node; If the required scheduling data volume is greater than the maximum scheduling data volume, then mark the excess data as a burst data stream and send it to other edge computing nodes for analysis and processing. The cloud processor receives a partial data volume corresponding to the required scheduling data volume of the edge computing node at the maximum scheduling data volume for auxiliary processing; If the required scheduling data volume is less than or equal to the maximum scheduling data volume, then perform data extraction on the real-time processing task based on the required scheduling data volume of the edge computing node and send it to the cloud processor for auxiliary processing.

8. The intelligent resource scheduling method based on edge computing according to claim 7, wherein: The sending to other edge computing nodes specifically includes: Obtain the performance parameters of the transmission channel of the edge computing node, and combine the maximum latency processing time of the real-time processing task to obtain the farthest transmission distance of the edge computing node; Taking the edge computing node as the center and the farthest transmission distance as the radius, divide to obtain the maximum transmission range of the edge computing node; Within the maximum transmission range, select other edge computing nodes with the lowest real-time load evaluation value, and record this other edge computing node as the auxiliary edge computing node. After the auxiliary edge computing node receives the burst data stream, scan the burst data stream to obtain the basic parameters, analyze and process to obtain the basic feature value of the burst data stream. Map and match the basic feature value of the burst data stream with the energy expenditure adjustment parameter corresponding to the pre-stored basic feature value to obtain the energy expenditure adjustment parameter of the auxiliary edge computing node and apply it.

9. The intelligent resource scheduling method based on edge computing according to claim 1, wherein: If the high-load diagnosis result of the edge computing node is low load, then adjust the energy expenditure adjustment parameter of the edge computing node based on the real-time load evaluation value of the edge computing node. Specifically, it includes: If the high-load diagnosis result of the edge computing node is low load, the difference between the real-time load evaluation value of the edge computing node and the high-load diagnosis threshold is obtained, and it is mapped and matched with the energy expenditure adjustment parameter corresponding to the low-load diagnosis difference pre-stored in the local built-in database of the edge computing node to obtain the energy expenditure adjustment parameter of the edge computing node and apply it.

10. An intelligent resource scheduling system based on edge computing, characterized in that: A real-time load evaluation module, configured to obtain the real-time computing task volume of the edge computing node, and comprehensively analyze it with the real-time energy expenditure index of the edge computing node to obtain the real-time load evaluation value of the edge computing node; A high-load diagnosis module, configured to obtain the high-load diagnosis threshold of the edge computing node based on the performance parameters of the edge computing node, and obtain the high-load diagnosis result of the edge computing node based on the real-time load evaluation value of the edge computing node; A high-load energy expenditure adjustment module, configured to, if the high-load diagnosis result of the edge computing node is high load, obtain the priority of the real-time processing task of the edge computing node, and adjust the energy expenditure adjustment parameter of the edge computing node based on the priority of the real-time processing task; A low-load energy expenditure adjustment module, configured to, if the high-load diagnosis result of the edge computing node is low load, adjust the energy expenditure adjustment parameter of the edge computing node based on the real-time load evaluation value of the edge computing node.

Images