Virtualized computing resource scheduling method and system based on power wireless local area network

Abstract

The application provides a virtualized computing resource scheduling method and system based on a power wireless local area network. Firstly, power equipment operation load data of an access device in the power wireless local area network coverage is acquired, including real-time current fluctuation and other characteristics. Then, load characteristic extraction is performed on the power equipment operation load data to obtain a load characteristic set containing device power consumption fluctuation and other characteristics. Then, dynamic resource demand prediction is performed on the set based on a preset load prediction model to generate a virtualized resource demand prediction result containing a computing resource allocation level. A virtualized resource scheduling strategy containing an edge computing node resource allocation topology is generated according to the virtualized resource demand prediction result. Finally, virtualized computing resources are dynamically adjusted based on the virtualized resource scheduling strategy, resource reallocation is triggered, and a global resource state mapping table is updated to achieve effective scheduling of virtualized computing resources of the power wireless local area network.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of cloud computing, in particular to a virtualized computing resource scheduling method and system based on a power wireless local area network. BACKGROUND

[0002] In the field of power wireless local area networks, with the increasing number of access devices and the increasing complexity of power device operation, effective scheduling of virtualized computing resources has become a key problem to be solved.

[0003] Existing power wireless local area network resource scheduling technologies are mostly based on simple device connection states and fixed resource allocation rules. Generally, resource scheduling is performed only according to whether a device is online and a preset rough resource allocation ratio, which completely ignores the rich information of power device operation load data. In actual operation, key data such as real-time current fluctuation characteristics, voltage phase offset characteristics, and device operation cycle parameters of power devices contain important clues about device operation state and resource demand, but existing technologies fail to effectively utilize these data.

[0004] Due to the lack of in-depth mining of load characteristics, existing technologies cannot accurately grasp the actual resource demand of each access device. When allocating resources, only a one-size-fits-all approach is used, resulting in resource surplus for some devices, causing resource waste, and resource deficiency for some devices, affecting the normal operation of the devices and the processing efficiency of power services. At the same time, existing resource scheduling strategies are often static and cannot be dynamically adjusted according to real-time load changes of devices, making it difficult to adapt to scenarios with rapid changes in load in power wireless local area networks. SUMMARY

[0005] In view of the above-mentioned problems, in combination with the first aspect of the present application, the present application embodiment provides a virtualized computing resource scheduling method based on a power wireless local area network, the method comprising:

[0006] obtaining power device operation load data of all access devices within the coverage range of the power wireless local area network, the power device operation load data including real-time current fluctuation characteristics, voltage phase offset characteristics, and device operation cycle parameters;

[0007] extracting load characteristics from the power device operation load data to obtain a load characteristic set of each access device, the load characteristic set including device power consumption fluctuation characteristics, load response time efficiency characteristics, and resource request priority characteristics;

[0008] performing dynamic resource demand prediction processing on the load feature set based on a preset load prediction model, to generate a virtualization resource demand prediction result for each access device, the virtualization resource demand prediction result including a computing resource allocation magnitude and a storage resource allocation time-efficiency requirement;

[0009] generating a virtualization resource scheduling strategy based on the virtualization resource demand prediction result, the virtualization resource scheduling strategy including a resource allocation topology of an edge computing node and a resource redundancy backup instruction of a central cloud node;

[0010] performing dynamic adjustment on virtualization computing resources of the power wireless local area network based on the virtualization resource scheduling strategy, to trigger a resource reallocation operation and update a global resource state mapping table.

[0011] In still another aspect, an embodiment of the present application also provides a virtualization computing resource scheduling system based on a power wireless local area network, including a processor, a machine readable storage medium, the machine readable storage medium being connected with the processor, the machine readable storage medium being used for storing programs, instructions or codes, and the processor being used for executing the programs, instructions or codes in the machine readable storage medium to realize the above method.

[0012] Based on the above aspects, an embodiment of the present application integrates power equipment operation load data of access devices within a coverage range of a power wireless local area network, and then constructs a dynamic resource demand prediction mechanism based on a preset load prediction model, to generate a virtualization resource demand prediction result for each access device, which not only explicitly defines a computing resource allocation magnitude, but also defines a storage resource allocation time-efficiency requirement. On this basis, a generated virtualization resource scheduling strategy combines a resource allocation topology of an edge computing node and a resource redundancy backup instruction of a central cloud node, to form a multi-level and high-elasticity resource scheduling system, which not only guarantees efficient utilization of local resources, but also ensures reliable backup of global resources. Finally, by dynamically adjusting virtualization computing resources of the power wireless local area network, a resource reallocation operation is triggered and a global resource state mapping table is updated in real time, to realize intelligentization, automation and global optimization of resource scheduling, and significantly improve resource utilization efficiency, response speed and system stability of the power wireless local area network in a complex load environment. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1 is an execution flow schematic diagram of a virtualization computing resource scheduling method based on a power wireless local area network provided by an embodiment of the present application.

[0014] Figure 2 is a schematic diagram of exemplary hardware and software components of a virtualization computing resource scheduling system based on a power wireless local area network provided by an embodiment of the present application. DETAILED DESCRIPTION

[0015] The application will be described in detail below with reference to the accompanying drawings, Figure 1 is a flowchart of a virtualized computing resource scheduling method based on a power wireless local area network provided by an embodiment of the application, and the virtualized computing resource scheduling method based on the power wireless local area network will be described in detail below.

[0016] Step S110: Obtain power equipment operation load data of all access devices in the coverage range of the power wireless local area network, wherein the power equipment operation load data includes real-time current fluctuation characteristics, voltage phase offset characteristics, and device operation cycle parameters.

[0017] In the actual operation scenario of the power wireless local area network, there are various access devices, such as smart meters, power sensors, monitoring cameras, etc., which generate different power loads in the operation process. In order to achieve accurate scheduling of virtualized computing resources, it is necessary to first obtain the power equipment operation load data of these access devices. The real-time current fluctuation characteristics can reflect the change of the current in the operation process of the device, the voltage phase offset characteristics reflect the deviation between the voltage and the standard phase, and the device operation cycle parameters describe the periodicity of the device in performing tasks. The specific implementation is as follows:

[0018] Step S111: Send a periodic load monitoring instruction to each access device in the power wireless local area network, wherein the periodic load monitoring instruction includes a device identifier and a monitoring time window parameter.

[0019] In a power wireless local area network including multiple access devices, the central scheduling node is used to perform the task of sending instructions to each access device. Taking a power wireless local area network with 50 access devices as an example, the central scheduling node can send a periodic load monitoring instruction to each access device in a certain order. The device identifier is the unique identifier of each access device, which is used to distinguish different devices. For example, the identifier of device 1 can be set as “Device001”, the identifier of device 2 is “Device002”, and so on, until the identifier of device 50 is “Device050”. The monitoring time window parameter specifies the time period for data collection, which is to ensure that the load data of each device is obtained within a unified time range for effective comparison and analysis. It is assumed that the monitoring time window is set to 30 minutes from the current time, that is, the central scheduling node will require each access device to monitor and record its load data within the next 30 minutes.

[0020] Step S112: Receive the original load monitoring data packet returned by each access device within the monitoring time window, wherein the original load monitoring data packet includes a current waveform sampling sequence, a voltage waveform sampling sequence, and a device operation state log.

[0021] Upon receiving the periodic load monitoring instruction, each access device starts data collection within the specified monitoring time window. Taking the access device "Device001" as an example, it can sample the current and voltage at a certain sampling frequency. Assuming the sampling frequency is 10 times per second, within a 30-minute (i.e., 1800 seconds) monitoring time window, 18000 current sampling values and 18000 voltage sampling values can be obtained, which are arranged in time sequence to form a current waveform sampling sequence and a voltage waveform sampling sequence. At the same time, the running state of itself can be recorded, including the start time, end time, resource usage, and other information, which are sorted into a device running state log. After the end of the monitoring time window, the access device can send the original load monitoring data packet containing the current waveform sampling sequence, voltage waveform sampling sequence, and device running state log back to the central scheduling node. The central scheduling node receives and stores these data packets for subsequent further processing and analysis.

[0022] Step S113: performing fluctuation feature extraction processing on the current waveform sampling sequence to obtain the real-time current fluctuation feature, wherein the real-time current fluctuation feature includes a current peak value offset, a waveform distortion rate, and a frequency harmonic distribution parameter.

[0023] Taking the current waveform sampling sequence of the access device "Device001" as an example, first, the current peak value offset is calculated. The current peak value offset refers to the difference between the actual current peak value and the theoretical current peak value. Under normal circumstances, the current peak value of the device should be within a relatively stable range, but due to the influence of factors such as the running state of the device and the external environment, the actual current peak value may deviate from the theoretical value. By analyzing and comparing each peak value in the current waveform sampling sequence, the current peak value offset can be calculated. For example, in a certain sampling period, the theoretical current peak value is 5 amperes, and the actual current peak value is 5.2 amperes, so the current peak value offset in that sampling period is 0.2 amperes.

[0024] Then, the waveform distortion rate is calculated. The waveform distortion rate reflects the degree of deviation of the current waveform from the standard sine wave. By comparing the actual current waveform with the standard sine wave, the difference between them can be calculated to obtain the waveform distortion rate. Specifically, each sampling point in the current waveform sampling sequence can be compared with the corresponding standard sine wave sampling point to calculate the sum of the squares of the differences between them, then take the average and take the square root to obtain the waveform distortion rate. Assuming that the current waveform distortion rate of the access device "Device001" is 0.05, which means that the current waveform of the device deviates from the standard sine wave to a certain extent.

[0025] Finally, the frequency harmonic distribution parameters are calculated. Frequency harmonics refer to the frequency components in the current waveform other than the fundamental frequency. By performing Fourier transform on the current waveform sampling sequence, it can be converted from the time domain to the frequency domain, thereby obtaining the frequency component distribution of the current waveform. In the frequency domain, the amplitude and phase of each frequency harmonic can be analyzed, and these parameters constitute the frequency harmonic distribution parameters. For example, after Fourier transform, it is detected that in the current waveform of the access device "Device001", there are 3rd and 5th harmonics in addition to the fundamental frequency, and their amplitudes are 0.2 amperes and 0.1 amperes, and their phases are 30 degrees and 60 degrees, respectively. The above amplitude and phase information is part of the frequency harmonic distribution parameters. The current peak value offset, waveform distortion rate and frequency harmonic distribution parameters calculated are combined together, and the real-time current fluctuation characteristics of the access device "Device001" are obtained.

[0026] Step S114: performing phase offset analysis processing on the voltage waveform sampling sequence to obtain the voltage phase offset characteristics, wherein the voltage phase offset characteristics include phase synchronization error, voltage drop amplitude and transient response delay parameters.

[0027] Similarly, the central scheduling node can perform phase offset analysis processing on the voltage waveform sampling sequence. Taking the voltage waveform sampling sequence of the access device "Device001" as an example, first, the phase synchronization error is calculated. The phase synchronization error refers to the difference between the actual voltage phase and the standard voltage phase. In the power system, the phase of the voltage should be synchronized with the standard phase, but due to various factors, the actual voltage phase may deviate. By analyzing and comparing the phase of each sampling point in the voltage waveform sampling sequence, the phase synchronization error can be calculated. For example, in a certain sampling period, the standard voltage phase is 0 degrees, and the actual voltage phase is 5 degrees, so the phase synchronization error in this sampling period is 5 degrees.

[0028] Then, the voltage drop amplitude is calculated. Voltage drop refers to the phenomenon that the voltage suddenly drops in a short time. The voltage drop amplitude can be calculated by comparing the maximum voltage value and the minimum voltage value in the voltage waveform sampling sequence. Suppose that in a certain time period, the maximum voltage value in the voltage waveform sampling sequence of the access device "Device001" is 220 volts, and the minimum voltage value is 200 volts, then the voltage drop amplitude is 20 volts.

[0029] Finally, the transient response delay parameter is calculated. The transient response delay refers to the time required for the voltage to recover to a stable state when the device is subjected to external interference or load changes. By analyzing the changes in the voltage waveform sample sequence after being disturbed, the time interval from when the voltage begins to drop to when it recovers to a stable state can be determined, which is the transient response delay parameter. For example, when the load of the access device "Device001" suddenly increases, the voltage begins to drop, and after 50 milliseconds, the voltage recovers to a stable state. Therefore, the transient response delay parameter is 50 milliseconds. By combining the phase synchronization error, voltage drop amplitude, and transient response delay parameters calculated above, the voltage phase offset characteristics of the access device "Device001" are obtained.

[0030] Step S115: Analyze the periodic task trigger records in the device running state log to generate the device running period parameter, which includes the task execution interval duration, task processing time threshold, and resource occupation upper limit.

[0031] Taking the device running state log of the access device "Device001" as an example, the task execution interval duration is analyzed first. The task execution interval duration refers to the time interval between the triggering of two adjacent tasks. By analyzing the task trigger times in the device running state log, the task execution interval duration can be calculated. For example, if the device running state log records that task A was triggered at the 10th minute, the 20th minute, and the 30th minute, then the task execution interval duration of task A is 10 minutes.

[0032] Next, the task processing time threshold is determined. The task processing time threshold refers to the maximum time allowed for the device to complete a task. By analyzing the task start and end times in the device running state log, the processing time of different tasks can be calculated, and a reasonable task processing time threshold can be determined based on these data. Suppose that after analyzing the device running state log, it is detected that the processing time of task B is mostly within 5 minutes, then the task processing time threshold of task B can be set to 6 minutes.

[0033] Finally, the upper limit of resource occupation ratio is calculated. The upper limit of resource occupation ratio refers to the maximum resource ratio that the device is allowed to occupy during task execution. By analyzing the resource usage in the device running state log, the resource ratio occupied by different tasks during execution can be counted, and a reasonable upper limit of resource occupation ratio can be determined according to these data. For example, after analyzing the device running state log, it is detected that the CPU resource ratio occupied by task C during execution is mostly within 30%, so the upper limit of resource occupation ratio of task C can be set to 35%. Combining these calculated task execution interval duration, task processing time threshold and resource occupation ratio upper limit, the device running cycle parameters of the access device "Device001" are obtained.

[0034] Step S120: Load feature extraction is performed on the power device running load data to obtain a load feature set of each access device, which includes device power consumption fluctuation features, load response time efficiency features and resource request priority features.

[0035] After obtaining the power device running load data, further processing is needed to extract key features that can reflect the device load condition and form a load feature set. The device power consumption fluctuation feature can reflect the change of power consumption during device operation, the load response time efficiency feature reflects the response speed and timeliness of the device to resource requests, and the resource request priority feature is used to determine the priority of the device in resource allocation.

[0036] Step S121: A device power consumption calculation function is constructed according to the real-time current fluctuation feature and the voltage phase offset feature, and the device power consumption fluctuation feature is generated based on the device power consumption calculation function, wherein the device power consumption fluctuation feature includes unit time power consumption change rate, peak power consumption duration and low power consumption state switching frequency.

[0037] Taking the access device "Device001" as an example, the central scheduling node will construct a device power consumption calculation function according to the previously calculated real-time current fluctuation feature and voltage phase offset feature. The current peak offset, waveform distortion rate and frequency harmonic distribution parameters in the real-time current fluctuation feature, as well as the phase synchronization error, voltage drop amplitude and transient response delay parameters in the voltage phase offset feature will all affect the power consumption of the device. By considering these factors comprehensively, a function that can accurately calculate the power consumption of the device can be constructed.

[0038] After obtaining the device power consumption calculation function, the device power consumption fluctuation characteristics can be calculated. First, the unit time power consumption change rate is calculated. The unit time power consumption change rate refers to the change proportion of the power consumption of the device in unit time. The power consumption change amount in unit time can be obtained by calculating and comparing the power consumption of the device at different time points, and then the power consumption change amount in unit time is divided by the average power consumption of the time period to obtain the unit time power consumption change rate. For example, in a certain minute, the initial power consumption of the access device "Device001" is 100 watts, and the power consumption at the end is 110 watts, so the power consumption change amount in the minute is 10 watts, assuming that the average power consumption in the minute is 105 watts, then the unit time power consumption change rate is 10 ÷ 105 ≈ 0.095, that is, 9.5%.

[0039] Then the peak power consumption duration is calculated. The peak power consumption duration refers to the time length that the device maintains the peak power consumption after reaching the peak power consumption. The time point of the occurrence of the peak power consumption and the time point of the end can be determined by analyzing the power consumption curve of the device with time, and the time interval between the two is the peak power consumption duration. For example, the access device "Device001" reaches a peak power consumption of 200 watts in a certain time period, and the time interval from reaching the peak power consumption to the power consumption dropping to 190 watts is 5 minutes, so the peak power consumption duration is 5 minutes.

[0040] Finally, the low-power state switching frequency is calculated. The low-power state switching frequency refers to the frequency of entering and exiting the low-power state of the device during operation. The number of times of entering the low-power state (i.e. the power consumption is lower than a certain set threshold) and the number of times of exiting the low-power state can be counted by analyzing the power consumption curve of the device with time, and the sum of the two is the low-power state switching frequency. Assuming that in an hour, the access device "Device001" enters the low-power state 5 times and exits the low-power state 5 times, then the low-power state switching frequency is 10 times / hour. Thus, the unit time power consumption change rate, the peak power consumption duration and the low-power state switching frequency calculated are combined together to obtain the device power consumption fluctuation characteristics of the access device "Device001".

[0041] Step S122: perform response time modeling processing on the device operation period parameters to generate the load response time characteristics, wherein the load response time characteristics include task processing delay tolerance, resource request response timeout threshold and real-time performance guarantee level identifier.

[0042] Taking the device running cycle parameter of the access device "Device001" as an example, first, the task processing delay tolerance is determined. The task processing delay tolerance refers to the task processing delay time that the device can tolerate. The task processing delay tolerance can be determined according to the task execution interval time length and the task processing time threshold in the device running cycle parameter. For example, if the task execution interval time length is 10 minutes and the task processing time threshold is 5 minutes, the task processing delay tolerance can be set to 3 minutes, that is, the device starts processing the task within 3 minutes after the task is triggered, which is considered to be within the tolerable range.

[0043] Then, the resource request response timeout threshold is determined. The resource request response timeout threshold refers to the longest time that the device can wait for a response after sending a resource request. The resource request response timeout threshold can be determined according to the business needs of the device and the urgency of the task. For example, for some tasks with high real-time requirements, the resource request response timeout threshold can be set to a short time, such as 1 second; and for some tasks with low real-time requirements, the resource request response timeout threshold can be set to a long time, such as 10 seconds. Assuming that a task of the access device "Device001" has high real-time requirements, the resource request response timeout threshold of the task can be set to 2 seconds.

[0044] Finally, the real-time guarantee level identifier is determined. The real-time guarantee level identifier is used to represent the degree of requirement of the device for real-time. The real-time guarantee level identifier can be determined according to the task processing delay tolerance and the resource request response timeout threshold. For example, the real-time guarantee level can be divided into three levels: high, medium and low. If the task processing delay tolerance is short and the resource request response timeout threshold is also short, the real-time guarantee level identifier can be set to "high"; if the task processing delay tolerance and the resource request response timeout threshold are moderate, the real-time guarantee level identifier can be set to "medium"; if the task processing delay tolerance is long and the resource request response timeout threshold is also long, the real-time guarantee level identifier can be set to "low". Assuming that the task processing delay tolerance of the access device "Device001" is 2 minutes and the resource request response timeout threshold is 3 seconds, the real-time guarantee level identifier of the access device "Device001" can be set to "medium". Combining the determined task processing delay tolerance, resource request response timeout threshold and real-time guarantee level identifier, the load response timeliness feature of the access device "Device001" is obtained.

[0045] Step S123: performing priority weight allocation processing according to the resource request records in the device running state log to generate the resource request priority feature, the resource request priority feature including an emergency task trigger identifier, a key device identifier and a quality of service level parameter.

[0046] Taking the device running state log of the access device "Device001" as an example, first, the emergency task trigger identifier is determined. The emergency task trigger identifier is used to indicate whether the device triggers an emergency task. By analyzing the task type and task description in the device running state log, it can be judged whether there is an emergency task. For example, if the device running state log records that a certain task is "emergency fault handling task", the emergency task trigger identifier of the device can be set to "yes"; otherwise, it is set to "no".

[0047] Then the key device identifier is determined. The key device identifier is used to indicate the importance of the device in the power system. The key device identifier can be determined according to the function and role of the device. For example, for some devices that undertake important monitoring and control tasks, their key device identifier can be set to "key device"; while for some ordinary auxiliary devices, their key device identifier can be set to "ordinary device". Assuming that the access device "Device001" is an important device for monitoring the voltage of the power system, its key device identifier can be set to "key device".

[0048] Finally, the service quality level parameter is determined. The service quality level parameter is used to indicate the quality requirement of the device for resource allocation. The service quality level parameter can be determined according to the business demand and real-time guarantee level identifier of the device. For example, the service quality level can be divided into three levels: high, medium and low. If the real-time guarantee level identifier of the device is "high" and the quality requirement of the business demand for resources is high, the service quality level parameter can be set to "high"; if the real-time guarantee level identifier of the device is "medium" and the quality requirement of the business demand for resources is moderate, the service quality level parameter can be set to "medium"; if the real-time guarantee level identifier of the device is "low" and the quality requirement of the business demand for resources is low, the service quality level parameter can be set to "low". Assuming that the real-time guarantee level identifier of the access device "Device001" is "medium" and the quality requirement of its business demand for resources is moderate, its service quality level parameter can be set to "medium". Combining the determined emergency task trigger identifier, key device identifier and service quality level parameter, the resource request priority feature of the access device "Device001" is obtained.

[0049] Step S124: The device power consumption fluctuation feature, the load response time efficiency feature and the resource request priority feature are normalized and spliced to obtain the load feature set, which is stored in the global resource state mapping table.

[0050] In this embodiment, the normalization processing is to eliminate the difference in dimension and numerical range between different features, so that each feature can be spliced and compared in a unified scale.

[0051] First, the device power fluctuation feature is normalized. Taking the access device "Device001" as an example, its device power fluctuation feature includes the unit time power consumption change rate, the peak power consumption duration and the low power state switching frequency. For the unit time power consumption change rate, assuming that its value range is between -0.1 and 0.2, it is normalized to the range of 0 to 1. The linear normalization method can be used, first find the minimum value and the maximum value of the feature in all access devices, assuming that the minimum value is -0.1 and the maximum value is 0.2. The unit time power consumption change rate of "Device001" is 0.095, the normalized value calculation process is: subtract the minimum value from the unit time power consumption change rate of the device, that is, 0.095-(-0.1) = 0.195, and then divide by the difference between the maximum value and the minimum value, that is, 0.2-(-0.1) = 0.3, to get the normalized value 0.195÷0.3≈0.65.

[0052] For the peak power consumption duration, assuming that the peak power consumption duration of all access devices ranges from 1 minute to 10 minutes, the peak power consumption duration of "Device001" is 5 minutes. Similarly, the linear normalization method is used, subtracting the minimum value 1 minute from the peak power consumption duration of the device, resulting in 5-1 = 4 minutes, and then dividing by the difference between the maximum value and the minimum value 10-1 = 9 minutes, the normalized value is 4÷9≈0.44.

[0053] For the low power state switching frequency, assuming that the low power state switching frequency of all access devices ranges from 2 times / hour to 20 times / hour, the low power state switching frequency of "Device001" is 10 times / hour. Subtracting the minimum value 2 times / hour from the low power state switching frequency of the device, resulting in 10-2 = 8 times / hour, and then dividing by the difference between the maximum value and the minimum value 20-2 = 18 times / hour, the normalized value is 8÷18≈0.44.

[0054] After the above processing, the device power fluctuation feature of "Device001" is normalized to a vector containing three normalized values (0.65, 0.44, 0.44).

[0055] Then the load response timeliness features are normalized. The load response timeliness features include task processing delay tolerance, resource request response timeout threshold, and real-time guarantee level identifier. For the task processing delay tolerance, it is assumed that the task processing delay tolerance of all access devices ranges from 1 minute to 10 minutes, and the task processing delay tolerance of "Device001" is 3 minutes. A linear normalization method is used, and the task processing delay tolerance of the device is subtracted from the minimum value of 1 minute to obtain 3-1 = 2 minutes, and then divided by the difference between the maximum value and the minimum value of 10-1 = 9 minutes, and the normalized value is 2 ÷ 9 = 0.22.

[0056] For the resource request response timeout threshold, it is assumed that the resource request response timeout threshold of all access devices ranges from 0.5 seconds to 10 seconds, and the resource request response timeout threshold of "Device001" is 2 seconds. The resource request response timeout threshold of the device is subtracted from the minimum value of 0.5 seconds to obtain 2-0.5 = 1.5 seconds, and then divided by the difference between the maximum value and the minimum value of 10-0.5 = 9.5 seconds, and the normalized value is 1.5 ÷ 9.5 = 0.16.

[0057] For the real-time guarantee level identifier, since it is a categorical variable (high, medium, low), it can be processed using one-hot encoding. It is assumed that "high" is encoded as (1, 0, 0), "medium" is encoded as (0, 1, 0), and "low" is encoded as (0, 0, 1). The real-time guarantee level identifier of "Device001" is "medium", and after encoding it is (0, 1, 0).

[0058] The normalized values of the task processing delay tolerance and the resource request response timeout threshold are concatenated with the encoding of the real-time guarantee level identifier, and the normalized load response timeliness features of "Device001" are a vector containing five values (0.22, 0.16, 0, 1, 0).

[0059] Then the resource request priority features are normalized. The resource request priority features include emergency task trigger identifier, critical device identifier, and quality of service level parameter. For the emergency task trigger identifier, it is a binary variable (yes, no), which can be represented by 1 for "yes" and 0 for "no". It is assumed that the emergency task trigger identifier of "Device001" is "no", and its value is 0.

[0060] For the critical device identifier, it is also a categorical variable (critical device, ordinary device), and one-hot encoding is used. It is assumed that "critical device" is encoded as (1, 0) and "ordinary device" is encoded as (0, 1). The critical device identifier of "Device001" is "critical device", and after encoding it is (1, 0).

[0061] For the service quality level parameter, which is also a categorical variable (high, medium, low), one-hot encoding is used. Assuming that "high" is encoded as (1, 0, 0), "medium" as (0, 1, 0), and "low" as (0, 0, 1), the service quality level parameter of "Device001" is "medium", and after encoding, it becomes (0, 1, 0).

[0062] The value of the emergency task trigger identifier, the key device identifier, and the encoding of the service quality level parameter are spliced together, and the resource request priority feature of "Device001" is normalized into a vector containing four values (0, 1, 0, 0, 1, 0).

[0063] After normalizing each feature, the splicing process is performed. The vector (0.65, 0.44, 0.44) of the device power consumption fluctuation feature of "Device001", the vector (0.22, 0.16, 0, 1, 0) of the load response time efficiency feature, and the vector (0, 1, 0, 0, 1, 0) of the resource request priority feature are spliced in order to obtain a vector containing twelve values (0.65, 0.44, 0.44, 0.22, 0.16, 0, 1, 0, 0, 1, 0, 0). This vector is the load feature set of "Device001".

[0064] Finally, the load feature set is associated and stored in the global resource state mapping table. The global resource state mapping table is a database used to record the resource state and related features of all access devices in the power wireless local area network. When storing, the identifier "Device001" of the access device is used as the index, and the load feature set is associated with the device. For example, a new record is created in the global resource state mapping table, the primary key of the record is "Device001", and the load feature set (0.65, 0.44, 0.44, 0.22, 0.16, 0, 1, 0, 0, 1, 0, 0) is stored in the load feature field corresponding to the primary key. In this way, in the subsequent resource scheduling process, the corresponding load feature set can be quickly queried according to the device identifier, providing a basis for resource allocation.

[0065] Step S130: based on the preset load prediction model, the load feature set is processed for dynamic resource demand prediction, and a virtualized resource demand prediction result of each access device is generated, which includes a computing resource allocation magnitude and a storage resource allocation time efficiency requirement.

[0066] After obtaining the load feature set of each access device, it is necessary to utilize the preset load prediction model to perform dynamic resource demand prediction processing on the load feature set, so as to determine the virtualization resources required by each access device in the future. The computing resource allocation magnitude is used to determine the number of computing resources allocated to the device, and the storage resource allocation time requirement is used to specify the time requirement of the device for reading and writing and backup of the storage resource. The specific prediction process is as follows:

[0067] Step S131: calling a pre-trained load prediction model to perform time series analysis processing on the load feature set, to generate a short-term resource demand prediction sequence and a long-term resource demand trend distribution; wherein the short-term resource demand prediction sequence contains a computing resource demand peak value and a storage resource demand fluctuation range within a preset time period in the future; and the long-term resource demand trend distribution contains a resource occupation growth slope, a periodic task superposition influence factor, and a burst load event trigger probability.

[0068] Taking the access device "Device001" as an example, a pre-trained load prediction model is called to perform time series analysis processing on its load feature set (0.65, 0.44, 0.44, 0.22, 0.16, 0, 1, 0, 0, 1, 0, 0). The pre-trained load prediction model is obtained by training a large amount of historical data, and it can learn the relationship between the load feature set and the resource demand.

[0069] In the short-term resource demand prediction, it is assumed that the preset short-term time period is 1 hour in the future. The load prediction model will predict the computing resource demand peak value and the storage resource demand fluctuation range within 1 hour in the future according to the load feature set of "Device001", combined with historical data and the current running state. For the prediction of the computing resource demand peak value, the model will analyze the device power fluctuation feature, the load response time feature, and the resource request priority feature in the load feature set, and consider factors such as the task execution of the device, the real-time requirement, and the resource request priority. For example, if the real-time guarantee level of the device is "high" and the task processing delay tolerance is short, the model will predict that the device may have a high computing resource demand peak value within 1 hour in the future. It is assumed that after prediction, the computing resource demand peak value of "Device001" within 1 hour in the future is 80% CPU usage and 60% memory usage.

[0070] For the prediction of the storage resource demand fluctuation range, the model will consider the business demand and data generation of the device. If the device is a data acquisition device, it will continuously generate a large amount of data, and the model will predict that its storage resource demand fluctuation range is large. It is assumed that the storage resource demand fluctuation range of "Device001" within 1 hour in the future is between 10GB and 20GB.

[0071] When performing long-term resource demand trend distribution prediction, a long-term time period of 1 month in the future is assumed. The load prediction model analyzes the periodic task trigger records and resource request history in the load feature set, calculates the resource occupancy growth slope, periodic task superposition influence factor, and burst load event trigger probability. The resource occupancy growth slope reflects the growth rate of the device's resource occupancy over a long period. Through analysis of historical data, if it is detected that the computing resource occupancy rate of "Device001" grows at a rate of 5% per month, then the resource occupancy growth slope is 5%.

[0072] The periodic task superposition influence factor takes into account the impact of the device's periodic tasks on resource demand. If "Device001" has multiple periodic tasks that execute simultaneously in certain time periods, the periodic task superposition influence factor will be large. Assuming that through analysis, the periodic task superposition influence factor of "Device001" is 1.2, indicating that due to the superposition of periodic tasks, resource demand will increase by 20%.

[0073] The burst load event trigger probability refers to the likelihood that the device will encounter a burst load event in the future. Through analysis of abnormal situations in historical data and the device's operating environment, it is predicted that the burst load event trigger probability of "Device001" in the next 1 month is 10%.

[0074] Combining the short-term resource demand prediction sequence (CPU usage of 80% and memory usage of 60% for a computing resource demand peak, and a storage resource demand fluctuation range of 10GB to 20GB) and the long-term resource demand trend distribution (resource occupancy growth slope of 5%, periodic task superposition influence factor of 1.2, and burst load event trigger probability of 10%) together, the resource demand prediction result of "Device001" is obtained.

[0075] Step S132: Perform resource allocation magnitude calculation processing according to the short-term resource demand prediction sequence and the long-term resource demand trend distribution to generate the computing resource allocation magnitude, wherein the computing resource allocation magnitude includes CPU core number demand, memory capacity demand, and GPU computing power allocation proportion.

[0076] The computing resource allocation magnitude is calculated according to the short-term resource demand prediction sequence and long-term resource demand trend distribution of "Device001". First, the CPU core number demand is calculated. Considering that the computing resource demand peak within the next 1 hour is 80% CPU usage, and the long-term resource occupation growth slope is 5% and the periodic task superposition influence factor is 1.2. Assuming that the current device uses 4 CPU cores, and the usage rate of each CPU core is evenly distributed under normal circumstances. Since the computing resource demand peak is 80%, in order to ensure that the device can run normally under peak load, a certain number of CPU cores need to be added. According to experience and historical data, when the CPU usage rate exceeds 70%, performance bottlenecks may occur. Therefore, in order to cope with future growth and the impact of periodic task superposition, the CPU core number needs to be appropriately increased.

[0077] The peak CPU usage rate after considering growth and superposition impact can be calculated first, that is, 80% x (1+5%) x 1.2 = 100.8%. This means that the current 4 CPU cores may not meet the demand. In order to control the CPU usage rate within a reasonable range, assuming that the target CPU usage rate is controlled at 70%, the required CPU core number is 4 x 100.8% ÷ 70% ≈ 5.76, rounded up to 6 CPU cores.

[0078] Next, the memory capacity demand is calculated. Considering that the storage resource demand fluctuation range within the next 1 hour is between 10GB and 20GB, and the impact of long-term resource occupation growth and periodic task superposition. Assuming that the current device has a memory capacity of 16GB, due to the large fluctuation range of storage resource demand, and the existence of growth and superposition impact, a certain amount of memory capacity needs to be added. According to the upper limit of storage resource demand 20GB, and considering the growth and superposition impact coefficient 1.2 x (1+5%) = 1.26, the required memory capacity is calculated as 20GB x 1.26 = 25.2GB, rounded up to 32GB.

[0079] Finally, the GPU computing power allocation proportion is calculated. If "Device001" business demand involves some graphics processing or deep learning tasks, a certain amount of GPU computing power needs to be allocated. According to the device load feature set and resource demand prediction result, assuming that after analysis, 30% of the workload of "Device001" tasks can be processed by GPU acceleration, then the GPU computing power allocation proportion is 30%.

[0080] Combining the calculated CPU core number demand of 6, memory capacity demand of 32GB and GPU computing power allocation proportion of 30%, the computing resource allocation magnitude of "Device001" is obtained.

[0081] Step S133: Perform storage resource timeliness matching processing based on the resource request priority feature and the load response timeliness feature, to generate the storage resource allocation timeliness requirement, which includes a data read-write delay upper limit, a storage bandwidth guarantee threshold, and a redundant backup response time.

[0082] The storage resource timeliness matching processing is performed according to the resource request priority feature and the load response timeliness feature of “Device001”. First, the data read-write delay upper limit is determined. The quality of service level parameter in the resource request priority feature of “Device001” is “medium”, and the real-time guarantee level identifier in the load response timeliness feature is “medium”, which indicates that the device has moderate real-time requirements for data read-write. According to experience and historical data, for devices with a service quality level of “medium” and a real-time guarantee level of “medium”, the data read-write delay upper limit can be set to 100 milliseconds.

[0083] Next, the storage bandwidth guarantee threshold is determined. Considering the storage resource demand fluctuation range of “Device001” within the next 1 hour, which is between 10GB and 20GB, as well as its resource request priority and load response timeliness feature. In order to ensure that the device can complete the data read-write operation within the specified time, a certain storage bandwidth guarantee needs to be provided. Assuming that according to the upper limit of storage resource demand 20GB and the time range 1 hour (3600 seconds), and considering a certain redundancy and fluctuation, the required storage bandwidth is calculated as 20GB ÷ 3600 seconds × 1.2 ≈ 6.67MB / s, rounded up to 7MB / s, which is the storage bandwidth guarantee threshold.

[0084] Finally, the redundant backup response time is determined. Since “Device001” is a critical device, it needs to be backed up redundantly to prevent data loss. According to its resource request priority feature and load response timeliness feature, considering that its real-time guarantee level is “medium”, the redundant backup response time can be set to 5 minutes. This means that after the device data changes, the redundant backup of the data needs to be completed within 5 minutes.

[0085] Combining the determined data read-write delay upper limit of 100 milliseconds, storage bandwidth guarantee threshold of 7MB / s, and redundant backup response time of 5 minutes, the storage resource allocation timeliness requirement of “Device001” is obtained.

[0086] Step S134: Perform joint encoding processing on the computing resource allocation level and the storage resource allocation timeliness requirement, to generate the virtualization resource demand prediction result, and write the virtualization resource demand prediction result into the global resource state mapping table.

[0087] The computing resource allocation level (CPU core requirement of 6, memory capacity requirement of 32 GB, and GPU computing power allocation ratio of 30%) and the storage resource allocation time limit requirement (data read-write delay upper limit of 100 ms, storage bandwidth guarantee threshold of 7 MB / s, and redundant backup response time of 5 minutes) of "Device001" are jointly encoded. A fixed encoding format can be used to convert these information into a unified string or vector. For example, the computing resource allocation level and the storage resource allocation time limit requirement can be arranged in a certain order and separated by a separator to form an encoded string: "CPU core number: 6, memory capacity: 32 GB, GPU computing power allocation ratio: 30%, data read-write delay upper limit: 100 ms, storage bandwidth guarantee threshold: 7 MB / s, redundant backup response time: 5 minutes".

[0088] The encoded string is used as the virtualization resource demand prediction result of "Device001", and is written into the global resource state mapping table. In the global resource state mapping table, the device identifier of "Device001" is used as the index to find the corresponding record item. The global resource state mapping table is usually stored in a database, and the database management system provides a series of operation interfaces to realize the writing of data.

[0089] First, a connection with the database is established. This requires the use of connection tools provided by the database management system. For example, for a relational database MySQL, a corresponding MySQL driver can be used to create a database connection object. Assuming that the database server address used is "192.168.1.100", the port number is 3306, the database name is "power_wlan_resource", the username is "admin", and the password is "password". Through the connection function provided by the driver, these parameters are passed in to establish a connection with the database.

[0090] Next, a SQL statement is constructed to update the record in the global resource status mapping table. Since the virtualized resource demand prediction result is to be written into the record corresponding to "Device001", an update statement needs to be written. Assume that the global resource status mapping table is named "device_resource_status", which contains the field "device_id" for storing the device identifier and the field "virtual_resource_prediction" for storing the virtualized resource demand prediction result. The constructed SQL update statement is: "UPDATE device_resource_status SET virtual_resource_prediction = 'CPU core number: 6, memory capacity: 32 GB, GPU computing power allocation ratio: 30%, data read-write delay upper limit: 100 ms, storage bandwidth guarantee threshold: 7 MB / s, redundant backup response time: 5 minutes' WHERE device_id = 'Device001'".

[0091] Then, the constructed SQL statement is sent to the database for execution. Through the previously established database connection object, the method for executing SQL statements is called, and the above update statement is passed in as a parameter. The database management system will parse and execute the statement, find the record with "device_id" as "Device001" in the "device_resource_status" table, and update the value of the "virtual_resource_prediction" field to the encoded virtualized resource demand prediction result.

[0092] After executing the SQL statement, the execution result needs to be checked. The database management system will return an execution status code or related information to determine whether the update operation is successful. If the execution status code indicates that the operation is successful, it means that the virtualized resource demand prediction result of "Device001" has been successfully written into the global resource status mapping table. If an error occurs during execution, the database management system will return the corresponding error information, such as SQL syntax error, table does not exist, insufficient permissions, etc. According to these error information, corresponding processing can be carried out, such as checking the correctness of the SQL statement, confirming the correctness of the table structure, checking the user permissions, etc.

[0093] Step S140: generating a virtualized resource scheduling strategy according to the virtualized resource demand prediction result, the virtualized resource scheduling strategy containing a resource allocation topology of an edge computing node and a resource redundant backup instruction of a central cloud node.

[0094] After obtaining the virtualization resource requirement prediction results of each access device, a virtualization resource scheduling strategy needs to be generated according to the results to reasonably allocate and manage the virtualization computing resources in the power wireless local area network. The edge computing node is close to the access device and can provide low-delay computing services, while the central cloud node has strong computing and storage capabilities and is used for redundant backup and global resource coordination of resources. The specific generation process is as follows:

[0095] Step S141: Analyzing the computing resource allocation level in the virtualization resource requirement prediction result, generating a resource allocation topology of the edge computing node, the resource allocation topology including the computing resource allocation proportion of the edge node, the cross-node load balancing path and the resource reservation buffer capacity.

[0096] Taking the virtualization resource requirement prediction result of “Device001” as an example, the computing resource allocation level (CPU core number requirement 6, memory capacity requirement 32 GB and GPU computing power allocation proportion 30%) is analyzed to generate the resource allocation topology of the edge computing node. It is assumed that there are three edge computing nodes in the power wireless local area network, namely “EdgeNode001”, “EdgeNode002” and “EdgeNode003”, each of which has different total amount of computing resources. “EdgeNode001” has 16 CPU cores, 64 GB of memory and a certain amount of GPU computing power, “EdgeNode002” has 8 CPU cores, 32 GB of memory and a certain amount of GPU computing power, and “EdgeNode003” has 12 CPU cores, 48 GB of memory and a certain amount of GPU computing power.

[0097] First, the computing resource allocation proportion of the edge node is determined. According to the computing resource requirement of “Device001” and the total amount of resources of each edge computing node, the principle of resource balanced allocation is used for calculation. For the CPU core number, “Device001” requires 6 CPU cores, and the total number of CPU cores of the three edge computing nodes is 16+8+12=36. The number of CPU cores allocated to “EdgeNode001” is 6x(16÷36)≈2.67, rounded up to 3; the number of CPU cores allocated to “EdgeNode002” is 6x(8÷36)≈1.33, rounded up to 2; and the number of CPU cores allocated to “EdgeNode003” is 6-3-2=1.

[0098] For memory capacity, "Device001" requires 32GB of memory, and the total memory capacity of the three edge computing nodes is 64+32+48=144GB. "EdgeNode001" is allocated a memory capacity of 32x(64÷144)≈14.22GB, rounded up to 15GB; "EdgeNode002" is allocated a memory capacity of 32x(32÷144)≈7.11GB, rounded up to 8GB; and "EdgeNode003" is allocated a memory capacity of 32-15-8=9GB.

[0099] For GPU computing power, the GPU computing power allocation ratio of "Device001" is 30%. Assuming that the GPU computing power of the three edge computing nodes can be allocated proportionally, the same calculation method is used to allocate according to the total GPU computing power of each edge computing node. After calculation, the GPU computing power ratio allocated to "Device001" by each edge computing node is obtained. These CPU core numbers, memory capacities, and GPU computing powers are arranged into the computing resource allocation ratio of the edge nodes.

[0100] Next, the cross-node load balancing path is determined. Considering factors such as network delay and resource utilization, it is necessary to determine the load balancing path between different edge computing nodes for "Device001". By monitoring and analyzing the network connection status between each edge computing node, network delay data between them is obtained. Assuming that the network delay between "Device001" and "EdgeNode001" is 10 milliseconds, the network delay between "Device001" and "EdgeNode002" is 15 milliseconds, and the network delay between "Device001" and "EdgeNode003" is 12 milliseconds. In order to achieve load balancing and reduce delay, when the computing task of "Device001" can be divided, part of the task is preferentially allocated to "EdgeNode001" with lower network delay. At the same time, according to the resource usage of each edge computing node, the allocation ratio of the task is dynamically adjusted. For example, when the resource utilization of "EdgeNode001" reaches 80%, part of the task is transferred to "EdgeNode003". In this way, the cross-node load balancing path of "Device001" between edge computing nodes is formed.

[0101] Finally, the resource reservation buffer capacity is determined. In order to cope with the sudden resource demand, each edge computing node needs to reserve a certain amount of resource buffer. According to the long-term resource demand trend distribution of "Device001" (resource occupation growth slope 5%, periodic task superposition influence factor 1.2, and burst load event trigger probability 10%), as well as the total amount of resources of the edge computing node and the current usage, the resource reservation buffer capacity is calculated. Taking "EdgeNode001" as an example, its current CPU core usage rate is 60%, and the memory usage rate is 50%. Considering the resource demand growth of "Device001" and the possibility of burst load events, the CPU core number reserved for "Device001" is 16x(1-60%)x10%xl.2x(1+5%)≈0.8, rounded up to 1; the reserved memory capacity is 64x(1-50%)x10%xl.2x(1+5%)≈4GB. These reserved resources for "Device001" are part of the resource reservation buffer capacity of "EdgeNode001", and the resource reservation buffer capacities of other edge computing nodes are calculated in the same way. The combination of the computing resource allocation proportion of the edge node, the cross-node load balancing path, and the resource reservation buffer capacity results in the resource allocation topology of the edge computing node for "Device001".

[0102] Step S142: generating a resource redundancy backup instruction of the center cloud node according to the storage resource allocation timeliness requirement, the resource redundancy backup instruction comprising a data backup period parameter, an off-site disaster recovery node identifier, and a backup data verification strategy.

[0103] According to the storage resource allocation timeliness requirement of "Device001" (data read-write delay upper limit 100 ms, storage bandwidth guarantee threshold 7 MB / s, and redundancy backup response time 5 minutes), the resource redundancy backup instruction of the center cloud node is generated.

[0104] First, determine the data backup period parameter. Considering that the storage resource requirement fluctuation range of "Device001" is between 10GB and 20GB, and the redundant backup response time requirement is 5 minutes. In order to ensure the completion of data backup within the specified time, it is necessary to calculate the data backup period according to the storage bandwidth guarantee threshold of 7MB / s. Assuming that the upper limit of 20GB of storage resource requirement is calculated, 20GB is converted to byte, which is 20×1024×1024×1024 bytes, and the storage bandwidth is 7MB / s, which is 7×1024×1024 bytes / s. Then the time required to complete 20GB data backup is (20×1024×1024×1024) ÷ (7×1024×1024) ≈ 2994 seconds, about 50 minutes. But considering the redundant backup response time is 5 minutes, the data backup period needs to be set shorter to ensure that the important data backup can be completed within 5 minutes. After comprehensive consideration, the data backup period parameter is set to 3 minutes, that is, the data of "Device001" is backed up every 3 minutes.

[0105] Then determine the off-site disaster recovery node identifier. The central cloud node usually has multiple off-site disaster recovery nodes for storing backup data to prevent single point failure. According to the network connection stability, storage capacity and geographical location and other factors, select the appropriate node from the preset off-site disaster recovery node list. Assuming that there are "DisasterRecoveryNode001", "DisasterRecoveryNode002" and "DisasterRecoveryNode003" in the preset off-site disaster recovery node list. Through real-time monitoring of the network connection status of these nodes, it is detected that "DisasterRecoveryNode001" has the smallest network delay with the central cloud node, and its storage capacity is sufficient to store the data of "Device001". Therefore, "DisasterRecoveryNode001" is selected as the off-site disaster recovery node of "Device001", and its identifier is "DisasterRecoveryNode001".

[0106] Finally, determine the backup data verification strategy. In order to ensure the integrity and accuracy of the backup data, a suitable backup data verification strategy needs to be adopted. The hash verification method can be used, such as MD5 or SHA-256 hash algorithm. In each backup data, the original data is hashed to get a hash value. After transmitting the data to the off-site disaster recovery node, the same hash calculation is performed on the backup data on the disaster recovery node to get a new hash value. Compare the two hash values, if the same, it means that the backup data is complete and accurate; if different, it means that the data may have errors in the transmission or storage process, and needs to be re-backed up. At the same time, in order to improve the accuracy and efficiency of the verification, the block verification method can be used, which divides the data into several small blocks, and calculates and compares the hash value of each small block. The data backup period parameter is 3 minutes, the off-site disaster recovery node identifier is "Disaster Recovery Node001", and the backup data verification strategy (SHA-256 hash algorithm for block verification) is combined together, and the resource redundancy backup instruction of the center cloud node for "Device001" is obtained.

[0107] Step S143: Based on the resource allocation topology and the resource redundancy backup instruction, a resource scheduling instruction set is constructed, which includes edge node resource allocation instruction, center cloud backup trigger instruction and global load balancing coordination instruction.

[0108] Based on the resource allocation topology of the edge computing node for "Device001" and the resource redundancy backup instruction of the center cloud node for "Device001" generated before, a resource scheduling instruction set is constructed.

[0109] First, the edge node resource allocation instruction is generated. According to the computing resource allocation proportion of the edge node (“EdgeNode001” allocates 3 CPU cores, 15 GB of memory, and the corresponding GPU computing power proportion, “EdgeNode002” allocates 2 CPU cores, 8 GB of memory, and the corresponding GPU computing power proportion, and “EdgeNode003” allocates 1 CPU core, 9 GB of memory, and the corresponding GPU computing power proportion), and the cross-node load balancing path and resource reservation buffer capacity, the edge node resource allocation instruction is constructed. The instruction contains the identifier of each edge computing node, the specific resource amount allocated to “Device001” and the resource reservation information, and the rules and conditions of load balancing. For example, the instruction content can be: “EdgeNode001: CPU core number allocation 3, memory capacity allocation 15 GB, GPU computing power allocation proportion [specific proportion value], resource reservation buffer: CPU core 1, memory 4 GB; when the resource utilization rate reaches 80%, part of the task is transferred to EdgeNode003; EdgeNode002: CPU core number allocation 2, memory capacity allocation 8 GB, GPU computing power allocation proportion [specific proportion value], resource reservation buffer [specific reservation amount]; load balancing rule [specific rule]; EdgeNode003: CPU core number allocation 1, memory capacity allocation 9 GB, GPU computing power allocation proportion [specific proportion value], resource reservation buffer [specific reservation amount]; load balancing rule [specific rule]”.

[0110] Then the center cloud backup trigger instruction is generated. According to the data backup period parameter 3 minutes in the resource redundancy backup instruction, the off-site disaster recovery node identifier “DisasterRecoveryNode001” and the backup data verification strategy (using SHA-256 hash algorithm for block verification), the center cloud backup trigger instruction is constructed. The instruction content is: “Backup the data of Device001 every 3 minutes to the off-site disaster recovery node DisasterRecoveryNode001, and use SHA-256 hash algorithm for block verification”.

[0111] Finally, the global load balancing coordination instruction is generated. Considering that there can be multiple access devices in the power wireless local area network, global load balancing coordination of the entire network resources is needed. According to the virtualized resource demand prediction results of each access device and the resource usage of the edge computing nodes, the global load balancing coordination instruction is formulated. For example, when the resource utilization of a certain edge computing node is too high, the tasks of some access devices are migrated to other edge computing nodes with lower resource utilization; or according to the real-time requirements and resource demand priorities of different access devices, the resource allocation is dynamically adjusted. The instruction content can be: "monitor the resource utilization of each edge computing node in real time, when the CPU utilization of a certain edge computing node exceeds 80% or the memory utilization exceeds 70%, start the load migration mechanism, and migrate the tasks of some low-priority access devices to other edge computing nodes; according to the real-time guarantee level and resource request priority of the access devices, the resource demand of high-priority devices is prioritized."

[0112] The edge node resource allocation instruction, the center cloud backup triggering instruction and the global load balancing coordination instruction are combined together to obtain a resource scheduling instruction set for "Device001".

[0113] Step S144: encapsulate the resource scheduling instruction set as the virtualized resource scheduling strategy, and store it in association with the global resource state mapping table.

[0114] The resource scheduling instruction set generated for "Device001" is encapsulated as a virtualized resource scheduling strategy. The resource scheduling instruction set can be encapsulated in formats such as JSON or XML. Taking the JSON format as an example, the edge node resource allocation instruction, the central cloud backup trigger instruction, and the global load balancing coordination instruction are converted into a JSON object. This JSON object is used as the virtualized resource scheduling strategy for "Device001" and is associated and stored in the global resource state mapping table. Similarly, through the database connection, an SQL update statement is constructed to update the record corresponding to "Device001" in the global resource state mapping table. Assuming that a new field "resource_scheduling_strategy" is added to the global resource state mapping table to store the virtualized resource scheduling strategy, the constructed SQL update statement is: "UPDATE device_resource_status SET resource_scheduling_strategy='{"edge_node_allocation_instructions":"EdgeNode001:CPU core number allocation 3, memory capacity allocation 15GB, GPU computing power allocation proportion [specific proportion value], resource reservation buffer: CPU core 1, memory 4GB; when the resource utilization rate reaches 80%, part of the task is transferred to EdgeNode003; EdgeNode002: CPU core number allocation 2, memory capacity allocation 8GB, GPU computing power allocation proportion [specific proportion value], resource reservation buffer: [specific reservation amount]; load balancing rule [specific rule]; EdgeNode003: CPU core number allocation 1, memory capacity allocation 9GB, GPU computing power allocation proportion [specific proportion value], resource reservation buffer: [specific reservation amount]; load balancing rule [specific rule]", "cloud_backup_trigger_instructions":"Backup data of Device001 every 3 minutes to DisasterRecoveryNode001 in a remote disaster recovery node, and use SHA-256 hash algorithm for block verification", "global_load_balancing_instructions":"Real-time monitoring of resource utilization of each edge computing node, when the CPU utilization of a certain edge computing node exceeds 80% or the memory utilization exceeds 70%, the load migration mechanism is started, part of the low priority access device task is migrated to other edge computing nodes; according to the real-time level of the access device and the resource request priority, the resource demand of high priority device is preferentially met"}'WHERE device_id='Device001'".

[0115] After constructing the above SQL update statement, it is sent to the database for execution. Through the previously established database connection object, the method for executing SQL statements is called, and the update statement is passed in as a parameter. After the database management system receives the statement, it will be parsed in detail. First, check if the syntax is correct, including the use of keywords, matching of quotes, legality of field names and table names, etc. If the syntax check passes, the database management system locates the record in the "device_resource_status" table with "device_id" as "Device001".

[0116] After finding the corresponding record, the database management system will assign the virtualized resource scheduling strategy JSON string specified in the update statement to the "resource_scheduling_strategy" field. During the assignment process, the database will check the type and length of the data to ensure that the JSON string meets the definition of the field. If the field is defined as a string type with a length limit, and the length of the JSON string exceeds this limit, the database will return an error message indicating that the data length is out of range.

[0117] After the update operation is executed, the database management system returns an execution result. If the update is successful, it will return a status code indicating the success of the operation and the number of records updated. For example, the returned status code may be 200 (indicating successful operation in some database systems), and the number of records updated is 1, indicating that the "Device001" corresponding record has been successfully updated, and its virtualized resource scheduling strategy has been associated and stored in the global resource state mapping table.

[0118] If an error occurs during execution, the database management system will return detailed error information. These error messages may include SQL syntax errors, table does not exist, field does not exist, data type mismatch, insufficient permissions, etc. Different processing measures need to be taken for different error types. For example, if there is a SQL syntax error, the update statement needs to be checked and the syntax error needs to be corrected; if the table does not exist or the field does not exist, the database table structure needs to be confirmed, and the corresponding table or field may need to be created; if the data type is not matched, the data format or field definition needs to be adjusted; if the permission is insufficient, the user's permission settings need to be checked to ensure that the user has the permission to update the table.

[0119] Step S150: Based on the virtualized resource scheduling strategy, dynamically adjust the virtualized computing resources of the power wireless local area network, trigger resource reallocation operation and update the global resource state mapping table.

[0120] After the virtualization resource scheduling strategy is generated and stored, the virtualization computing resources of the power wireless local area network need to be dynamically adjusted according to the strategy to meet the resource requirements of the access devices, improve the utilization rate of resources and the stability of the system. The specific operation process is as follows:

[0121] Step S151: send the edge node resource allocation instruction to the target edge computing node, trigger the target edge computing node to perform dynamic partitioning processing on the local resource pool according to the computing resource allocation ratio, and generate updated local resource pool configuration parameters.

[0122] Taking the edge node resource allocation instruction of "Device001" as an example, it specifies that "EdgeNode001", "EdgeNode002" and "EdgeNode003" are the three target edge computing nodes for allocating computing resources. The central scheduling node sends the edge node resource allocation instruction to the three target edge computing nodes through network connection. Then the target edge computing node will carry out a series of operations according to the instruction:

[0123] Step S1511: analyze the computing resource allocation ratio in the edge node resource allocation instruction to determine the CPU core allocation number, memory allocation capacity and GPU computing power allocation ratio of the target edge computing node.

[0124] Taking "EdgeNode001" as an example, when receiving the edge node resource allocation instruction, the computing resource allocation ratio for "Device001" is analyzed. The instruction specifies that a certain number of CPU cores, memory capacity and corresponding GPU computing power ratio are allocated to "Device001". Through accurate analysis of the instruction content, "EdgeNode001" determines to allocate 3 CPU cores, 15GB memory and a specific GPU computing power ratio (assuming that the ratio is 20% through complex resource evaluation and strategy calculation) to "Device001".

[0125] Step S1512: perform real-time resource occupancy detection processing on the local resource pool of the target edge computing node to generate current resource occupancy state parameters, which include allocated resource amount, idle resource amount and resource fragmentation degree.

[0126] The "EdgeNode001" uses system monitoring tools provided by the operating system or a dedicated resource management module to perform real-time resource occupancy detection on the local resource pool. For CPU cores, it detects that out of a total of 16 CPU cores, 10 have been allocated to other tasks, so there are 6 idle CPU cores. At the same time, through analysis of the CPU core allocation, it detects that there is a certain degree of resource fragmentation, for example, some CPU cores are allocated to multiple small tasks in a scattered manner, resulting in a certain impact on overall resource utilization efficiency. The degree of resource fragmentation is evaluated to be 20% (obtained by a fragmentation evaluation algorithm of the prior art, which comprehensively considers factors such as CPU core allocation granularity and task relevance). For memory, out of a total of 64GB of memory, 32GB has been allocated, and 32GB is free. Similarly, by analyzing the memory allocation, the degree of memory resource fragmentation is determined to be 15% (evaluated based on factors such as memory block allocation size and continuity). For GPU computing power, it is detected through monitoring that 30% of the computing power has been used, and the remaining 70% of the computing power is idle. The degree of GPU resource fragmentation is relatively low, at 5% (judged according to GPU computing task allocation and video memory occupancy). By integrating these allocated resource amounts, free resource amounts, and resource fragmentation degree information, the current resource occupancy state parameters of "EdgeNode001" are generated.

[0127] Step S1513: Perform resource reservation space calculation processing according to the computing resource allocation ratio and the current resource occupancy state parameters to generate a resource reservation buffer capacity and a resource recovery trigger threshold.

[0128] Based on the computing resources allocated to "Device001" (3 CPU cores, 15GB of memory, and 20% of GPU computing power) and the current resource occupancy state parameters of "EdgeNode001", the resource reservation space is calculated. For CPU cores, considering the possibility of burst tasks or resource demand fluctuations, in addition to the 3 CPU cores allocated to "Device001", an additional CPU core is reserved as a resource reservation buffer to deal with emergency situations. For memory, in addition to the 15GB allocated to "Device001", 4GB of memory is reserved as a buffer. For GPU computing power, based on the allocation of 20%, 5% of the computing power is reserved. In this way, the resource reservation buffer capacity is determined to be 1 CPU core, 4GB of memory, and 5% of GPU computing power. At the same time, in order to ensure effective utilization of resources and stable operation of the system, a resource recovery trigger threshold is set. When the CPU core utilization rate exceeds 90%, the memory utilization rate exceeds 85%, or the GPU computing power utilization rate exceeds 95%, the resource recovery operation is triggered to release part of the resources for use by other tasks.

[0129] Step S1514: Based on the resource reservation buffer capacity, the local resource pool of the target edge computing node is dynamically partitioned to generate updated local resource pool configuration parameters, which include newly allocated resource block addresses, resource reservation zone identifiers, and resource recycling strategies.

[0130] "EdgeNode001" dynamically partitions the local resource pool according to the determined resource reservation buffer capacity. Three CPU cores and 15GB of memory are divided into a dedicated resource block for "Device001", and a unique address is assigned to this resource block, such as "ResourceBlock001". At the same time, the resource reservation zone is marked, and its identifier is set as "ReservedBlock001". This reserved zone contains 1 CPU core, 4GB of memory, and 5% of GPU computing power. A resource recycling strategy is developed, which recycles and reallocates part of the resources when the resource usage of "Device001" is below a certain percentage (such as CPU usage below 30%, memory usage below 20%, GPU computing power usage below 10%) or the system is in resource shortage (reaches the resource recycling trigger threshold). The newly allocated resource block addresses, resource reservation zone identifiers, and resource recycling strategies are integrated to generate updated local resource pool configuration parameters.

[0131] Step S1515: Return the updated local resource pool configuration parameters to the central scheduling node and trigger the synchronous update operation of the global resource state mapping table.

[0132] "EdgeNode001" sends the updated local resource pool configuration parameters back to the central scheduling node through the network. After receiving these parameters, the central scheduling node immediately triggers the synchronous update operation of the global resource state mapping table. The central scheduling node will construct an SQL update statement to accurately associate the updated local resource pool configuration parameters of "EdgeNode001" to the corresponding records of "Device001" and "EdgeNode001" in the global resource state mapping table. Similarly, "EdgeNode002" and "EdgeNode003" also operate according to the above steps S1511-S1515, generate their own updated local resource pool configuration parameters, and return them to the central scheduling node for updating the global resource state mapping table.

[0133] Step S152: Send the central cloud backup trigger instruction to the central cloud node, trigger the central cloud node to perform redundant backup processing on the target data according to the backup data verification strategy, and generate a backup completion confirmation signal and a disaster recovery node state identifier.

[0134] The central scheduling node sends a central cloud backup trigger instruction for "Device001" to the central cloud node. After receiving the instruction, the central cloud node will perform redundant backup processing on the target data according to the following detailed steps:

[0135] Step S1521: Analyze the backup data verification strategy in the central cloud backup trigger instruction to determine the verification algorithm type, block backup size, and disaster recovery node selection priority of the target data.

[0136] After receiving the instruction, the central cloud node analyzes the backup data verification strategy therein. Through analysis of the instruction content, it is determined that the target data is verified using the SHA-256 hash algorithm, and the block backup size is set to 1 GB (this size is determined by considering factors such as data transmission efficiency, storage device performance, and network bandwidth). At the same time, the disaster recovery node selection priority is clear, and in this case, the offsite disaster recovery node "DisasterRecoveryNode001" is specified as the preferred choice for data backup.

[0137] Step S1522: According to the disaster recovery node selection priority, filter the target disaster recovery node from the preset disaster recovery node list, and send a backup preparation ready instruction to the target disaster recovery node.

[0138] According to the determined disaster recovery node selection priority, the central cloud node filters the target disaster recovery node "DisasterRecoveryNode001" from the preset disaster recovery node list. Then, the central cloud node sends a backup preparation ready instruction to "DisasterRecoveryNode001" through the network, informing it that the data backup operation of "Device001" is about to be performed, and requiring it to prepare to receive data.

[0139] Step S1523: Receive the storage space allocation confirmation signal returned by the target disaster recovery node, and perform block processing on the target data based on the block backup size to generate multiple data blocks and corresponding block check codes.

[0140] After receiving the backup preparation ready instruction, "DisasterRecoveryNode001" checks its storage space and status. If there is sufficient storage space and the status is normal, it can return a storage space allocation confirmation signal to the central cloud node. After receiving the signal, the central cloud node performs block processing on the target data of "Device001" according to the 1 GB block backup size. Assuming that the total amount of target data of "Device001" is 15 GB, it will be divided into 15 data blocks. For each data block, a simple summation verification algorithm is used to generate a corresponding block check code for preliminary checking of data integrity during transmission.

[0141] Step S1524: Hash calculation is performed on each data block using the check algorithm type, a block hash value is generated and compared with the block check code for verification, and a data integrity verification result is generated.

[0142] The central cloud node performs hash calculation on each data block using the SHA-256 hash algorithm. Each data block is taken as input, and a corresponding block hash value is generated through the algorithm. Then, the block hash value is compared with the previously generated block check code for verification. If they are consistent, it is considered that the data block is complete before transmission; if they are not consistent, it is considered that the data block may have errors, and the data block is marked as a possibly damaged data block, and a corresponding data integrity verification result is generated.

[0143] Step S1525: When the data integrity verification result is passed, the plurality of data blocks are transmitted in parallel to the target disaster recovery node, and a backup completion confirmation signal and a disaster recovery node state identifier returned by the target disaster recovery node are received.

[0144] When the data integrity verification results of all data blocks are passed, the central cloud node transmits the plurality of data blocks in parallel to the target disaster recovery node "DisasterRecoveryNode001" through a high-speed network. In the transmission process, a reliable transmission protocol is used to ensure accurate transmission of data. After receiving the data blocks, "DisasterRecoveryNode001" performs verification again, including recalculating the block hash value and the block check code, and comparing them with those sent by the central cloud node. If the verification is passed, "DisasterRecoveryNode001" returns a backup completion confirmation signal and a disaster recovery node state identifier to the central cloud node. The disaster recovery node state identifier can represent the current state of the disaster recovery node, such as "normal", "busy", "faulty", etc. It is assumed that the returned disaster recovery node state identifier is "normal".

[0145] Step S153: Receive the updated local resource pool configuration parameters and the backup completion confirmation signal, and perform real-time update processing on the global resource state mapping table of the power wireless local area network to generate an updated global resource state mapping table.

[0146] The central scheduling node receives the updated local resource pool configuration parameters from the target edge computing node and the backup completion confirmation signal and the disaster recovery node state identifier from the central cloud node.

[0147] For the updated local resource pool configuration parameters, the central scheduling node associates them to the corresponding records in the global resource state mapping table. Taking "EdgeNode001" as an example, its newly allocated resource block address "ResourceBlock001", resource reservation area identifier "ReservedBlock001", and resource recycling strategy information are accurately updated to the records corresponding to "Device001" and "EdgeNode001". By constructing SQL update statements, these information is written to the corresponding fields of the global resource state mapping table, ensuring data consistency and accuracy.

[0148] For the backup completion confirmation signal and the disaster recovery node state identifier, the central scheduling node updates them to the backup-related records of "Device001" in the global resource state mapping table. If the backup completion confirmation signal indicates that the backup is successful, and the disaster recovery node state identifier is "normal", the backup state is updated to "completed", and the state information of the disaster recovery node is recorded. Similarly, through SQL update statements, these backup-related information is accurately updated to the corresponding fields of the global resource state mapping table.

[0149] When updating the global resource state mapping table, the central scheduling node will again check the consistency and integrity of the data. Check whether the newly allocated resource quantity is consistent with the previous resource allocation instruction, whether the backup information is accurately recorded, etc. For example, verify whether the CPU core number, memory capacity, and other resource information allocated for "Device001" are consistent with the edge node resource allocation instruction, whether the backup completion time, disaster recovery node state, and other backup information are complete and accurate. If inconsistent data or errors are detected, appropriate error handling will be performed, such as reacquiring data, checking network connection, or communicating with related nodes for confirmation.

[0150] After the data consistency and integrity check is correct, the central scheduling node executes the series of SQL update statements constructed before, accurately writes the updated local resource pool configuration parameters, backup completion confirmation signal, and disaster recovery node state identifier into the corresponding fields. After completing these operations, the updated global resource state mapping table is generated, which reflects the latest resource state and backup situation of each node in the current power wireless local area network.

[0151] Step S154: According to the global load balancing coordination instruction, cross-node load migration processing is performed on the access devices with incomplete resource allocation, and load migration log and resource allocation exception alarm event are generated.

[0152] The central scheduling node obtains the list of access devices with incomplete resource allocation from the global resource state mapping table. Then, according to the following steps, cross-node load migration processing is performed on these devices:

[0153] Step S1541: Obtain a list of access devices with unfulfilled resource allocation from the global resource status mapping table, the list containing device identifiers, types of unfulfilled resource requirements, and remaining waiting time lengths.

[0154] The central scheduling node extracts a list of access devices with unfulfilled resource allocation from the global resource status mapping table, which contains device identifiers, types of unfulfilled resource requirements, and remaining waiting time lengths, etc. Assume that there is "Device002" in the list, whose type of unfulfilled resource requirement is insufficient CPU cores, and the remaining waiting time length is 5 minutes.

[0155] Step S1542: Filter a set of available resource nodes from other edge computing nodes according to the type of unfulfilled resource requirement, the set containing node identifiers, available resource amounts, and network transmission delay parameters.

[0156] Filter a set of available resource nodes from other edge computing nodes according to the type of unfulfilled resource requirement. For the CPU core requirement of "Device002", the central scheduling node checks the resource usage of each edge computing node. Assume that "EdgeNode004" has 4 idle CPU cores and "EdgeNode005" has 2 idle CPU cores, these two nodes are selected as the set of available resource nodes, which contains node identifiers ("EdgeNode004" and "EdgeNode005"), available resource amounts (4 and 2 CPU cores), and network transmission delay parameters (assume that the network transmission delay between "EdgeNode004" and "Device002" is 12 milliseconds, and the network transmission delay between "EdgeNode005" and "Device002" is 15 milliseconds).

[0157] Step S1543: Perform migration path optimization processing based on the network transmission delay parameters and the available resource amounts, and generate an optimal migration path and an estimated migration time consumption.

[0158] Perform migration path optimization processing based on the network transmission delay parameters and the available resource amounts. In order to reduce the delay and improve the efficiency in the migration process, preferentially select nodes with small network transmission delay and sufficient available resource amounts to meet the requirements. In this example, "EdgeNode004" has smaller network transmission delay and more available resource amounts, so it is selected as the target available resource node. Calculate the optimal migration path, which is assumed to be the optimal path to migrate directly from the current node to "EdgeNode004", and estimate the migration time consumption. The estimated migration time consumption can be estimated according to factors such as data size, network bandwidth, and transmission delay, and is assumed to be 2 minutes.

[0159] Step S1544: Migrate the load data of the access device with uncompleted resource allocation to the target available resource node according to the optimal migration path, and generate a load migration log containing the migration data volume, actual migration time consumption, and target node resource update status.

[0160] The load data of the access device "Device002" with uncompleted resource allocation is migrated to the target available resource node "EdgeNode004" according to the optimal migration path. During the migration process, a detailed load migration log is recorded. The load migration log contains the migration data volume (assumed to be 5 GB), actual migration time consumption (assumed to be 2.5 minutes), and target node resource update status (1 free CPU core remaining after "EdgeNode004" is updated).

[0161] Step S1545: When the actual migration time consumption exceeds the estimated migration time consumption or data loss occurs during the migration process, generate a resource allocation abnormality alarm event and trigger a resource rollback operation.

[0162] When the actual migration time consumption exceeds the estimated migration time consumption (2.5 minutes is greater than 2 minutes) or data loss occurs during the migration process, a resource allocation abnormality alarm event is generated and a resource rollback operation is triggered. The resource rollback operation restores the load data of "Device002" to the state before migration, and records the abnormal situation for subsequent troubleshooting and processing.

[0163] Step S155: Based on the load migration log and the resource allocation abnormality alarm event, iteratively optimize the virtualized resource scheduling strategy, generate an optimized virtualized resource scheduling strategy, and update the global resource state mapping table.

[0164] For example, the central scheduling node analyzes the actual migration time consumption in the load migration log and the resource allocation abnormality alarm event, and then performs the following iterative optimization processing:

[0165] Step S1551: Analyze the actual migration time consumption in the load migration log and the resource allocation abnormality alarm event, and determine the resource scheduling efficiency bottleneck factor and the abnormal trigger root cause.

[0166] For the load migration of "Device002", the actual migration time consumption of 2.5 minutes exceeds the estimated 2 minutes, and a resource allocation abnormality alarm event is generated. Through in-depth analysis of the load migration log and related network data, the central scheduling node determines that the resource scheduling efficiency bottleneck factor is insufficient network bandwidth, and the abnormal trigger root cause is unstable data transmission caused by network jitter during the migration process.

[0167] Step S1552: According to the resource scheduling efficiency bottleneck factor, the resource allocation topology in the virtualization resource scheduling strategy is optimized in topological structure, and an optimized resource allocation topology is generated, which contains an added edge node identifier, a load balancing path adjustment parameter and a resource reservation buffer expansion ratio.

[0168] According to the resource scheduling efficiency bottleneck factor, the resource allocation topology in the virtualization resource scheduling strategy is optimized in topological structure. In the resource allocation topology, consider adding an edge computing node with larger network bandwidth, or adjusting the load balancing path to avoid passing through nodes with insufficient network bandwidth. Assuming that "EdgeNode006" is detected to have larger network bandwidth and sufficient resources, it is added as a new edge node to the resource allocation topology, and its identifier is "EdgeNode006". At the same time, the load balancing path is adjusted, and when there is an access device that needs to migrate load, it is preferred to pass through "EdgeNode006". In addition, the capacity of the resource reservation buffer is appropriately expanded to cope with possible network fluctuations and sudden increases in resource demand, and the resource reservation buffer expansion ratio is assumed to be set to 20%. An optimized resource allocation topology is generated, containing the added edge node identifier "EdgeNode006", the load balancing path adjustment parameter (prefer to pass through "EdgeNode006"), and the resource reservation buffer expansion ratio 20%.

[0169] Step S1553: Based on the abnormal trigger root cause, the resource redundancy backup instruction is modified in strategy, and an optimized resource redundancy backup instruction is generated, which contains a backup data compression algorithm, a disaster recovery node dynamic selection strategy and a backup process retry mechanism.

[0170] Since the exception is caused by unstable data transmission caused by network jitter, the backup data compression algorithm is optimized, and a more efficient compression algorithm such as LZMA algorithm is selected to reduce data transmission volume. At the same time, the disaster recovery node dynamic selection strategy is adopted, and the network connection status and load of each disaster recovery node are monitored in real time during the backup process, and the disaster recovery node with stable network connection and low load is selected for backup. In addition, the backup process retry mechanism is set, and when an error occurs in the backup process, it is automatically retried for a certain number of times, and the retry number is assumed to be set to 3 times. An optimized resource redundancy backup instruction is generated, containing a backup data compression algorithm (LZMA algorithm), a disaster recovery node dynamic selection strategy and a backup process retry mechanism (retry 3 times).

[0171] Step S1554: The optimized resource allocation topology and the optimized resource redundancy backup instruction are processed in strategy fusion, an optimized virtualization resource scheduling strategy is generated, and the policy version identifier in the global resource state mapping table is updated.

[0172] Integrate the information such as the newly added edge node information, load balancing path adjustment, resource reservation buffer expansion and the like with the backup data compression algorithm, disaster recovery node dynamic selection strategy and backup process retry mechanism and the like to generate the optimized virtualization resource scheduling strategy.

[0173] Through the database connection, an SQL update statement is constructed to update the optimized virtualization resource scheduling strategy to the record corresponding to "Device002" in the global resource state mapping table. Meanwhile, the strategy version identifier in the global resource state mapping table is updated to distinguish the virtualization resource scheduling strategies of different versions, facilitating the subsequent management and tracking. In this way, the iterative optimization processing of the virtualization resource scheduling strategy is completed, and the global resource state mapping table is updated.

[0174] Figure 2 A schematic diagram of exemplary hardware and software components of the power wireless local area network based virtualization computing resource scheduling system 100 according to some embodiments of the present application is shown. For example, the processor 120 can be used in the power wireless local area network based virtualization computing resource scheduling system 100 and used to execute the functions in the present application.

[0175] The power wireless local area network based virtualization computing resource scheduling system 100 can be a general-purpose server or a special-purpose server, both of which can be used to implement the power wireless local area network based virtualization computing resource scheduling method of the present application. Although only one server is shown in the present application, for the sake of convenience, the functions described in the present application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0176] For example, the power wireless local area network based virtualization computing resource scheduling system 100 can include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and different forms of storage media 140, such as a disk, a ROM, or a RAM, or any combination thereof. Exemplarily, the power wireless local area network based virtualization computing resource scheduling system 100 can also include program instructions stored in a ROM, a RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present application can be implemented according to these program instructions. The power wireless local area network based virtualization computing resource scheduling system 100 also includes an input / output (I / O) interface 150 between the computer and other input / output devices.

[0177] For the convenience of description, only one processor is described in the power wireless local area network based virtualization computing resource scheduling system 100. However, it should be noted that the power wireless local area network based virtualization computing resource scheduling system 100 in the present application can also include multiple processors, and thus the steps performed by one processor described in the present application can also be jointly performed by multiple processors or individually performed. For example, if the processor of the power wireless local area network based virtualization computing resource scheduling system 100 performs steps A and B, it should be understood that steps A and B can also be jointly performed by two different processors or individually performed in one processor. For example, a first processor performs step A, a second processor performs step B, or the first processor and the second processor jointly perform steps A and B.

[0178] In addition, the embodiment of the present application further provides a readable storage medium, wherein computer executable instructions are preset, and when a processor executes the computer executable instructions, the power wireless local area network based virtualization computing resource scheduling method is realized.

[0179] It should be noted that, in order to simplify the description of the present application and to help the understanding of one or more embodiments of the present application, in the foregoing description of the embodiments of the present application, various features are sometimes combined into one embodiment, figure or description thereof.

Claims

1. A virtualized computing resource scheduling method based on power wireless local area network, characterized in that, The method includes: Acquire power equipment operating load data of all access devices within the coverage area of ​​the power wireless local area network. The power equipment operating load data includes real-time current fluctuation characteristics, voltage phase offset characteristics, and equipment operating cycle parameters. The load features of the power equipment operating load data are extracted to obtain a load feature set for each connected device. The load feature set includes device power consumption fluctuation features, load response timeliness features, and resource request priority features. Based on a preset load prediction model, the load feature set is dynamically predicted to generate virtualization resource demand prediction results for each access device. The virtualization resource demand prediction results include the computing resource allocation scale and storage resource allocation timeliness requirements. A virtualization resource scheduling strategy is generated based on the virtualization resource demand prediction results. The virtualization resource scheduling strategy includes the resource allocation topology of edge computing nodes and the resource redundancy backup instructions of central cloud nodes. Based on the virtualization resource scheduling strategy, the virtualization computing resources of the power wireless local area network are dynamically adjusted, triggering resource reallocation operations and updating the global resource status mapping table. The dynamic resource demand prediction process based on the preset load prediction model is applied to the load feature set to generate virtualization resource demand prediction results for each access device, including: The pre-trained load prediction model is invoked to perform time series analysis on the load feature set, generating a short-term resource demand prediction sequence and a long-term resource demand trend distribution. The short-term resource demand prediction sequence includes the peak value of computing resource demand and the fluctuation range of storage resource demand within a preset future time period. The long-term resource demand trend distribution includes the resource occupancy growth slope, the influence factor of periodic task superposition, and the probability of sudden load events. The resource allocation magnitude is calculated based on the short-term resource demand forecast sequence and the long-term resource demand trend distribution to generate the computing resource allocation magnitude, wherein the computing resource allocation magnitude includes the CPU core number requirement, memory capacity requirement and GPU computing power allocation ratio. Based on the resource request priority characteristics and the load response time characteristics, storage resource timeliness matching processing is performed to generate the storage resource allocation timeliness requirements. The storage resource allocation timeliness requirements include the upper limit of data read / write latency, the storage bandwidth guarantee threshold, and the redundant backup response time. The computational resource allocation magnitude and the storage resource allocation timeliness requirement are jointly encoded to generate the virtualization resource demand prediction result, and the virtualization resource demand prediction result is written into the global resource status mapping table.

2. The virtualized computing resource scheduling method based on power wireless local area network according to claim 1, characterized in that, The acquisition of power equipment operating load data for all access devices within the coverage area of ​​the power wireless local area network includes: Send a periodic load monitoring instruction to each access device in the power wireless local area network, the periodic load monitoring instruction including a device identifier and a monitoring time window parameter; Receive raw load monitoring data packets returned by each access device within the monitoring time window. The raw load monitoring data packets include current waveform sampling sequences, voltage waveform sampling sequences, and device operation status logs. The current waveform sampling sequence is subjected to fluctuation feature extraction processing to obtain the real-time current fluctuation feature, wherein the real-time current fluctuation feature includes current peak offset, waveform distortion rate and frequency harmonic distribution parameters. The voltage waveform sampling sequence is subjected to phase offset analysis to obtain the voltage phase offset characteristics, wherein the voltage phase offset characteristics include phase synchronization error, voltage drop amplitude and transient response delay parameters; The periodic task trigger records in the device operation status log are parsed to generate the device operation cycle parameters, which include the task execution interval duration, task processing time threshold, and upper limit of resource usage ratio.

3. The virtualized computing resource scheduling method based on power wireless local area network according to claim 2, characterized in that, The process of extracting load features from the power equipment's operating load data to obtain a load feature set for each connected device includes: A device power consumption calculation function is constructed based on the real-time current fluctuation characteristics and the voltage phase offset characteristics. The device power consumption fluctuation characteristics are generated based on the device power consumption calculation function. The device power consumption fluctuation characteristics include the power consumption change rate per unit time, the peak power consumption duration, and the low power consumption state switching frequency. The device operation cycle parameters are processed by response timeliness modeling to generate the load response timeliness characteristics, which include task processing delay tolerance, resource request response timeout threshold and real-time assurance level identifier. Based on the resource request records in the device operation status log, priority weight allocation is performed to generate the resource request priority feature, which includes an emergency task trigger identifier, a critical device identifier, and a service quality level parameter. The device power consumption fluctuation characteristics, the load response time characteristics, and the resource request priority characteristics are normalized and concatenated to obtain the load characteristic set, which is then associated and stored in the global resource status mapping table.

4. The virtualized computing resource scheduling method based on power wireless local area network according to claim 1, characterized in that, The step of generating a virtualization resource scheduling strategy based on the virtualization resource demand prediction results includes: The computational resource allocation magnitude in the virtualization resource demand prediction results is analyzed to generate the resource allocation topology of the edge computing nodes. The resource allocation topology includes the computational resource allocation ratio of the edge nodes, the cross-node load balancing path, and the resource reserved buffer capacity. Based on the storage resource allocation timeliness requirements, a resource redundancy backup instruction for the central cloud node is generated. The resource redundancy backup instruction includes data backup cycle parameters, off-site disaster recovery node identifiers, and backup data verification strategies. Based on the resource allocation topology and the resource redundancy backup instructions, a resource scheduling instruction set is constructed. The resource scheduling instruction set includes edge node resource allocation instructions, central cloud backup trigger instructions, and global load balancing coordination instructions. The set of resource scheduling instructions is encapsulated into the virtualization resource scheduling policy and stored in the global resource status mapping table.

5. The virtualized computing resource scheduling method based on power wireless local area network according to claim 4, characterized in that, The dynamic adjustment of virtualized computing resources in the power wireless local area network based on the virtualized resource scheduling strategy, triggering resource reallocation operations and updating the global resource status mapping table, includes: Send the edge node resource allocation instruction to the target edge computing node, triggering the target edge computing node to dynamically divide the local resource pool according to the computing resource allocation ratio, and generate updated local resource pool configuration parameters; Send the central cloud backup trigger command to the central cloud node, triggering the central cloud node to perform redundant backup processing on the target data according to the backup data verification strategy, and generate a backup completion confirmation signal and a disaster recovery node status identifier. Upon receiving the updated local resource pool configuration parameters and the backup completion confirmation signal, the global resource status mapping table of the power wireless local area network is updated in real time to generate an updated global resource status mapping table. According to the global load balancing coordination instructions, cross-node load migration is performed on access devices that have not completed resource allocation, generating load migration logs and resource allocation anomaly alarm events. Based on the load migration logs and the resource allocation anomaly alarm events, the virtualization resource scheduling strategy is iteratively optimized to generate an optimized virtualization resource scheduling strategy and update the global resource status mapping table.

6. The virtualized computing resource scheduling method based on power wireless local area network according to claim 5, characterized in that, The process of triggering the target edge computing node to dynamically partition its local resource pool according to the computing resource allocation ratio, and generating updated local resource pool configuration parameters, includes: The computing resource allocation ratio in the edge node resource allocation instruction is analyzed to determine the number of CPU cores, memory allocation capacity, and GPU computing power allocation ratio of the target edge computing node. Real-time resource occupancy detection is performed on the local resource pool of the target edge computing node to generate current resource occupancy status parameters, which include the amount of allocated resources, the amount of idle resources, and the degree of resource fragmentation. Based on the calculated resource allocation ratio and the current resource occupancy status parameters, the resource reservation space is calculated to generate the resource reservation buffer capacity and the resource reclamation trigger threshold. Based on the resource reservation buffer capacity, the local resource pool of the target edge computing node is dynamically partitioned to generate updated local resource pool configuration parameters. The updated local resource pool configuration parameters include the newly allocated resource block address, resource reservation area identifier, and resource reclamation strategy. The updated local resource pool configuration parameters are returned to the central scheduling node, and the synchronization update operation of the global resource status mapping table is triggered.

7. The virtualized computing resource scheduling method based on power wireless local area network according to claim 5, characterized in that, The process of triggering the central cloud node to perform redundant backup of the target data according to the backup data verification strategy, generating a backup completion confirmation signal and a disaster recovery node status identifier, includes: Analyze the backup data verification strategy in the central cloud backup trigger command to determine the verification algorithm type, block backup size, and disaster recovery node selection priority of the target data; Based on the priority of the disaster recovery node selection, the target disaster recovery node is selected from the preset disaster recovery node list, and a backup preparation ready instruction is sent to the target disaster recovery node. Receive the storage space allocation confirmation signal returned by the target disaster recovery node, divide the target data into blocks based on the block backup size, and generate multiple data blocks and corresponding block check codes; The aforementioned verification algorithm type is used to perform hash calculation on each data block to generate a block hash value, which is then compared and verified with the block check code to generate a data integrity verification result. When the data integrity verification result is passed, the multiple data blocks are transmitted in parallel to the target disaster recovery node, and the backup completion confirmation signal and disaster recovery node status identifier returned by the target disaster recovery node are received.

8. The virtualized computing resource scheduling method based on power wireless local area network according to claim 5, characterized in that, The process of performing cross-node load migration for access devices that have not completed resource allocation, generating load migration logs and resource allocation anomaly alarm events, includes: Obtain a list of access devices that have not completed resource allocation from the global resource status mapping table. The list of access devices includes device identifiers, types of unmet resource requirements, and remaining waiting time. Based on the type of unmet resource requirement, a set of available resource nodes is selected from other edge computing nodes. The set of available resource nodes includes node identifiers, available resource quantities, and network transmission delay parameters. Based on the network transmission delay parameters and the available resources, the migration path is optimized to generate the optimal migration path and the estimated migration time. According to the optimal migration path, the load data of the access devices that have not completed resource allocation are migrated to the target available resource node, and a load migration log is generated. The load migration log includes the amount of migrated data, the actual migration time, and the resource update status of the target node. When the actual migration time exceeds the estimated migration time or data loss occurs during the migration process, a resource allocation anomaly alarm event is generated and a resource rollback operation is triggered.

9. A virtualized computing resource scheduling system based on a power wireless local area network, characterized in that, The device includes a processor and a memory, the memory being connected to the processor. The memory is used to store programs, instructions, or code, and the processor is used to execute the programs, instructions, or code in the memory to implement the virtualized computing resource scheduling method based on a power wireless local area network as described in any one of claims 1-8.

Images