A parameter configuration optimization method and system for flexible equipment operation module

Abstract

The present invention relates to the field of information technology, and discloses a parameter configuration optimization method and system for flexible device operation modules. The operation modules in the flexible device are hierarchically divided through a hierarchical management method to form a multi-level operation module structure; historical operation-related data of the operation modules at each level are collected to determine their performance evaluation thresholds, operation status thresholds and configuration parameter sensitivity; by real-time monitoring of the operation status and performance evaluation data of the operation modules at each level, a configuration parameter table is generated when the threshold is exceeded, and the correlation between the operation status and performance is analyzed, the importance of the configuration parameters in the configuration parameter table is extracted and calculated, and a configuration parameter set is generated; the operation modules at each level are grouped according to sensitivity and a mapping relationship between the configuration parameter adjustment amount and the performance is constructed, and a progressive adjustment rule is generated to optimize the configuration parameter set, so as to achieve fine adjustment of the configuration parameters of the operation modules in the flexible device, so as to improve the operation efficiency and stability.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a parameter configuration optimization method and system for a flexible device operation module. Background Art

[0002] With the continuous innovation of flexible electronic technology in materials, processes and design, flexible devices have been widely used in many industries. In view of the significant differences in performance requirements and resource allocation requirements of various operating modules in flexible devices, existing technologies usually use heuristic algorithms or priority scheduling methods to optimize the adjustment parameters of each operating module for resource allocation. However, existing methods do not consider the sensitivity of each operating module to parameter configuration changes, resulting in limited optimization effects.

[0003] It can be seen that how to achieve accurate adjustment and optimization of the configuration parameters of each operating module in a flexible device under limited computing resources and communication bandwidth has become a technical problem that technical personnel in this field need to solve urgently. Summary of the invention

[0004] The present invention provides a parameter configuration optimization method and system for a flexible device operating module, which solves the problem of how to accurately adjust and optimize the configuration parameters of each operating module in the flexible device under limited computing resources and communication bandwidth.

[0005] In order to solve the above technical problems, the first aspect of the present invention provides a parameter configuration optimization method for a flexible device operation module, comprising:

[0006] By using the hierarchical management method, each operating module in the flexible equipment is hierarchically divided to obtain a multi-level operating module, and the historical operating status data of each level of operating module and its corresponding historical performance evaluation data and historical configuration parameter change data are collected to build a historical database;

[0007] Determine the performance evaluation threshold and the operation state threshold of each of the hierarchical operation modules according to the historical database, and analyze the influence of the configuration parameter change in each of the hierarchical operation modules on the operation performance of the flexible device based on the historical database to obtain the configuration parameter sensitivity of each of the hierarchical operation modules;

[0008] Acquire real-time operation status data and real-time performance evaluation data of each of the hierarchical operation modules, and generate a configuration parameter table of each of the hierarchical operation modules when the corresponding operation status threshold or the corresponding performance evaluation threshold is exceeded;

[0009] Analyzing the correlation between the operating status and performance of each of the hierarchical operating modules based on the historical database, and extracting configuration parameters from the configuration parameter table to perform importance calculation according to the correlation, so as to generate a configuration parameter set for each of the hierarchical operating modules;

[0010] The sensitivity of each of the hierarchical operation modules is grouped according to the sensitivity of each of the configuration parameters to obtain a multi-level sensitivity operation module, so as to construct a mapping relationship between the configuration parameter adjustment amount and the performance of each level of sensitivity operation module through a regression tree, generate a progressive adjustment rule to optimize the configuration parameter set of each level of operation module, and control the parameter configuration of each operation module in the flexible device based on the optimization result.

[0011] As one of the preferred solutions, the hierarchical management method is used to hierarchically divide the operation modules in the flexible device to obtain multi-level operation modules, including:

[0012] Obtaining the number of configuration parameters, computational load values, and logic block numbers of each operating module in the flexible device to construct a module complexity quantification matrix of each operating module;

[0013] A directed acyclic graph is established based on the complexity quantization matrix of each module, and a clustering operation is performed through the resource occupancy peak value and the average occupancy value of each running module to perform an initial grouping of each running module to obtain a resource occupancy grouping result;

[0014] The operation delay data corresponding to each of the operation modules is obtained, and the operation delay data is compared based on a preset delay threshold, and the resource occupancy grouping result is adjusted according to the comparison result to obtain a multi-level operation module.

[0015] As one preferred solution, the performance evaluation threshold includes a stability requirement reference value, a resource threshold, a bandwidth threshold and an energy consumption threshold; wherein,

[0016] The step of determining the performance evaluation threshold and the operation status threshold of each level operation module according to the historical database includes:

[0017] Extracting the number of failures, the number of crashes, and the abnormal recovery time of each level of operation module from the historical database and using them as sample data to train the recurrent neural network, so as to output the stability requirement benchmark value of each level of operation module through the trained recurrent neural network;

[0018] Extracting the processor occupancy rate, memory usage and disk read / write queue depth of each level operation module under various load scenarios from the historical database, and performing smoothing processing through a hyperbolic tangent function to obtain a resource threshold value of each level operation module;

[0019] Extracting the number of uplink data packets, the number of downlink data packets, the average size of data packets and the retransmission rate of each level operation module from the historical database, and calculating the corresponding bandwidth occupancy rate by sliding average to obtain the bandwidth threshold of each level operation module;

[0020] The current and voltage sampling values ​​of each of the hierarchical operation modules under various operation states are extracted from the historical database, and the relationship between the load and power consumption under the corresponding operation state is fitted using linear regression to determine the energy consumption threshold of each of the hierarchical operation modules.

[0021] As one of the preferred solutions, the determining of the performance evaluation threshold and the operation status threshold of each level operation module according to the historical database further includes:

[0022] The performance evaluation thresholds of the operating modules at each level are weighted by a fuzzy logic algorithm to obtain the priority order of the operating modules at each level;

[0023] Extracting the historical operation status data of each level operation module from the historical database and preprocessing it to obtain standardized historical operation status data to construct an operation data curve;

[0024] The operation data curves are weightedly calculated according to the priority order of the operation modules at each level to determine the operation status threshold of the operation modules at each level.

[0025] As one of the preferred solutions, the extracting the historical operation status data of each level operation module from the historical database and preprocessing it to obtain the standardized historical operation status data includes:

[0026] Extract the historical operation status data of each level operation module from the historical database, perform weighted median filtering using a sliding window, locate abnormal data by the residual sum of squares, and perform secondary sampling on the historical operation status data whose residual value exceeds a preset standard deviation to obtain filtered status data;

[0027] Obtaining a baseline range of operating status indicators to apply minimum and maximum normalization transformation to the filtered status data to obtain normalized status data of each of the hierarchical operating modules;

[0028] The first-order derivative of the normalized state data of each of the hierarchical operation modules is calculated to obtain the inflection point of each of the normalized state data, and cubic spline interpolation is performed on the area where the inflection point of the change has a preset data gap to obtain the standardized historical operation state data of each of the hierarchical operation modules.

[0029] As one of the preferred solutions, the analysis of the influence of the configuration parameter changes in each of the hierarchical operation modules on the operation performance of the flexible device based on the historical database to obtain the configuration parameter sensitivity of each of the hierarchical operation modules includes:

[0030] Extracting historical performance evaluation data and historical configuration parameter change data of each of the hierarchical operation modules from the historical database to train the long short-term memory network as training samples, and obtaining a relationship model that characterizes the correlation between configuration parameter changes and operation performance in each of the hierarchical operation modules;

[0031] Based on the relationship model, the performance change values ​​corresponding to the configuration parameter change values ​​of each level operation module within a preset time period are counted, and the Pearson correlation calculation is performed on the performance change values ​​according to the preset time window to obtain the parameter correlation influence index of each level operation module;

[0032] The preset weight values ​​of the hierarchical operation modules are used to perform weighted operations on the corresponding parameter correlation impact indicators, and the weighted operation results are grouped according to the preset sensitivity thresholds to obtain a parameter sensitivity quantization table to determine the configuration parameter sensitivity of the hierarchical operation modules.

[0033] As one of the preferred solutions, when the corresponding operating state threshold or the corresponding performance evaluation threshold is exceeded, a configuration parameter table of each level operating module is generated, including:

[0034] When the real-time operation status data of each of the hierarchical operation modules exceeds the corresponding operation status threshold, or the real-time performance evaluation data of each of the hierarchical operation modules exceeds the corresponding performance evaluation threshold, the over-limit degree of each of the hierarchical operation modules is calculated by using a regression decision tree to divide the over-limit interval of each of the hierarchical operation modules;

[0035] Based on the obtained over-limit interval division result, the parameter adjustment frequency, parameter adjustment granularity and parameter adjustment complexity of each level operation module are determined to generate a configuration parameter table of each level operation module.

[0036] As one of the preferred solutions, the correlation between the operating status and performance of each of the hierarchical operating modules is analyzed based on the historical database, and configuration parameters are extracted from the configuration parameter table according to the correlation to calculate the importance, so as to generate a configuration parameter set for each of the hierarchical operating modules, including:

[0037] Extracting the historical operation status data of each of the hierarchical operation modules and its corresponding historical performance evaluation data from the historical database, and calculating the similarity by Euclidean distance, so as to cluster the loads of each of the hierarchical operation modules according to the similarity calculation results to obtain clustered load intervals;

[0038] Calculate the average value, standard deviation, peak value, valley value and volatility of the corresponding load in the cluster load interval to construct a feature vector matrix, and use nonlinear regression method to quantify the correlation between the operating state and the performance index to obtain a multidimensional correlation matrix of each hierarchical operating module;

[0039] Based on the multidimensional association matrix, configuration parameters are extracted from the configuration parameter table of each hierarchical operation module, and the influence relationship between the extracted configuration parameters and the operation status is quantified through a gradient boosting tree, so as to quantify the importance of the extracted configuration parameters based on the influence relationship, and the configuration parameters whose importance calculation results exceed the importance threshold are selected to form the configuration parameter set of each hierarchical operation module.

[0040] As one of the preferred solutions, the sensitivity of each hierarchical operation module is grouped according to the sensitivity of each configuration parameter to obtain a multi-level sensitivity operation module, a mapping relationship between the resource adjustment amount and the performance change amount of each hierarchical sensitivity operation module is constructed through a regression tree, a progressive adjustment rule is generated to optimize the configuration parameter set of each hierarchical operation module, and each operation module in the flexible device is controlled to perform parameter configuration based on the optimization result, which includes:

[0041] Acquire the performance indicator time series data of each operating module in the flexible device after parameter configuration, and use the performance indicator time series data as input features to train a generative adversarial network, so as to perform real-time mapping of virtual data through the trained generative adversarial network to obtain virtual mapping data;

[0042] Based on the virtual mapping data, the upper limit of parameter fluctuation of each level operation module is obtained as a reference threshold, and a virtual operation environment including three sets of virtual configuration parameters of processor scheduling, memory allocation and task queue is constructed by incremental compression;

[0043] Taking the minimum weighted sum of the performance indicator time series data as the optimization goal, the virtual configuration parameters of each level operation module in the virtual operation environment are optimized by using Bayesian optimization to obtain a virtual parameter combination of each level operation module;

[0044] The operation modules of each level are simulated based on each virtual parameter combination, and when the performance indicator data obtained from the operation converges to a preset convergence threshold, the operation modules in the flexible device are controlled to perform parameter configuration according to the virtual parameter combination of each level operation module.

[0045] A second aspect of the present invention provides a parameter configuration optimization system for a flexible device operation module, comprising:

[0046] A database construction module is used to hierarchically divide each operating module in the flexible device through a hierarchical management method to obtain a multi-level operating module, and collect historical operating status data of each level of operating module and its corresponding historical performance evaluation data and historical configuration parameter change data to construct a historical database;

[0047] A sensitivity analysis module, used to determine the performance evaluation threshold and the operation state threshold of each of the hierarchical operation modules according to the historical database, and analyze the influence of the configuration parameter changes in each of the hierarchical operation modules on the operation performance of the flexible device based on the historical database, so as to obtain the configuration parameter sensitivity of each of the hierarchical operation modules;

[0048] A parameter table generating module, used to obtain the real-time operation status data and real-time performance evaluation data of each of the hierarchical operation modules, and generate a configuration parameter table of each of the hierarchical operation modules when the corresponding operation status threshold or the corresponding performance evaluation threshold is exceeded;

[0049] A parameter set building module, used for analyzing the correlation between the operating status and performance of each of the hierarchical operating modules based on the historical database, and extracting configuration parameters from the configuration parameter table according to the correlation to calculate the importance, so as to generate a configuration parameter set for each of the hierarchical operating modules;

[0050] A parameter optimization module is used to group the sensitivity of each level of operation modules according to the sensitivity of each configuration parameter to obtain a multi-level sensitivity operation module, so as to construct a mapping relationship between the configuration parameter adjustment amount and the performance of each level of sensitivity operation module through a regression tree, generate a progressive adjustment rule to optimize the configuration parameter set of each level of operation module, and control the parameter configuration of each operation module in the flexible device based on the optimization result.

[0051] Compared with the prior art, the embodiments of the present invention have the following advantages:

[0052] (1) Through the hierarchical management method, it is possible to carry out refined management of operating modules at different levels to ensure that each module can receive appropriate attention and optimization; using the rich data in the historical database, it is possible to more accurately determine the performance evaluation threshold, operating status threshold and configuration parameter sensitivity, providing strong support for decision-making;

[0053] (2) By real-time monitoring of operating status and performance evaluation data, potential performance problems can be discovered and responded to in a timely manner to ensure the stable operation of flexible equipment. By building mapping relationships and generating progressive adjustment rules, configuration parameters can be gradually optimized to avoid the risks that may be caused by large-scale one-time adjustments.

[0054] (3) The system can fine-tune the configuration parameters of the operating modules in the flexible device, improve the operating efficiency and stability of the flexible device, and reduce energy consumption, extend the life of the device, and enhance the user experience by accurately optimizing the configuration parameters of the operating modules. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solution of the present invention, the drawings required for use in the implementation mode will be briefly introduced below. Obviously, the drawings described below are only some implementation modes of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0056] Figure 1 It is a flow chart of a parameter configuration optimization method for a flexible device operation module provided by a certain embodiment of the present invention;

[0057] Figure 2 It is a structural diagram of a parameter configuration optimization system for a flexible device operation module provided by a certain embodiment of the present invention. DETAILED DESCRIPTION

[0058] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings and embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] In the description of this application, the terms "first", "second", "third", etc. are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first", "second", "third", etc. may explicitly or implicitly include one or more of the feature. In the description of this application, unless otherwise specified, "plurality" means two or more.

[0060] In the description of the present application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two components. The terms "vertical", "horizontal", "left", "right", "upper", "lower" and similar expressions used herein are for illustrative purposes only, and do not indicate or imply that the system or component referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0061] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as those commonly understood by those skilled in the art. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood by specific circumstances.

[0062] In one embodiment, if Figure 1 As shown, the first aspect of the present invention provides a parameter configuration optimization method for a flexible device operation module, comprising:

[0063] S1. hierarchically divide each operating module in the flexible device through a hierarchical management method to obtain a multi-level operating module, and collect historical operating status data of each level of operating module and its corresponding historical performance evaluation data and historical configuration parameter change data to build a historical database;

[0064] Specifically, the flexible device in the present invention can be understood as a flexible electronic device or a flexible intelligent device, which is a production equipment with the ability to quickly adapt to production changes. It can quickly adjust production capacity, product models and production methods, and is widely used in many fields, such as flexible stone production and processing machinery and equipment, soft porcelain production line soft tile soft stone machinery and equipment, flexible feeding automatic feeding equipment, flexible visual feeding system equipment, flexible circuit board double-sided cleaning and dust removal equipment, flexible material vacuum coating machine, etc., which can all be called flexible equipment. Each flexible device contains several parallel operation modules, and the type of module depends on the specific function of the flexible device, such as sensor module, control system module, actuator module, human-computer interaction module, etc. The present invention takes a flexible display device including a display drive module, a touch control module and an ambient light perception module as an example to develop the scheme; wherein the display drive module has functions such as display parameter configuration, refresh rate control, and display cache management, the touch control module has functions such as touch data collection, coordinate analysis, and gesture recognition, and the ambient light perception module has functions such as light intensity data collection and data smoothing. It should be noted that the functions of the operating modules contained in the exemplary flexible device are simply described here and below as examples, and the flexible device of the present invention and the operating modules contained therein are not limited to these contents.

[0065] In one embodiment, the hierarchical management method is used to hierarchically divide the operation modules in the flexible device to obtain multi-level operation modules, including:

[0066] Obtaining the number of configuration parameters, computational load values, and logic block numbers of each operating module in the flexible device to construct a module complexity quantification matrix of each operating module;

[0067] A directed acyclic graph is established based on the complexity quantization matrix of each module, and a clustering operation is performed through the resource occupancy peak value and the average occupancy value of each running module to perform an initial grouping of each running module to obtain a resource occupancy grouping result;

[0068] The operation delay data corresponding to each of the operation modules is obtained, and the operation delay data is compared based on a preset delay threshold, and the resource occupancy grouping result is adjusted according to the comparison result to obtain a multi-level operation module.

[0069] Specifically, the present invention is divided based on the functional complexity, configuration parameter quantity, and performance requirement characteristics of each operating module in the flexible device, and a module complexity quantization matrix of each operating module is constructed by the number of configuration parameters, computing load value, and number of logic blocks of each operating module in the flexible device; wherein, the functional complexity of the module is quantified and marked by the code magnitude and the logic block counter, the performance requirement characteristics can be represented by the performance index benchmark data obtained from the module parameter reading table, and the number of configuration parameters and computing load value are obtained by statistically analyzing the parameters inside each module; for example, the number of logic blocks of the display driver module is 156, the number of configuration parameters is 243, the computing load value is 89, and the performance benchmark index is 95; the number of logic blocks of the touch control module is 92, the number of configuration parameters is 157, the computing load value is 76, and the performance benchmark index is 82; the number of logic blocks of the ambient light sensing module is 37, the number of configuration parameters is 45, the computing load value is 31, and the performance benchmark index is 41;

[0070] Then, the modules are sorted according to the values ​​of the above indicators, and the data dependencies between the modules are traversed in a directed acyclic graph, and the peak and average resource occupancy values ​​during the operation of the modules are obtained. For example, the peak CPU occupancy of the display driver module reaches 78%, the average occupancy is 45%, the peak memory occupancy is 384MB, and the average occupancy is 156MB. The peak CPU occupancy of the touch control module is 52%, the average occupancy is 28%, the peak memory occupancy is 196MB, and the average occupancy is 87MB. The peak CPU occupancy of the ambient light perception module is only 15%, the average occupancy is 8%, the peak memory occupancy is 43MB, and the average occupancy is 26MB. Then, based on these resource occupancy data, the modules are grouped using K-means clustering to obtain high resource occupancy group, medium resource occupancy group and low resource occupancy group, that is, the resource occupancy grouping results;

[0071] Finally, the operation delay data of each operation module is obtained from the real-time data monitoring interface, and the delay sensitivity value is calculated based on the data flow map, which is compared with the preset layered delay threshold. The resource occupancy grouping result is adjusted according to the comparison result, and a multi-level operation module structure including a core layer operation module, a middle layer operation module and an edge layer operation module is obtained. The preset layered delay threshold is determined according to the actual situation. For example, the core layer delay threshold is set to 10ms, the middle layer delay threshold is set to 100ms, and the edge layer delay threshold is set to 500ms. The display driver module has a direct impact on the display frame rate, the operating delay fluctuation range is between 5-8ms, and the delay sensitivity value is 0.92. The touch control module affects the user interaction response speed, the operating delay fluctuation range is between 45-75ms, the delay sensitivity value is 0.68, and the ambient light perception module operating delay fluctuation range is between 280-450ms, and the delay sensitivity value is 0.23. Then the operating module display driver module belongs to the core layer, the touch control module belongs to the middle layer, and the ambient light perception module belongs to the edge layer. And according to the hierarchical results, the inter-layer data communication rules are established, the core layer and middle layer module data packets are transmitted in duplex mode using the message bus, and the edge layer module data packets are transmitted in simplex mode, and the computing resource quota is allocated according to the module results.

[0072] The present invention can comprehensively and accurately reflect the complexity of each module by collecting the number of configuration parameters, computing load values ​​and the number of logic blocks of each operating module in the flexible device. The module complexity quantification matrix is ​​constructed using the collected data, which can intuitively display the complexity differences between the modules, providing a solid foundation for the subsequent module complexity quantification and helping to achieve refined management and optimization of the modules. A directed acyclic graph is established based on the module complexity quantification matrix, which can clearly display the dependencies and complexity levels between the modules. The initial grouping of the operating modules through clustering operations is helpful to achieve reasonable classification of the modules and resource integration. The management method of multi-level operating modules can make full use of equipment resources and achieve optimal allocation and efficient utilization of resources.

[0073] After hierarchical classification of the operating modules, the present invention collects historical operating status data of the operating modules at each level and their corresponding historical performance evaluation data and historical configuration parameter change data to build a historical database; wherein the operating status, performance evaluation, configuration parameter change and other data of a single operating module in the historical database correspond one-to-one on a time scale, facilitating subsequent analysis and processing.

[0074] S2. Determine the performance evaluation threshold and the operation state threshold of each of the hierarchical operation modules according to the historical database, and analyze the influence of the configuration parameter change in each of the hierarchical operation modules on the operation performance of the flexible device based on the historical database to obtain the configuration parameter sensitivity of each of the hierarchical operation modules;

[0075] In one embodiment, the performance evaluation threshold includes a stability requirement reference value, a resource threshold, a bandwidth threshold and an energy consumption threshold; wherein,

[0076] The step of determining the performance evaluation threshold and the operation status threshold of each level operation module according to the historical database includes:

[0077] Extracting the number of failures, the number of crashes, and the abnormal recovery time of each level of operation module from the historical database and using them as sample data to train the recurrent neural network, so as to output the stability requirement benchmark value of each level of operation module through the trained recurrent neural network;

[0078] Extracting the processor occupancy rate, memory usage and disk read / write queue depth of each level operation module under various load scenarios from the historical database, and performing smoothing processing through a hyperbolic tangent function to obtain a resource threshold value of each level operation module;

[0079] Extracting the number of uplink data packets, the number of downlink data packets, the average size of data packets and the retransmission rate of each level operation module from the historical database, and calculating the corresponding bandwidth occupancy rate by sliding average to obtain the bandwidth threshold of each level operation module;

[0080] The current and voltage sampling values ​​of each of the hierarchical operation modules under various operation states are extracted from the historical database, and the relationship between the load and power consumption under the corresponding operation state is fitted using linear regression to determine the energy consumption threshold of each of the hierarchical operation modules.

[0081] The present invention uses the stability requirements, computing resources, communication bandwidth and energy consumption budget of the running module as indicators for evaluating its performance, specifically:

[0082] The operation data of each level of operation modules is collected from the historical database every 5 minutes. Fault records, crash times, and abnormal recovery time are extracted from the operation logs. A 4-layer network structure is constructed through a recurrent neural network. The input features include the processor occupancy rate, memory usage, and storage read and write rate of each operation module. The module stability score is output to map the correlation between the stability score and resource consumption and generate a stability requirement benchmark value.

[0083] For modules running at each level in four load scenarios (idle, light load, medium load, and heavy load), the processor performance counter is used to record the time slice occupancy rate every 100 milliseconds, the memory page table scanner is used to record the space usage every 200 milliseconds, and the disk monitor is used to record the read and write queue depth every 500 milliseconds. The hyperbolic tangent function is used to smooth the data curve to make the resource consumption curve more continuous. The resource threshold of modules running at each level is determined by combining the curve with actual needs.

[0084] Set a 60-second sampling window to read data packet statistics from the historical database, such as the number of uplink data packets, the number of downlink data packets, the average size of data packets, and the retransmission rate. Use the sliding average to calculate the bandwidth occupancy rate, and then determine the bandwidth threshold of each level of operation module based on the calculation results and actual needs;

[0085] The current and voltage sampling values ​​of each level of operating modules under various operating conditions are extracted from the historical database, such as feature extraction of data with a current less than 100 mA and a voltage fluctuation less than 0.1 V in the idle state, a current less than 300 mA and a voltage fluctuation less than 0.2 V in the normal operating state, and a current less than 500 mA and a voltage fluctuation less than 0.3 V in the peak load state. Linear regression is used to fit the relationship between load and power consumption to determine the energy consumption thresholds of each level of operating modules.

[0086] The present invention uses a variety of data analysis and machine learning technologies to accurately determine the stability requirement benchmark values ​​of operating modules at each level, determine resource thresholds, bandwidth thresholds and energy consumption thresholds, and achieve refined management and intelligent optimization of the stability, resources, bandwidth and energy consumption of flexible devices. It can comprehensively improve the stability, performance and energy efficiency of the system, help ensure the stability and reliability of flexible devices under various loads and operating conditions, and provide a strong guarantee for improving the overall system performance and energy efficiency.

[0087] In one embodiment, the determining the performance evaluation threshold and the operation status threshold of each level operation module according to the historical database further includes:

[0088] The performance evaluation thresholds of the operating modules at each level are weighted by a fuzzy logic algorithm to obtain the priority order of the operating modules at each level;

[0089] Extracting the historical operation status data of each level operation module from the historical database and preprocessing it to obtain standardized historical operation status data to construct an operation data curve;

[0090] The operation data curves are weightedly calculated according to the priority order of the operation modules at each level to determine the operation status threshold of the operation modules at each level.

[0091] Specifically, the present invention applies the rules in the fuzzy logic rule base to weight-map the performance evaluation thresholds of the operating modules at each level; wherein the fuzzy logic rule base includes processor occupancy and memory usage, data throughput, and response delay as input variables, which correspond to the computing resources, communication bandwidth, and stability requirements in the performance indicators, respectively, so as to divide each variable into three levels: low, medium, and high, and combine to form a plurality of specific rules. For example, when the processor occupancy of the display driver module reaches 85%, the rule triggers the weight adjustment, the display driver module weight is increased to 0.6, and the bandwidth occupancy threshold is correspondingly increased to 400MB / s , while the weight of the touch control module is reduced to 0.3, and the ambient light perception module is reduced to 0.1, so that resources are tilted toward the core layer; then the historical operation status data of the operation modules at each level are extracted from the historical database with a period of 60 seconds and preprocessed to obtain the standardized historical operation status data to construct the operation data curve, and the operation data curves at each level are weighted in the order of priority of 0.6, 0.3, and 0.1, and then the operation status thresholds of the operation modules at each level are determined in combination with actual needs. The operation status thresholds include at least the processor occupancy threshold, memory usage threshold, data throughput threshold, and response delay threshold.

[0092] The present invention processes imprecise and ambiguous information through fuzzy logic algorithm, so that the solution can flexibly adapt to the characteristics of different modules; comprehensively consider multiple performance evaluation thresholds to more comprehensively evaluate the performance of modules running at each level; use historical data for weighted calculation to determine the operating status threshold of each module, thereby improving the stability and reliability of the system.

[0093] In one embodiment, extracting the historical operation status data of each of the hierarchical operation modules from the historical database and preprocessing the data to obtain standardized historical operation status data includes:

[0094] Extract the historical operation status data of each level operation module from the historical database, perform weighted median filtering using a sliding window, locate abnormal data by the residual sum of squares, and perform secondary sampling on the historical operation status data whose residual value exceeds a preset standard deviation to obtain filtered status data;

[0095] Obtaining a baseline range of operating status indicators to apply minimum and maximum normalization transformation to the filtered status data to obtain normalized status data of each of the hierarchical operating modules;

[0096] The first-order derivative of the normalized state data of each of the hierarchical operation modules is calculated to obtain the inflection point of each of the normalized state data, and cubic spline interpolation is performed on the area where the inflection point of the change has a preset data gap to obtain the standardized historical operation state data of each of the hierarchical operation modules.

[0097] Specifically, the present invention extracts data of operation modules at each level from the historical database according to the monitoring and collection rules of each layer of the flexible device. The sampling period of the core layer operation module is set to 10 milliseconds, the sampling period of the middle layer operation module is set to 50 milliseconds, and the sampling period of the edge layer operation module is set to 100 milliseconds. At each sampling point, the values ​​of four indicators, namely, processor occupancy, memory usage, data throughput, and response delay, are collected as historical operation status data, and the sliding window size is set to 20 data points, and the sliding step is 5 data points. The noise of these data is smoothed by weighted median filtering, and the abnormal data points are located by the residual sum of squares. The data points whose residual values ​​exceed 2 times the standard deviation are resampled to obtain filtered status data.

[0098] Obtain the benchmark range of the operating status indicators, where the benchmark range of processor occupancy is 0 to 100, the benchmark range of memory usage is 0 to 1024, the benchmark range of data throughput is 0 to 1000, and the benchmark range of response delay is 0 to 1000. Apply the minimum and maximum normalization transformation to the filtered status data, calculate the score according to the weight of indicators at different levels, with the weight of the core layer being 0.5, the weight of the middle layer being 0.3, and the weight of the edge layer being 0.2, and obtain the normalized status data of the operating modules at each level;

[0099] The first-order derivative of the normalized state data curve of the operating modules at each level is calculated, and the absolute value threshold of the derivative is set to 0.05 to identify the inflection point of the data change. The inflection point characteristics of the data curve reflect the state change of the operating module. For the area where the interval between adjacent inflection points exceeds 50 data points, cubic spline interpolation is performed to fill in the points and maintain the continuity of the curve, so as to generate complete and smooth standardized historical operating status data for the operating modules at each level.

[0100] The present invention locates abnormal data by performing sliding window weighted median filtering and residual sum of squares on the historical operating status data of operating modules at each level, which can better retain the edge information of the data, reduce information loss in the data smoothing process, and timely discover and process abnormal situations in the data. The secondary sampling method is simple and effective, and can reduce the impact of abnormal data on the overall data quality. The minimum and maximum standardization conversion is used to eliminate the differences between operating modules at different levels due to different dimensions, thereby improving the comparability and analyzability of the data.

[0101] In one embodiment, the analyzing, based on the historical database, the influence of the configuration parameter changes in each of the hierarchical operation modules on the operation performance of the flexible device to obtain the configuration parameter sensitivity of each of the hierarchical operation modules includes:

[0102] Extracting historical performance evaluation data and historical configuration parameter change data of each of the hierarchical operation modules from the historical database to train the long short-term memory network as training samples, and obtaining a relationship model that characterizes the correlation between configuration parameter changes and operation performance in each of the hierarchical operation modules;

[0103] Based on the relationship model, the performance change values ​​corresponding to the configuration parameter change values ​​of each level operation module within a preset time period are counted, and the Pearson correlation calculation is performed on the performance change values ​​according to the preset time window to obtain the parameter correlation influence index of each level operation module;

[0104] The preset weight values ​​of the hierarchical operation modules are used to perform weighted operations on the corresponding parameter correlation impact indicators, and the weighted operation results are grouped according to the preset sensitivity thresholds to obtain a parameter sensitivity quantization table to determine the configuration parameter sensitivity of the hierarchical operation modules.

[0105] Specifically, the five historical configuration parameter change data of processor time slice, thread priority, task queue depth, memory allocation upper limit and bandwidth limit corresponding to the data are obtained from the historical database, and the historical performance evaluation data of the corresponding module configuration parameters 10 minutes, 1 hour and 24 hours before and after the change is extracted according to the extraction rate of the core layer running module per second, the middle layer running module per minute and the edge layer running module per hour. A 4-layer network structure is constructed using a long short-term memory network. The input features include the values ​​of 5 configuration parameters, and the output features are performance indicators. The training sample size is set to 1000 groups, and the learning rate is set to 0.01. A relationship model that reveals the characteristics of the impact of configuration parameter changes on performance is established.

[0106] Then, based on the relationship model, the performance change values ​​before and after the configuration parameters are changed are counted, and the Pearson correlation coefficient is calculated according to the 30-minute time window. The correlation between the core layer configuration parameters and the processor occupancy rate, the correlation between the middle layer configuration parameters and the response delay, and the correlation between the edge layer configuration parameters and the resource occupancy are quantitatively calculated to generate the correlation impact index of the configuration parameters of the operating modules at each level with a normalized range of 0 to 1;

[0107] The sensitivity of configuration parameters is graded according to the parameter correlation impact index to reflect the difference in importance of parameters of operating modules at different levels. Weighted calculation is performed according to the core layer parameter weight of 0.5, the middle layer parameter weight of 0.3, and the edge layer parameter weight of 0.2. Sensitivity thresholds are set to 0.8, 0.5, and 0.2 accordingly. The parameter sensitivities are grouped, and a sensitivity quantification table of parameters at different levels is established. The sensitivity of configuration parameters of operating modules at each level is determined according to the sensitivity quantification table to guide the priority and adjustment strategy of configuration parameter adjustment.

[0108] The present invention builds a relational model through data-driven and machine learning support, which can more accurately reflect the dynamic relationship in the actual system and provide a solid foundation for subsequent parameter correlation analysis and sensitivity evaluation; through the calculation of parameter correlation influence indicators and weighted operations and sensitivity quantification of data, the sensitivity of different operating modules to configuration parameter adjustment is taken into account, so that the subsequent adjustment effect is better.

[0109] S3. Acquire the real-time operation status data and real-time performance evaluation data of each level operation module, and generate a configuration parameter table of each level operation module when the corresponding operation status threshold or the corresponding performance evaluation threshold is exceeded; wherein, the real-time data of each level operation module is collected using collection rules.

[0110] In one embodiment, when the corresponding operating state threshold or the corresponding performance evaluation threshold is exceeded, a configuration parameter table of each level operating module is generated, including:

[0111] When the real-time operation status data of each of the hierarchical operation modules exceeds the corresponding operation status threshold, or the real-time performance evaluation data of each of the hierarchical operation modules exceeds the corresponding performance evaluation threshold, the over-limit degree of each of the hierarchical operation modules is calculated by using a regression decision tree to divide the over-limit interval of each of the hierarchical operation modules;

[0112] Based on the obtained over-limit interval division result, the parameter adjustment frequency, parameter adjustment granularity and parameter adjustment complexity of each level operation module are determined to generate a configuration parameter table of each level operation module.

[0113] Specifically, when the real-time operation status data of the operation modules at each level exceeds the corresponding operation status threshold, or the real-time performance evaluation data of the operation modules at each level exceeds the corresponding performance evaluation threshold, the regression decision tree is used to calculate the degree of excess score, the weight of the operation status excess score is 0.6, the weight of the performance index excess score is 0.4, and the excess data is divided into three intervals of slight excess, moderate excess, and severe excess according to the score; and the parameter adjustment frequency, parameter adjustment granularity and parameter adjustment complexity of the operation modules at each level are determined according to the result of the excess interval division, such as adjusting the parameters once per second for slight excess in the core layer, adjusting the parameters once every 0.5 seconds for moderate excess, and adjusting the parameters once every 0.1 seconds for severe excess, adjusting the parameters once every minute for slight excess in the middle layer, adjusting the parameters once every 30 seconds for moderate excess, and adjusting the parameters once every 10 seconds for severe excess, adjusting the parameters once every hour for slight excess in the edge layer, and adjusting the parameters once every 30 minutes for moderate excess. , severe overlimit, adjust parameters once every 10 minutes; the adjustment amount for light overlimit in the core layer is 1%, the adjustment amount for medium overlimit is 3%, and the adjustment amount for severe overlimit is 5%; the adjustment amount for light overlimit in the middle layer is 5%, the adjustment amount for medium overlimit is 10%, and the adjustment amount for severe overlimit is 15%; the adjustment amount for light overlimit in the edge layer is 10%, the adjustment amount for medium overlimit is 20%, and the adjustment amount for severe overlimit is 30%; the complexity weight reflects the influence of different resource occupation on module performance, the core layer processor occupation complexity is 0.5, memory occupation complexity is 0.3, and bandwidth occupation complexity is 0.2; the middle layer processor occupation complexity is 0.4, memory occupation complexity is 0.4, and bandwidth occupation complexity is 0.2; the edge layer processor occupation complexity is 0.3, memory occupation complexity is 0.4, and bandwidth occupation complexity is 0.3. The configuration parameters are prioritized according to the complexity weight to obtain the configuration parameter table of modules running at each level.

[0114] The present invention uses a regression decision tree to calculate the degree of over-limit of operating modules at each level, so as to make an accurate assessment based on complex data relationships, and through the over-limit interval division, it can more accurately understand the severity of the over-limit of each module, so as to take more targeted measures; according to the over-limit interval division result, the flexible device can dynamically determine the parameter adjustment frequency, adjustment granularity and adjustment complexity of the operating modules at each level, so as to ensure that the adjustment measures are neither too frequent nor too rough, while maintaining the feasibility of the adjustment; the automated generation of configuration parameter tables improves work efficiency, reduces human intervention, and reduces the risk of configuration errors.

[0115] S4, analyzing the correlation between the operating status and performance of each of the hierarchical operating modules based on the historical database, and extracting configuration parameters from the configuration parameter table according to the correlation to perform importance calculation, so as to generate a configuration parameter set for each of the hierarchical operating modules;

[0116] In one embodiment, step S4 includes:

[0117] Extracting the historical operation status data of each of the hierarchical operation modules and its corresponding historical performance evaluation data from the historical database, and calculating the similarity by Euclidean distance, so as to cluster the loads of each of the hierarchical operation modules according to the similarity calculation results to obtain clustered load intervals;

[0118] Calculate the average value, standard deviation, peak value, valley value and volatility of the corresponding load in the cluster load interval to construct a feature vector matrix, and use nonlinear regression method to quantify the correlation between the operating state and the performance index to obtain a multidimensional correlation matrix of each hierarchical operating module;

[0119] Based on the multidimensional association matrix, configuration parameters are extracted from the configuration parameter table of each hierarchical operation module, and the influence relationship between the extracted configuration parameters and the operation status is quantified through a gradient boosting tree, so as to quantify the importance of the extracted configuration parameters based on the influence relationship, and the configuration parameters whose importance calculation results exceed the importance threshold are selected to form the configuration parameter set of each hierarchical operation module.

[0120] Specifically, the present invention extracts the historical operation status data of each level of operation modules and their corresponding historical performance evaluation data from the historical database, and uses the Euclidean distance meter to measure the similarity between the two to divide the operation modules of each level into three layers of high load, medium load and low load clusters to obtain the cluster load interval;

[0121] The five statistical eigenvalues ​​of mean value, standard deviation, peak value, valley value and volatility are calculated for each layer of data to construct the eigenvector matrix. The nonlinear regression formula is applied to the eigenvector matrix to reveal the correlation characteristics between the operating status and the performance indicators. The operating status characteristics are used as independent variables and the performance indicator characteristics are used as dependent variables. The core layer processor occupancy rate and response delay, the middle layer memory usage rate and throughput, the edge layer bandwidth occupancy rate and execution efficiency are fitted, the correlation coefficient is calculated, and a 15-dimensional correlation matrix is ​​generated.

[0122] Based on the data contained in the association matrix, multiple configuration parameters of the operating modules of each layer, such as the core layer, the middle layer, and the edge layer, are read from the configuration parameter table, and the corresponding relationship between the values ​​of these configuration parameters and the state performance changes is extracted using the gradient boosting tree. The parameter importance analysis reveals the degree of influence of the configuration parameters through the gradient boosting tree, in which the gradient boosting tree sets the number of learning rounds to 500 rounds, the tree depth to 6 layers, and the feature sampling ratio to 0.8, and the importance of the parameters is quantitatively calculated; according to the parameter importance value, the screening threshold is set to 0.6, and the importance of the configuration parameters corresponding to the operating modules at each level is calculated, and the parameters whose importance exceeds the threshold are selected to form a configuration parameter set. It should be noted that the various threshold values ​​involved in the present invention can be determined according to actual needs or industry experience, and no specific numerical restrictions are made here.

[0123] The present invention performs Euclidean distance calculation and cluster load interval division on the historical operating status data of operating modules at each level and their corresponding historical performance evaluation data, which helps to more carefully analyze the operating status and performance indicators of operating modules at each level under different load conditions and improve the accuracy of analysis; through the construction of characteristic vector matrix and quantification of multidimensional correlation matrix, the intrinsic connection between the operating status and performance indicators is revealed, providing an important basis for subsequent configuration parameter extraction and quantitative analysis; through the gradient boosting tree quantification of influence relationship and the quantification and screening of configuration parameter importance, the influence of configuration parameters on the operating status is accurately reflected, and configuration parameters with significant influence on the operating status are screened out, providing key information for subsequent optimization and adjustment.

[0124] S5, grouping the operation modules of each level according to the sensitivity of each configuration parameter to obtain a multi-level sensitivity operation module, constructing a mapping relationship between the configuration parameter adjustment amount and the performance of each level of sensitivity operation module through a regression tree, generating a progressive adjustment rule to optimize the configuration parameter set of each level of operation module, and controlling each operation module in the flexible device to perform parameter configuration based on the optimization result;

[0125] Specifically, the present invention uses the K-means clustering algorithm to group the sensitivity of the operating modules at each level according to the sensitivity of each configuration parameter; wherein, the sensitivity of the high-sensitivity parameter group of the core layer is greater than 0.8, the sensitivity of the medium-sensitivity parameter group of the middle layer is between 0.4 and 0.8, and the sensitivity of the low-sensitivity parameter group of the edge layer is less than 0.4; the performance response time of the sensitivity operating modules at each level is statistically analyzed, and the adjustment cycle parameters and configuration parameter adjustment data of the configuration parameter set within the response time are extracted, and a regression tree is used to construct a mapping relationship between the processor time slice adjustment amount and the response delay change amount, the memory allocation amount and the throughput change amount, and the bandwidth limitation and the resource utilization change amount, and a progressive adjustment rule is generated in which the core layer operating module adjusts 1% per second, the middle layer operating module adjusts 5% per minute, and the edge layer operating module adjusts 10% per hour.

[0126] According to the progressive adjustment rule, dynamic parameter adjustment is achieved through the proportional-integral controller to optimize the configuration parameter set of the operating modules at each level. The core layer proportional coefficient is set to 0.6 and the integral coefficient is set to 0.3, the middle layer proportional coefficient is set to 0.4 and the integral coefficient is set to 0.2, and the edge layer proportional coefficient is set to 0.1. When the performance fluctuation caused by parameter adjustment exceeds the preset threshold, the adjustment step size is automatically reduced to 0.5 times the original step size. After the performance stabilizes, the original step size is restored, and the operating modules in the flexible device are controlled to perform parameter configuration according to the optimized results.

[0127] In one embodiment, after step S5, the following steps are included:

[0128] Acquire the performance indicator time series data of each operating module in the flexible device after parameter configuration, and use the performance indicator time series data as input features to train a generative adversarial network, so as to perform real-time mapping of virtual data through the trained generative adversarial network to obtain virtual mapping data;

[0129] Based on the virtual mapping data, the upper limit of parameter fluctuation of each level operation module is obtained as a reference threshold, and a virtual operation environment including three sets of virtual configuration parameters of processor scheduling, memory allocation and task queue is constructed by incremental compression;

[0130] Taking the minimum weighted sum of the performance indicator time series data as the optimization goal, the virtual configuration parameters of each level operation module in the virtual operation environment are optimized by using Bayesian optimization to obtain a virtual parameter combination of each level operation module;

[0131] The operation modules of each level are simulated based on each virtual parameter combination, and when the performance indicator data obtained from the operation converges to a preset convergence threshold, the operation modules in the flexible device are controlled to perform parameter configuration according to the virtual parameter combination of each level operation module.

[0132] Specifically, the present invention obtains the performance indicator time series data of each operating module in the flexible device after parameter configuration, and uses the generator in the generative adversarial network to adopt an 8-layer convolution structure, the discriminator adopts a 6-layer convolution structure, the input features include three groups of time series data of processor occupancy, memory usage, and response delay, the training samples are set to 10,000 groups per level, the learning rate is set to 0.001, and the real-time mapping of physical data to virtual data is realized through data replay caching to obtain virtual mapping data;

[0133] The virtual mapping data of the operating modules at each level are synchronized with the physical operating status at a sampling frequency of one second, and the upper limit of parameter fluctuation of the operating modules at each level is obtained, such as the upper limit of parameter fluctuation of 20% for the core layer, 50% for the middle layer, and 80% for the edge layer as the benchmark threshold. Incremental data compression transmission is adopted to establish a virtual operating environment including three sets of parameters: processor scheduling, memory allocation, and task queue.

[0134] Extract state data from the virtual operating environment, and use Bayesian optimization to sample parameter configurations. The number of sampling rounds is set to 200, and 30 sets of candidate parameters are generated in each round. The optimization goal is set to minimize the weighted sum of response delay and resource occupancy rate. Optimization weights of 0.6, 0.3, and 0.1 are used for modules at different levels to generate virtual parameter combinations with balanced performance. The Bayesian optimization process shows the convergence characteristics of parameter configuration. In 200 rounds of optimization, 30 sets of candidate parameters are generated in each round. The time slice parameters of the display driver module start from 2 milliseconds and finally converge to 1.8 milliseconds after multiple rounds of search. At this time, the response delay is reduced to 8 milliseconds, and the processor occupancy rate is maintained at 75%. The optimization weight of 0.6 ensures the priority of the core layer performance indicators, while taking into account the optimization goals of the 0.3 weight of the middle layer and the 0.1 weight of the edge layer.

[0135] According to the virtual parameter combination, the virtual module is simulated for 300 seconds to obtain the processor occupancy fluctuation value, response delay change rate, and resource utilization change curve. The parameter adaptive adjustment is achieved through a proportional-integral-based feedback controller. The proportional coefficient is set to 0.4 and the integration time is set to 10 seconds. When the optimization index converges to the set threshold, the virtual parameter combination configuration is synchronously updated to the physical module to control the parameter configuration of each running module in the flexible device. Among them, taking the display driver module as an example, it uses a proportional coefficient of 0.4 for parameter adjustment. When the processor occupancy fluctuates by 15%, the controller suppresses the fluctuation to less than 5% within the 10-second integration time, and the response delay change rate is reduced from the initial 20% to 5%. The resource utilization curve tends to be stable after 10 iterations. When the difference between the optimization index and the target threshold is less than 3%, the 1.8 millisecond time slice parameters verified in the virtual environment are synchronously updated to the physical module, realizing the closed-loop optimization of the virtual and real environments.

[0136] The present invention uses the performance index time series data of each operating module in the flexible device after parameter configuration as input features to train a generative adversarial network (GAN) to perform real-time mapping of virtual data, which can make full use of existing data resources and generate a large amount of virtual data similar to actual data, thereby improving data diversity and analysis accuracy; constructing a virtual operating environment based on virtual mapping data to simulate the operating status of the actual device and provide an experimental platform for subsequent parameter optimization; by obtaining the parameter fluctuation upper limit of each level of operating modules as a reference threshold, it can be ensured that the parameters will not exceed a reasonable range during the optimization process, thereby enhancing the feasibility and safety of the optimization; using Bayesian optimization to optimize the virtual configuration parameters of each level of operating modules in the virtual operating environment, it can efficiently find a parameter combination that minimizes the weighted sum of the performance index time series data, which has higher efficiency and accuracy than traditional optimization methods; through simulation operation, the influence of different parameter combinations on the performance index can be evaluated, providing a basis for subsequent parameter configuration; by optimizing the virtual parameter combination obtained, the parameter configuration of each operating module in the flexible device is controlled, thereby achieving flexibility and customizability of parameter configuration, helping to meet performance requirements in different application scenarios, and improving the overall performance and operation efficiency of the flexible device.

[0137] In the embodiment of the present application, based on the problem of how to achieve accurate adjustment and optimization of the configuration parameters of each operating module in a flexible device under limited computing resources and communication bandwidth, a parameter configuration optimization method for the operating module of a flexible device is designed, which realizes hierarchical division of each operating module in the flexible device through a hierarchical management method to obtain multi-level operating modules, and collects historical operating status data of each level of operating modules and their corresponding historical performance evaluation data and historical configuration parameter change data to build a historical database; determines the performance evaluation threshold and operating status threshold of each level of operating module based on the historical database, and analyzes the impact of configuration parameter changes in each level of operating module on the operating performance of the flexible device based on the historical database to obtain the configuration parameter sensitivity of each level of operating module; obtains real-time operating status data and real-time performance evaluation data of each level of operating module, and when exceeded When the corresponding operating state threshold or the corresponding performance evaluation threshold is reached, a configuration parameter table of each level operating module is generated; the correlation between the operating state and performance of each level operating module is analyzed based on the historical database, and the configuration parameters are extracted from the configuration parameter table according to the correlation to calculate the importance, so as to generate a configuration parameter set of each level operating module; the sensitivity of each level operating module is grouped according to the sensitivity of each configuration parameter to obtain a multi-level sensitivity operating module, so as to construct a mapping relationship between the configuration parameter adjustment amount and the performance of each level sensitivity operating module through a regression tree, generate a progressive adjustment rule to optimize the configuration parameter set of each level operating module, and control the parameter configuration of each operating module in the flexible device based on the optimization result. The technical solution; the refined adjustment of the configuration parameters of the operating module in the flexible device is realized to improve the operating efficiency and stability of the flexible device.

[0138] It should be noted that although the steps in the above flowchart are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders.

[0139] In another embodiment, if Figure 2 As shown, the second aspect of the present invention provides a parameter configuration optimization system for a flexible device operation module, comprising:

[0140] The database construction module 10 is used to hierarchically divide each operation module in the flexible device through a hierarchical management method to obtain a multi-level operation module, and collect the historical operation status data of the operation modules at each level and their corresponding historical performance evaluation data and historical configuration parameter change data to construct a historical database;

[0141] A sensitivity analysis module 20 is used to determine the performance evaluation threshold and the operation state threshold of each of the hierarchical operation modules according to the historical database, and to analyze the influence of the configuration parameter changes in each of the hierarchical operation modules on the operation performance of the flexible device based on the historical database, so as to obtain the configuration parameter sensitivity of each of the hierarchical operation modules;

[0142] The parameter table generating module 30 is used to obtain the real-time operation status data and the real-time performance evaluation data of each of the hierarchical operation modules, and generate the configuration parameter table of each of the hierarchical operation modules when the corresponding operation status threshold value or the corresponding performance evaluation threshold value is exceeded;

[0143] A parameter set building module 40 is used to analyze the correlation between the operating status and performance of each of the hierarchical operating modules based on the historical database, and extract configuration parameters from the configuration parameter table according to the correlation to calculate the importance, so as to generate a configuration parameter set for each of the hierarchical operating modules;

[0144] The parameter optimization module 50 is used to group the sensitivity of each of the hierarchical operation modules according to the sensitivity of each of the configuration parameters to obtain a multi-level sensitivity operation module, so as to construct a mapping relationship between the configuration parameter adjustment amount and the performance of each level of sensitivity operation module through a regression tree, generate a progressive adjustment rule to optimize the configuration parameter set of each level of operation module, and control the parameter configuration of each operation module in the flexible device based on the optimization result.

[0145] It should be noted that each module in the above-mentioned parameter configuration optimization system for a flexible device operation module can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules. For the specific definition of a parameter configuration optimization system for a flexible device operation module, please refer to the definition of a parameter configuration optimization method for a flexible device operation module in the above text. The two have the same functions and effects, which will not be repeated here.

[0146] In summary, the present invention relates to the field of information technology, and discloses a parameter configuration optimization method and system for flexible device operation modules. The operation modules in the flexible device are hierarchically divided through a hierarchical management method to form a multi-level operation module structure; the historical operation-related data of the operation modules at each level are collected to determine their performance evaluation thresholds, operation status thresholds and configuration parameter sensitivity; by real-time monitoring of the operation status and performance evaluation data of the operation modules at each level, a configuration parameter table is generated when the threshold is exceeded, and the correlation between the operation status and performance is analyzed, the importance of the configuration parameters in the configuration parameter table is extracted and calculated, and a configuration parameter set is generated; the operation modules at each level are grouped according to their sensitivity and a mapping relationship between the configuration parameter adjustment amount and the performance is constructed, and a progressive adjustment rule is generated to optimize the configuration parameter set, so as to achieve fine-grained adjustment of the configuration parameters of the operation modules in the flexible device to improve operation efficiency and stability.

[0147] Each embodiment in this specification is described in a progressive manner, and the same or similar parts of each embodiment can be directly referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. It should be noted that the technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above-mentioned embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0148] The above-mentioned embodiments only express several preferred implementation modes of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in the technical field, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be regarded as the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be based on the protection scope of the claims.

Claims

1. A parameter configuration optimization method for a flexible device operation module, characterized in that: include: By using the hierarchical management method, each operating module in the flexible equipment is hierarchically divided to obtain a multi-level operating module, and the historical operating status data of each level of operating module and its corresponding historical performance evaluation data and historical configuration parameter change data are collected to build a historical database; Determine the performance evaluation threshold and the operation state threshold of each of the hierarchical operation modules according to the historical database, and analyze the influence of the configuration parameter change in each of the hierarchical operation modules on the operation performance of the flexible device based on the historical database to obtain the configuration parameter sensitivity of each of the hierarchical operation modules; Acquire real-time operation status data and real-time performance evaluation data of each of the hierarchical operation modules, and generate a configuration parameter table of each of the hierarchical operation modules when the corresponding operation status threshold or the corresponding performance evaluation threshold is exceeded; Analyzing the correlation between the operating status and performance of each of the hierarchical operating modules based on the historical database, and extracting configuration parameters from the configuration parameter table to perform importance calculation according to the correlation, so as to generate a configuration parameter set for each of the hierarchical operating modules; According to the sensitivity of each of the configuration parameters, the sensitivity of each of the hierarchical operation modules is grouped to obtain a multi-level sensitivity operation module, so as to construct a mapping relationship between the configuration parameter adjustment amount and the performance of each level of sensitivity operation module through a regression tree, generate a progressive adjustment rule to optimize the configuration parameter set of each level of operation module, and control each operation module in the flexible device to perform parameter configuration based on the optimization result; The hierarchical management method is used to hierarchically divide the operation modules in the flexible device to obtain multi-level operation modules, including: Obtaining the number of configuration parameters, computational load values, and logic block numbers of each operating module in the flexible device to construct a module complexity quantification matrix of each operating module; A directed acyclic graph is established based on the complexity quantization matrix of each module, and a clustering operation is performed through the resource occupancy peak value and the average occupancy value of each running module to perform an initial grouping of each running module to obtain a resource occupancy grouping result; Obtaining operation delay data corresponding to each of the operation modules, comparing the operation delay data with a preset delay threshold, and adjusting the resource occupancy grouping result according to the comparison result to obtain a multi-level operation module; The performance evaluation thresholds include stability requirement benchmarks, resource thresholds, bandwidth thresholds, and energy consumption thresholds; wherein, The step of determining the performance evaluation threshold and the operation status threshold of each level operation module according to the historical database includes: Extracting the number of failures, the number of crashes, and the abnormal recovery time of each level of operation module from the historical database and using them as sample data to train the recurrent neural network, so as to output the stability requirement benchmark value of each level of operation module through the trained recurrent neural network; Extracting the processor occupancy rate, memory usage and disk read / write queue depth of each level operation module under various load scenarios from the historical database, and performing smoothing processing through a hyperbolic tangent function to obtain a resource threshold value of each level operation module; Extracting the number of uplink data packets, the number of downlink data packets, the average size of data packets and the retransmission rate of each level operation module from the historical database, and calculating the corresponding bandwidth occupancy rate by sliding average to obtain the bandwidth threshold of each level operation module; Extracting the current and voltage sampling values ​​of each of the hierarchical operation modules under various operation states from the historical database, and fitting the relationship between the load and the power consumption under the corresponding operation state by linear regression to determine the energy consumption threshold of each of the hierarchical operation modules; The step of determining the performance evaluation threshold and the operation status threshold of each level operation module according to the historical database further includes: The performance evaluation thresholds of the operating modules at each level are weighted by a fuzzy logic algorithm to obtain the priority order of the operating modules at each level; Extracting the historical operation status data of each level operation module from the historical database and preprocessing it to obtain standardized historical operation status data to construct an operation data curve; Performing weighted calculation on each of the operation data curves according to the priority order of each of the hierarchical operation modules to determine the operation state threshold of each of the hierarchical operation modules; The extracting the historical operation status data of each level operation module from the historical database and preprocessing it to obtain standardized historical operation status data includes: Extract the historical operation status data of each level operation module from the historical database, perform weighted median filtering using a sliding window, locate abnormal data by the residual sum of squares, and perform secondary sampling on the historical operation status data whose residual value exceeds a preset standard deviation to obtain filtered status data; Obtaining a baseline range of operating status indicators to apply minimum and maximum normalization transformation to the filtered status data to obtain normalized status data of each of the hierarchical operating modules; Calculate the first-order derivative of the normalized state data of each of the hierarchical operation modules to obtain the change inflection point of each of the normalized state data, and perform cubic spline interpolation to fill in the area where the change inflection point has a preset data gap, to obtain the standardized historical operation state data of each of the hierarchical operation modules; When the corresponding operating state threshold or the corresponding performance evaluation threshold is exceeded, a configuration parameter table of each level operating module is generated, including: When the real-time operation status data of each of the hierarchical operation modules exceeds the corresponding operation status threshold, or the real-time performance evaluation data of each of the hierarchical operation modules exceeds the corresponding performance evaluation threshold, the over-limit degree of each of the hierarchical operation modules is calculated by using a regression decision tree to divide the over-limit interval of each of the hierarchical operation modules; Based on the obtained over-limit interval division result, the parameter adjustment frequency, parameter adjustment granularity and parameter adjustment complexity of each level operation module are determined to generate a configuration parameter table of each level operation module.

2. A parameter configuration optimization method for a flexible device operation module according to claim 1, characterized in that: The analyzing, based on the historical database, the influence of the configuration parameter changes in each of the hierarchical operation modules on the operation performance of the flexible device to obtain the configuration parameter sensitivity of each of the hierarchical operation modules includes: Extracting historical performance evaluation data and historical configuration parameter change data of each of the hierarchical operation modules from the historical database to train the long short-term memory network as training samples, and obtaining a relationship model that characterizes the correlation between configuration parameter changes and operation performance in each of the hierarchical operation modules; Based on the relationship model, the performance change values ​​corresponding to the configuration parameter change values ​​of each level operation module within a preset time period are counted, and the Pearson correlation calculation is performed on the performance change values ​​according to the preset time window to obtain the parameter correlation influence index of each level operation module; The preset weight values ​​of the hierarchical operation modules are used to perform weighted operations on the corresponding parameter correlation impact indicators, and the weighted operation results are grouped according to the preset sensitivity thresholds to obtain a parameter sensitivity quantization table to determine the configuration parameter sensitivity of the hierarchical operation modules.

3. A parameter configuration optimization method for a flexible device operation module according to claim 1, characterized in that: The analyzing the correlation between the operating status and performance of each of the hierarchical operating modules based on the historical database, and extracting configuration parameters from the configuration parameter table to calculate the importance according to the correlation, so as to generate a configuration parameter set for each of the hierarchical operating modules, includes: Extracting the historical operation status data of each of the hierarchical operation modules and its corresponding historical performance evaluation data from the historical database, and calculating the similarity by Euclidean distance, so as to cluster the loads of each of the hierarchical operation modules according to the similarity calculation results to obtain clustered load intervals; Calculate the average value, standard deviation, peak value, valley value and volatility of the corresponding load in the cluster load interval to construct a feature vector matrix, and use nonlinear regression method to quantify the correlation between the operating state and the performance index to obtain a multidimensional correlation matrix of each hierarchical operating module; Based on the multidimensional association matrix, configuration parameters are extracted from the configuration parameter table of each hierarchical operation module, and the influence relationship between the extracted configuration parameters and the operation status is quantified through a gradient boosting tree, so as to quantify the importance of the extracted configuration parameters based on the influence relationship, and the configuration parameters whose importance calculation results exceed the importance threshold are selected to form the configuration parameter set of each hierarchical operation module.

4. A parameter configuration optimization method for a flexible device operation module according to claim 1, characterized in that: The method comprises: performing sensitivity grouping of each hierarchical operation module according to the sensitivity of each configuration parameter to obtain a multi-level sensitivity operation module, constructing a mapping relationship between the resource adjustment amount and the performance change amount of each hierarchical sensitivity operation module through a regression tree, generating a progressive adjustment rule to optimize the configuration parameter set of each hierarchical operation module, and controlling each operation module in the flexible device to perform parameter configuration based on the optimization result, and then comprising: Acquire the performance indicator time series data of each operating module in the flexible device after parameter configuration, and use the performance indicator time series data as input features to train a generative adversarial network, so as to perform real-time mapping of virtual data through the trained generative adversarial network to obtain virtual mapping data; Based on the virtual mapping data, the upper limit of parameter fluctuation of each level operation module is obtained as a reference threshold, and a virtual operation environment including three sets of virtual configuration parameters of processor scheduling, memory allocation and task queue is constructed by incremental compression; Taking the minimum weighted sum of the performance indicator time series data as the optimization goal, the virtual configuration parameters of each level operation module in the virtual operation environment are optimized by using Bayesian optimization to obtain a virtual parameter combination of each level operation module; The operation modules of each level are simulated based on each virtual parameter combination, and when the performance indicator data obtained from the operation converges to a preset convergence threshold, the operation modules in the flexible device are controlled to perform parameter configuration according to the virtual parameter combination of each level operation module.

5. A parameter configuration optimization system for a flexible device operation module, characterized in that: include: A database construction module is used to hierarchically divide each operating module in the flexible device through a hierarchical management method to obtain a multi-level operating module, and collect historical operating status data of each level of operating module and its corresponding historical performance evaluation data and historical configuration parameter change data to construct a historical database; A sensitivity analysis module, used to determine the performance evaluation threshold and the operation state threshold of each of the hierarchical operation modules according to the historical database, and analyze the influence of the configuration parameter changes in each of the hierarchical operation modules on the operation performance of the flexible device based on the historical database, so as to obtain the configuration parameter sensitivity of each of the hierarchical operation modules; A parameter table generating module, used to obtain the real-time operation status data and real-time performance evaluation data of each of the hierarchical operation modules, and generate a configuration parameter table of each of the hierarchical operation modules when the corresponding operation status threshold or the corresponding performance evaluation threshold is exceeded; A parameter set building module, used for analyzing the correlation between the operating status and performance of each of the hierarchical operating modules based on the historical database, and extracting configuration parameters from the configuration parameter table according to the correlation to calculate the importance, so as to generate a configuration parameter set for each of the hierarchical operating modules; A parameter optimization module, used to group the sensitivity of each of the hierarchical operation modules according to the sensitivity of each of the configuration parameters to obtain a multi-level sensitivity operation module, so as to construct a mapping relationship between the configuration parameter adjustment amount and the performance of each level of sensitivity operation module through a regression tree, generate a progressive adjustment rule to optimize the configuration parameter set of each level of operation module, and control the parameter configuration of each operation module in the flexible device based on the optimization result; The hierarchical management method is used to hierarchically divide the operation modules in the flexible device to obtain multi-level operation modules, including: Obtaining the number of configuration parameters, computational load values, and logic block numbers of each operating module in the flexible device to construct a module complexity quantification matrix of each operating module; A directed acyclic graph is established based on the complexity quantization matrix of each module, and a clustering operation is performed through the resource occupancy peak value and the average occupancy value of each running module to perform an initial grouping of each running module to obtain a resource occupancy grouping result; Obtaining operation delay data corresponding to each of the operation modules, comparing the operation delay data with a preset delay threshold, and adjusting the resource occupancy grouping result according to the comparison result to obtain a multi-level operation module; The performance evaluation thresholds include stability requirement benchmarks, resource thresholds, bandwidth thresholds, and energy consumption thresholds; wherein, The step of determining the performance evaluation threshold and the operation status threshold of each level operation module according to the historical database includes: Extracting the number of failures, the number of crashes, and the abnormal recovery time of each level of operation module from the historical database and using them as sample data to train the recurrent neural network, so as to output the stability requirement benchmark value of each level of operation module through the trained recurrent neural network; Extracting the processor occupancy rate, memory usage and disk read / write queue depth of each level operation module under various load scenarios from the historical database, and performing smoothing processing through a hyperbolic tangent function to obtain a resource threshold value of each level operation module; Extracting the number of uplink data packets, the number of downlink data packets, the average size of data packets and the retransmission rate of each level operation module from the historical database, and calculating the corresponding bandwidth occupancy rate by sliding average to obtain the bandwidth threshold of each level operation module; Extracting the current and voltage sampling values ​​of each of the hierarchical operation modules under various operation states from the historical database, and fitting the relationship between the load and the power consumption under the corresponding operation state by linear regression to determine the energy consumption threshold of each of the hierarchical operation modules; The step of determining the performance evaluation threshold and the operation status threshold of each level operation module according to the historical database further includes: The performance evaluation thresholds of the operating modules at each level are weighted by a fuzzy logic algorithm to obtain the priority order of the operating modules at each level; Extracting the historical operation status data of each level operation module from the historical database and preprocessing it to obtain standardized historical operation status data to construct an operation data curve; Performing weighted calculation on each of the operation data curves according to the priority order of each of the hierarchical operation modules to determine the operation state threshold of each of the hierarchical operation modules; The extracting the historical operation status data of each level operation module from the historical database and preprocessing it to obtain standardized historical operation status data includes: Extract the historical operation status data of each level operation module from the historical database, perform weighted median filtering using a sliding window, locate abnormal data by the residual sum of squares, and perform secondary sampling on the historical operation status data whose residual value exceeds a preset standard deviation to obtain filtered status data; Obtaining a baseline range of operating status indicators to apply minimum and maximum normalization transformation to the filtered status data to obtain normalized status data of each of the hierarchical operating modules; Calculate the first-order derivative of the normalized state data of each of the hierarchical operation modules to obtain the change inflection point of each of the normalized state data, and perform cubic spline interpolation to fill in the area where the change inflection point has a preset data gap, to obtain the standardized historical operation state data of each of the hierarchical operation modules; When the corresponding operating state threshold or the corresponding performance evaluation threshold is exceeded, a configuration parameter table of each level operating module is generated, including: When the real-time operation status data of each of the hierarchical operation modules exceeds the corresponding operation status threshold, or the real-time performance evaluation data of each of the hierarchical operation modules exceeds the corresponding performance evaluation threshold, the over-limit degree of each of the hierarchical operation modules is calculated by using a regression decision tree to divide the over-limit interval of each of the hierarchical operation modules; Based on the obtained over-limit interval division result, the parameter adjustment frequency, parameter adjustment granularity and parameter adjustment complexity of each level operation module are determined to generate a configuration parameter table of each level operation module.

Images