CN117539520A

CN117539520A - Firmware self-adaptive upgrading method, system and equipment

Info

Publication number: CN117539520A
Application number: CN202410035917.9A
Authority: CN
Inventors: 谭志成
Original assignee: Shenzhen East Lair Intelligent Technology Co ltd
Current assignee: Shenzhen East Lair Intelligent Technology Co ltd
Priority date: 2024-01-10
Filing date: 2024-01-10
Publication date: 2024-02-09
Anticipated expiration: 2044-01-10
Also published as: CN117539520B

Abstract

The application relates to the technical field of firmware upgrading and discloses a method, a system and equipment for adaptively upgrading firmware. The method comprises the following steps: monitoring the equipment state of the target firmware to obtain processor load data, memory use data and network connection data; creating a first firmware upgrading knowledge graph and performing knowledge graph expansion updating to obtain a second firmware upgrading knowledge graph; performing performance evaluation index calculation to obtain a plurality of performance evaluation indexes; performing index normalization and vector conversion to generate a performance evaluation input vector; establishing a firmware upgrading strategy decision model and carrying out firmware upgrading strategy decision to obtain a first firmware self-adaptive upgrading strategy; generating a second firmware self-adaptive upgrading strategy according to the first firmware self-adaptive upgrading strategy, carrying out self-adaptive upgrading, and outputting a firmware upgrading log.

Description

Firmware self-adaptive upgrading method, system and equipment

Technical Field

The present disclosure relates to the field of firmware upgrade technologies, and in particular, to a method, a system, and an apparatus for adaptively upgrading firmware.

Background

With the rapid development of the internet of things and embedded systems, firmware self-adaptive upgrade becomes a key means for maintaining device performance, repairing vulnerabilities and introducing new functions. In this context, researchers have focused their efforts on improving the intelligence and adaptability of firmware upgrades to cope with the variability and complexity of the device environment. The existing firmware upgrading method is usually too static and lacks dynamic perception of the real-time state of the equipment, so researchers are urgent to need a more intelligent firmware upgrading method which can monitor the state of the equipment in real time, evaluate the comprehensive performance and formulate a corresponding upgrading strategy.

However, in the current research of adaptive upgrading of firmware, some key problems still exist to be solved. The traditional upgrading method lacks of careful monitoring on the state of the equipment, and cannot effectively sense the dynamic change of the equipment in operation, so that the inaccuracy of upgrading decision is caused. Secondly, the existing method has limitations in performance evaluation and policy formulation, has limited comprehensive consideration degree on a plurality of performance indexes, and is difficult to realize flexibility and universality in different application scenes. In the process of firmware upgrade, the problems of network instability, device compatibility and the like are faced, so that the improvement of the reliability and fault tolerance of the upgrade is one of the challenges to be solved in the current research.

Disclosure of Invention

The application provides a method, a system and equipment for adaptively upgrading firmware, which are used for improving the accuracy of the adaptive upgrading of the firmware.

In a first aspect, the present application provides a firmware adaptive upgrade method, where the firmware adaptive upgrade method includes:

monitoring the equipment state of the target firmware to obtain processor load data, memory use data and network connection data;

creating a first firmware upgrading knowledge graph according to the processor load data, the memory use data and the network connection data, and performing knowledge graph expansion updating on the first firmware upgrading knowledge graph to obtain a second firmware upgrading knowledge graph;

performing performance evaluation index calculation on the target firmware based on the second firmware upgrading knowledge graph to obtain a plurality of performance evaluation indexes;

performing index normalization and vector conversion on the plurality of performance evaluation indexes to generate a performance evaluation input vector;

establishing a firmware upgrading strategy decision model, and carrying out firmware upgrading strategy decision on the performance evaluation input vector through the firmware upgrading strategy decision model to obtain a first firmware self-adaptive upgrading strategy;

Generating a second firmware self-adaptive upgrade strategy corresponding to the target firmware according to the first firmware self-adaptive upgrade strategy, carrying out self-adaptive upgrade on the target firmware through the second firmware self-adaptive upgrade strategy, and outputting a firmware upgrade log.

In a second aspect, the present application provides a firmware adaptive upgrade system, the firmware adaptive upgrade system comprising:

the monitoring module is used for monitoring the equipment state of the target firmware to obtain processor load data, memory use data and network connection data;

the creating module is used for creating a first firmware upgrading knowledge graph according to the processor load data, the memory use data and the network connection data, and carrying out knowledge graph expansion updating on the first firmware upgrading knowledge graph to obtain a second firmware upgrading knowledge graph;

the calculation module is used for calculating the performance evaluation indexes of the target firmware based on the second firmware upgrading knowledge graph to obtain a plurality of performance evaluation indexes;

the conversion module is used for carrying out index normalization and vector conversion on the plurality of performance evaluation indexes to generate a performance evaluation input vector;

The decision module is used for creating a firmware upgrading strategy decision model, and carrying out firmware upgrading strategy decision on the performance evaluation input vector through the firmware upgrading strategy decision model to obtain a first firmware self-adaptive upgrading strategy;

the output module is used for generating a second firmware self-adaptive upgrading strategy corresponding to the target firmware according to the first firmware self-adaptive upgrading strategy, carrying out self-adaptive upgrading on the target firmware through the second firmware self-adaptive upgrading strategy and outputting a firmware upgrading log.

A third aspect of the present application provides a firmware adaptive upgrade apparatus, including: a memory and at least one processor, the memory having instructions stored therein; the at least one processor invokes the instructions in the memory to cause the firmware adaptive upgrade apparatus to perform the firmware adaptive upgrade method described above.

In the technical scheme provided by the application, the method has real-time performance and dynamic performance through equipment state monitoring and knowledge graph expansion and update. The latest state of the target firmware can be obtained at any time, and the upgrade strategy can be formulated more accurately. By utilizing a firmware upgrading strategy decision model, the system can make more intelligent and comprehensive firmware upgrading decisions through comprehensive decisions of a random forest network and a support vector machine network and weighted fusion of an output layer. The device state monitoring data is clustered through the K-means clustering algorithm, and the first firmware self-adaptive upgrading strategy is optimized through the genetic algorithm, so that the complexity of the device state can be better understood and adapted, and the accuracy and the robustness of the algorithm are improved. By creating a multi-level knowledge graph, the system can more fully understand the structure and performance characteristics of the target firmware. This helps to improve the accuracy of firmware upgrade decisions, especially where multiple entities and relationships are considered. Through traversal of the knowledge graph, performance factor identification and sequencing, and mapping of performance evaluation indexes, the system can comprehensively evaluate the performance of the target firmware. This helps to get a finer understanding of the needs and direction of firmware upgrades. The genetic algorithm is utilized to optimize the self-adaptive upgrade strategy of the first firmware, so that the optimal upgrade strategy parameters can be searched, the self-adaptability and the performance of the system are improved, and the accuracy of the self-adaptive upgrade of the firmware is further improved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings may be obtained based on these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a schematic diagram of one embodiment of a firmware adaptive upgrade method according to an embodiment of the present application;

FIG. 2 is a schematic diagram of one embodiment of a firmware adaptive upgrade system according to embodiments of the present application.

Detailed Description

The embodiment of the application provides a firmware self-adaptive upgrading method, a system and equipment. The terms "first," "second," "third," "fourth" and the like in the description and in the claims of this application and in the above-described figures, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments described herein may be implemented in other sequences than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or apparatus.

For ease of understanding, the following describes a specific flow of an embodiment of the present application, referring to fig. 1, and one embodiment of a firmware adaptive upgrade method in an embodiment of the present application includes:

step S101, monitoring equipment state of target firmware to obtain processor load data, memory use data and network connection data;

it may be understood that the execution body of the present application may be a firmware adaptive upgrade system, and may also be a terminal or a server, which is not limited herein. The embodiment of the present application will be described by taking a server as an execution body.

Specifically, the device state monitoring is performed on the target firmware through a preset firmware management platform, and the processor load, the memory use condition and the network connection state of the device are monitored in real time. In this process, the platform will collect detailed device state monitoring data, which is the basis for subsequent cluster analysis and policy formulation. For further analysis and understanding of these monitored data, a preset K-means clustering algorithm is employed, which is a widely used unsupervised learning algorithm that can effectively aggregate a large amount of data into several clusters with similar characteristics. And carrying out cluster center calculation on the equipment state monitoring data through the K-means clustering algorithm, and finding out the inherent modes and structures in the data so as to determine a plurality of first cluster centers. Each cluster center represents a central point of a set of data, reflecting the main features of the set of data. Based on these first cluster centers, corresponding first cluster clusters are generated, and distances from each point in each cluster to its cluster center are calculated, and these distance information help to evaluate the closeness and effectiveness of the clusters. The average value of the point distance of each cluster can be obtained by carrying out average value operation on all the center point distances of each first cluster, and the average value provides a quantization index and reflects the consistency and aggregation degree inside the clusters. In order to improve the accuracy of clustering and reflect the dynamic change of the equipment state, the correction of the clustering center is performed. And adjusting the first clustering centers according to the point distance average value to form a plurality of corrected second clustering centers. This correction process is a key step in pursuing higher cluster quality and data representation, ensuring that the cluster center is better representative of the actual equipment state data. And carrying out secondary clustering on the equipment state monitoring data through a second clustering center, further refining and optimizing a clustering result, and ensuring that the clustering cluster can accurately reflect subtle changes of the equipment state. And respectively extracting corresponding processor load data, memory use data and network connection data according to the obtained second cluster clusters. The data are the results of two rounds of fine cluster analysis, not only reflect the current state of the equipment, but also disclose the potential relation and mode between the states of the equipment, and provide reliable and accurate basis for the subsequent firmware upgrading strategy formulation.

Step S102, creating a first firmware upgrading knowledge graph according to processor load data, memory use data and network connection data, and carrying out knowledge graph expansion updating on the first firmware upgrading knowledge graph to obtain a second firmware upgrading knowledge graph;

specifically, a plurality of processor load entities are determined according to the processor load data, a plurality of memory use entities are determined according to the memory use data, and a plurality of network connection entities are determined according to the network connection data. These entities are the basis for building knowledge maps and represent key device performance parameters in the firmware upgrade process. And carrying out deep relation analysis on the entities through a preset Bayesian algorithm. Bayesian algorithms, with the advantage of their probabilistic model, perform well in processing uncertainty information and finding potential associations. Through the relation analysis of the processor load entity and the memory use entity, a plurality of first entity relations can be obtained; and similarly, analyzing the memory using entity and the network connecting entity to obtain a plurality of second entity relations. These entity relationships reveal interactions and dependencies between different device performance parameters, which facilitate understanding of complex interactions during firmware upgrades. A first hierarchy between the processor load entity and the memory use entity is created from the first entity relationship, and a second hierarchy between the memory use entity and the network connection entity is created from the second entity relationship. The two hierarchical structures provide a framework for the knowledge graph, define the hierarchy and the connection mode between different entities, and enable the knowledge graph to reflect the real relationship between the equipment state and the performance parameters more accurately. And carrying out knowledge graph construction on the first hierarchical structure and the second hierarchical structure to generate a corresponding first firmware upgrading knowledge graph. The map not only comprises each entity and the attribute thereof, but also details the relation among the entities, and provides a set of comprehensive decision support system for the firmware upgrading strategy. And carrying out knowledge spectrum expansion updating on the first firmware upgrading knowledge spectrum. The process involves integration of new data, re-evaluation of entity relationships, and adjustment and optimization of the spectrum structure, resulting in a second firmware upgrade knowledge spectrum.

Step S103, calculating performance evaluation indexes of the target firmware based on the second firmware upgrading knowledge graph to obtain a plurality of performance evaluation indexes;

specifically, the target firmware in the second firmware upgrading knowledge graph is subjected to knowledge graph traversal, all relevant information about the target firmware in the knowledge graph is extracted, wherein the relevant information comprises historical performance data of the firmware, operation parameters of relevant equipment and past upgrading records, and a comprehensive knowledge graph traversal result is obtained. And according to the knowledge graph traversal result, identifying and sequencing the performance factors of the target firmware. Key factors affecting firmware performance are identified, including multiple dimensions such as processor load conditions, memory usage, network connection quality, etc. Having identified these performance factors, they are further ranked to determine which factors have the greatest impact on the performance of the firmware. This ordering is based not only on the degree of direct influence of the factors, but also takes into account interactions and dependencies between factors, ensuring that the actual influence of each factor can be fully and accurately understood. And performing performance evaluation index mapping on the plurality of performance factors, and converting the abstract performance factors into specific and quantifiable performance evaluation indexes. These metrics may be average load of the processor, maximum usage of memory, average of network delay, etc., which are specific metrics for measuring firmware performance, and may be used directly to evaluate the running status of the firmware. The process of performance evaluation index mapping needs to comprehensively consider various factors including importance of performance factors, measurability of indexes, correlation among indexes and the like, so that the obtained performance evaluation indexes are scientific, reasonable, practical and effective.

Step S104, performing index normalization and vector conversion on a plurality of performance evaluation indexes to generate a performance evaluation input vector;

specifically, the performance evaluation indexes are subjected to evaluation index normalization, and indexes of different orders or units are converted to the same scale, so that comparison and further analysis are facilitated. Through normalization processing, the dimension influence among the indexes can be eliminated, so that the comparison among the indexes is more reasonable and scientific. The normalized evaluation indexes are subjected to serialization conversion, and a plurality of normalized evaluation indexes are orderly arranged into a sequence, so that the relative positions and the relations among the indexes can be maintained, and structural input can be provided for subsequent vector conversion. This step considers the correlation and importance between the indices, ensuring that the arrangement of the sequences reflects the most important relationships and features between the indices. And performing vector conversion on the target evaluation index sequence to generate a performance evaluation input vector. Vector conversion is the conversion of serialized evaluation metrics into a mathematical vector form that can be efficiently processed by machine learning models and algorithms. Each serialized evaluation index is converted to a dimension in the vector space, thereby forming a multi-dimensional performance evaluation input vector. This vector integrates all important performance evaluation information and can be directly used as input to the firmware upgrade policy decision model. Vector conversion not only enhances the expression capability of data, but also enables the result of performance evaluation to be effectively processed by complex algorithms and models, thereby improving the intelligent degree and accuracy of firmware upgrading strategies.

Step S105, a firmware upgrading strategy decision model is created, and firmware upgrading strategy decision is carried out on the performance evaluation input vector through the firmware upgrading strategy decision model, so that a first firmware self-adaptive upgrading strategy is obtained;

specifically, a firmware upgrade policy decision model is created, which includes a random forest network and a support vector machine network, and an output layer. Random forests are an integrated learning method that improves prediction accuracy and controls overfitting by building multiple decision trees and aggregating their results. The support vector machine is a powerful classifier, can process high-dimensional data and optimize marginal errors, and is suitable for the conditions of complexity and large data volume. And processing and analyzing the performance evaluation input vector through a random forest network to obtain a first decision result. The random forest network can consider each dimension of the input vector, evaluate the necessity and urgency of firmware upgrade through the constructed decision tree, and output a preliminary decision result. The performance evaluation input vector is input into a support vector machine network, the support vector machine network analyzes data from different angles and tries to find the optimal classification boundary between the data, so that a second decision result is provided. The two networks analyze the same data from different angles, can provide complementary information, and increase the accuracy and reliability of decision making. Then, first network weight data of the random forest network and second network weight data of the support vector machine network are acquired. These weight data reflect the importance and contribution of each network in the firmware upgrade policy decision process. And inputting the weight data into an output layer of the model, and carrying out weighted fusion on the first decision result and the second decision result according to the weight data. Through the weighted fusion mode, not only the advantages of two networks can be integrated, but also the weight distribution can be adjusted according to the actual situation, so that the final decision result is more accurate and has strong adaptability.

Step S106, generating a second firmware self-adaptive upgrading strategy corresponding to the target firmware according to the first firmware self-adaptive upgrading strategy, carrying out self-adaptive upgrading on the target firmware through the second firmware self-adaptive upgrading strategy, and outputting a firmware upgrading log.

Specifically, according to the first firmware self-adaptive upgrade strategy, the target firmware is subjected to upgrade stage analysis, the current running state, the history upgrade record and the potential upgrade requirement of the firmware are understood, the corresponding target upgrade stage is obtained, the specific target and the expected effect of the firmware upgrade are indicated in the stage, and the direction is provided for subsequent strategy optimization. And initializing strategy optimization parameters of the first firmware self-adaptive upgrading strategy through a preset genetic algorithm. The genetic algorithm is a search algorithm imitating a natural evolution mechanism, and iterative optimization is carried out on candidate solutions through operations such as selection, propagation, crossover, mutation and the like. In this process, a plurality of first policy optimization parameters will be generated, each representing an upgrade policy scheme. To evaluate the effectiveness of these schemes, first matching degree data of each first policy optimization parameter is calculated separately. The matching degree data reflects the matching degree of each strategy scheme and the target upgrading stage, and is an important basis for parameter screening and optimization. And according to the first matching degree data, carrying out parameter group division on the plurality of first strategy optimization parameters to obtain a plurality of strategy optimization parameter groups. This partitioning is based on similarity and matching between parameters, aiming to categorize parameters with similar characteristics and performance into the same population. These populations of policy optimization parameters are then subjected to breeding, crossover and mutation operations to simulate natural selection and genetic mutation processes to generate a plurality of second policy optimization parameters. And respectively calculating second matching degree data of each second strategy optimization parameter, wherein the second matching degree data reflect the matching degree of the strategy parameters subjected to the generation evolution and the target upgrading stage. Based on these match data, a non-dominant line ordering analysis is performed on all second policy optimization parameters. The sorting analysis can identify the strategy parameters which are excellent in multiple evaluation dimensions, so that the selected strategy can meet the requirement of upgrading effect, and other factors such as cost and risk can be considered. And carrying out optimization analysis on a plurality of second strategy optimization parameters according to the strategy optimization parameter sequence, and selecting the optimal firmware self-adaptive upgrading strategy from all candidate schemes. The optimization analysis process involves complex computation and comparison, and needs to consider the comprehensive effect and adaptability of each policy parameter, and finally generates a second firmware self-adaptive upgrade policy corresponding to the target firmware. And carrying out self-adaptive upgrading on the target firmware through the generated second firmware self-adaptive upgrading strategy. The actual upgrade operation of the firmware is guided according to the strategy, including downloading new firmware, verifying the integrity of the firmware, installing updates, restarting the device, and the like. In the whole upgrading process, the system can continuously monitor the upgrading state and the equipment performance, record all key events and results and finally output a firmware upgrading log. This log records not only each step of the upgrade process, but also performance comparisons of the firmware before and after the upgrade, problems that occur, solutions, and the like.

In the embodiment of the application, the method has real-time performance and dynamic performance through equipment state monitoring and knowledge graph expansion and update. The latest state of the target firmware can be obtained at any time, and the upgrade strategy can be formulated more accurately. By utilizing a firmware upgrading strategy decision model, the system can make more intelligent and comprehensive firmware upgrading decisions through comprehensive decisions of a random forest network and a support vector machine network and weighted fusion of an output layer. The device state monitoring data is clustered through the K-means clustering algorithm, and the first firmware self-adaptive upgrading strategy is optimized through the genetic algorithm, so that the complexity of the device state can be better understood and adapted, and the accuracy and the robustness of the algorithm are improved. By creating a multi-level knowledge graph, the system can more fully understand the structure and performance characteristics of the target firmware. This helps to improve the accuracy of firmware upgrade decisions, especially where multiple entities and relationships are considered. Through traversal of the knowledge graph, performance factor identification and sequencing, and mapping of performance evaluation indexes, the system can comprehensively evaluate the performance of the target firmware. This helps to get a finer understanding of the needs and direction of firmware upgrades. The genetic algorithm is utilized to optimize the self-adaptive upgrade strategy of the first firmware, so that the optimal upgrade strategy parameters can be searched, the self-adaptability and the performance of the system are improved, and the accuracy of the self-adaptive upgrade of the firmware is further improved.

In a specific embodiment, the process of executing step S101 may specifically include the following steps:

(1) Monitoring the equipment state of the target firmware through a preset firmware management platform to obtain equipment state monitoring data;

(2) Inputting the equipment state monitoring data into a preset K-means clustering algorithm, and performing cluster center calculation on the equipment state monitoring data through the K-means clustering algorithm to obtain a plurality of corresponding first cluster centers;

(3) Generating a plurality of corresponding first cluster clusters according to the plurality of first cluster centers, calculating the distances between the plurality of first cluster clusters and the plurality of first cluster centers to obtain a plurality of center point distances of each first cluster, and carrying out mean value operation on the plurality of center point distances of each first cluster to obtain a mean value of the point distances of each first cluster;

(4) Performing cluster center correction on the first cluster centers through the point distance mean value to obtain a plurality of second cluster centers;

(5) And performing secondary clustering on the equipment state monitoring data through the plurality of second clustering centers to obtain a plurality of second clustering clusters, and generating corresponding processor load data, memory use data and network connection data according to the plurality of second clustering clusters.

Specifically, device state monitoring is performed on the target firmware through a preset firmware management platform, so as to obtain device state monitoring data, wherein the device state monitoring data comprise key performance indexes such as processor load, memory usage, network connection state and the like. Then, the device state monitoring data are input into a preset K-means clustering algorithm. K-means is a widely used clustering algorithm that optimizes the clustering result by iterative computation, divides the data into clusters, and determines the center point of each cluster. In this process, the algorithm needs to determine the number of clusters, i.e. the number of first cluster centers, which can be determined according to the characteristics and distribution of the device state data and the actual application requirements. The algorithm then randomly selects the initial cluster centers and assigns data to the nearest clusters based on the distance between the data points and these center points. And continuously and iteratively updating the positions of the cluster centers until convergence conditions are met, and finally obtaining a plurality of stable first cluster centers and corresponding first clusters. After the first cluster is obtained, calculating the distance from each point in each cluster to the cluster center, and carrying out average value operation on the distances to obtain the average value of the point distance of each cluster. This mean is an important indicator for evaluating the quality of the clusters, reflecting the closeness of the clusters. In general, the smaller the point distance average, the closer the data points within the cluster to the center, and the better the clustering effect. And carrying out cluster center correction on the plurality of first cluster centers through the point distance mean value to generate corrected second cluster centers. This correction process is to further optimize the clustering effect so that the clusters are more closely distributed to the actual data. And performing secondary clustering on the equipment state monitoring data. Through secondary clustering, the data is reassigned to the cluster where the nearest second cluster center is located, forming a plurality of second cluster clusters. Each cluster represents a somewhat homogenous set of device state data from which representative processor load data, memory usage data, and network connection data may be extracted. These extracted data not only reflect the current state of the device, but also implicate potential associations and patterns between the device states. For example, when firmware management is performed on a server in a data center, the management platform first collects data such as CPU utilization, memory occupation, and network throughput of the server in real time. By inputting into the K-means algorithm, the data is divided into clusters, each cluster corresponding to a specific type of server operating state, such as high load, idle or network congestion. These states are more accurately divided and identified by cluster center correction and secondary clustering. The management platform can identify the server which needs to be upgraded preferentially according to the clustering result, or formulate a more accurate firmware upgrading strategy aiming at a specific type of running state.

In a specific embodiment, the process of executing step S102 may specifically include the following steps:

(1) Determining a plurality of corresponding processor load entities according to the processor load data, determining a plurality of corresponding memory use entities according to the memory use data, and determining a plurality of corresponding network connection entities according to the network connection data;

(2) Carrying out relationship analysis on a plurality of processor load entities and a plurality of memory use entities through a preset Bayesian algorithm to obtain a plurality of first entity relationships, and carrying out relationship analysis on a plurality of memory use entities and a plurality of network connection entities through a Bayesian algorithm to obtain a plurality of second entity relationships;

(3) Creating a first hierarchical structure between a plurality of processor load entities and a plurality of memory use entities according to a plurality of first entity relationships, and creating a second hierarchical structure between a plurality of memory use entities and a plurality of network connection entities according to a plurality of second entity relationships;

(4) Carrying out knowledge graph construction on the first hierarchical structure and the second hierarchical structure to generate a corresponding first firmware upgrading knowledge graph;

(5) And carrying out knowledge graph expansion updating on the first firmware upgrading knowledge graph to obtain a second firmware upgrading knowledge graph.

Specifically, a plurality of corresponding processor load entities are determined according to the processor load data, a plurality of corresponding memory use entities are determined according to the memory use data, and a plurality of corresponding network connection entities are determined according to the network connection data. These entities are the basis for knowledge-graph construction and represent key dimensions in device state monitoring. And carrying out relation analysis on the entities through a preset Bayesian algorithm. A bayesian algorithm is a probabilistic model that predicts relationships between entities by computing conditional probabilities between them. And analyzing the relation between the processor load entity and the memory use entity to obtain a plurality of first entity relations. These relationships reveal interdependencies and patterns of impact between processor load and memory usage. And similarly, analyzing the relation between the memory using entity and the network connecting entity through a Bayesian algorithm to obtain a plurality of second entity relations. These relationships help understand how memory usage affects the performance of the network connection. A hierarchy between entities is created from these relationships. Creating a first hierarchy between a processor load entity and a memory use entity based on the first entity relationship; based on the second entity relationship, a second hierarchy is created between the memory use entity and the network connection entity. These hierarchies reflect the degree of association and the type of relationship between different entities. And integrating the hierarchical structures into the construction of the knowledge graph to generate a corresponding first firmware upgrading knowledge graph. The knowledge graph not only comprises each entity and the attribute thereof, but also comprises the relationship and the hierarchical structure among the entities, and provides comprehensive information support for firmware upgrading decision. And carrying out regular expansion updating on the first firmware upgrading knowledge graph so as to ensure that the first firmware upgrading knowledge graph is continuously and effectively. The development and update of the knowledge graph involve introducing new entities, updating entity attributes, adding new relationships or adjusting existing relationships, and the like. This process is based on newly collected device status monitoring data or on further analysis of existing data. Through these updates, a second firmware upgrade knowledge-graph may be obtained. For example, over time and with the collection of more data, new patterns are discovered, such as memory leaks that are more likely to result for a particular type of processor under a particular load, or network connection performance is related to memory access patterns over a particular period of time. These new findings will be used to update and expand the original knowledge-graph, generate a second firmware upgrade knowledge-graph, and provide more accurate upgrade guidance.

In a specific embodiment, the process of executing step S103 may specifically include the following steps:

(1) Performing knowledge graph traversal on the target firmware in the second firmware upgrading knowledge graph to obtain a knowledge graph traversal result;

(2) According to the knowledge graph traversal result, performing performance factor identification and sequencing on the target firmware to obtain a plurality of performance factors;

(3) And performing performance evaluation index mapping on the plurality of performance factors to obtain a plurality of performance evaluation indexes.

Specifically, the second firmware upgrading knowledge graph is traversed, and all relevant information about the target firmware in the graph is comprehensively understood and extracted. Knowledge graph traversal includes not only identifying each entity and relationship in a graph, but also understanding the hierarchical structure and semantic links between these entities and relationships. This process requires the use of a graph query and analysis tool, such as the SPARQL query language or query interface of a graph database, to perform the traversal operation. And carrying out performance factor identification and sequencing on the target firmware according to the result of the knowledge graph traversal. The performance factor is a key variable that affects the performance of the firmware, such as CPU occupancy, memory usage, network latency, etc. All factors affecting the target firmware performance are identified. This requires comprehensive consideration of information such as the operating principle of the firmware, the operating environment of the device, and historical performance data. The performance factors are ranked to determine which factors have the greatest impact on performance. This ordering process may be based on a variety of bases such as statistical importance of the factors, historical impact records, and the like. The result of the ranking will determine which performance factors should be prioritized in subsequent performance evaluation and firmware upgrade decisions. The performance evaluation index is mapped to these factors. Performance evaluation metrics are specific metrics for quantifying the impact of a performance factor, which translate the abstract concept of a performance factor into a numerical value that can be directly measured and evaluated. In this process, each performance factor is mapped to one or more evaluation indicators. For example, processor load entities may be mapped to indicators of CPU occupancy, average load per core, etc.; the memory usage entity maps to the total memory usage, the memory leakage rate, and other indicators. This mapping process requires sufficient consideration of the nature and manner of impact of the performance factors to ensure that the selected evaluation index effectively reflects the performance impact of the factors. After mapping is completed, a detailed set of performance evaluation metrics are obtained, which are used for subsequent performance evaluation and firmware upgrade policy formulation.

In a specific embodiment, the process of executing step S104 may specifically include the following steps:

(1) Performing evaluation index normalization on each performance evaluation index to generate a plurality of normalized evaluation indexes;

(2) Performing serialization conversion on a plurality of normalized evaluation indexes to generate a target evaluation index sequence;

(3) And performing vector conversion on the target evaluation index sequence to generate a performance evaluation input vector.

Specifically, the performance evaluation indexes are subjected to evaluation index normalization, so that the indexes can be compared and analyzed on the same scale. Normalization is typically done in a number of ways, such as max-min normalization (scaling the data to between 0 and 1), or Z-score normalization (conversion to a distribution with a mean of 0 and standard deviation of 1). And carrying out serialization conversion on the normalized evaluation indexes, and arranging the normalized indexes into a sequence according to a certain sequence so as to facilitate subsequent vector conversion and model processing. And after the serialization conversion is completed, carrying out vector conversion on the target evaluation index sequence. The purpose of vector conversion is to convert the serialized evaluation index into a numerically meaningful vector that can be used as an input to a machine learning model or other algorithm. Each serialized evaluation index is converted to a dimension in the vector space, thereby forming a multi-dimensional performance evaluation input vector. The vector integrates all relevant performance evaluation information, and can be directly used for subsequent tasks such as performance evaluation, firmware upgrading strategy formulation and the like. Vector conversion not only enhances the expression capability of data, but also enables the result of performance evaluation to be effectively processed by complex algorithms and models, and improves the accuracy and the intelligent degree of firmware upgrading decision.

In a specific embodiment, the process of executing step S105 may specifically include the following steps:

(1) Creating a firmware upgrade policy decision model, wherein the firmware upgrade policy decision model comprises: a random forest network, a support vector machine network and an output layer;

(2) Carrying out firmware upgrading strategy decision on the performance evaluation input vector through a random forest network to obtain a first decision result;

(3) Carrying out firmware upgrading strategy decision on the performance evaluation input vector through a support vector machine network to obtain a second decision result;

(4) Acquiring first network weight data of a random forest network and second network weight data of a support vector machine network;

(5) And carrying out result weighted fusion on the first decision result and the second decision result according to the first network weight data and the second network weight data through an output layer to obtain a first firmware self-adaptive upgrade strategy.

Specifically, a firmware upgrade policy decision model is created, and a composite structure comprising a random forest network, a support vector machine network and an output layer is built. Random forest networks are an integrated learning technique that improves overall prediction accuracy and stability by building multiple decision trees and synthesizing their prediction results. Each decision tree is trained on a random subset of the dataset, which improves the generalization ability of the model and reduces the risk of overfitting. In this embodiment, a random forest may be used to analyze multiple dimensions of the performance evaluation input vector to determine whether upgrades are needed and the priority and urgency of the upgrades. The support vector machine network is a supervised learning algorithm that finds the best decision boundaries in feature space for classification or regression problems. The support vector machine improves the generalization ability of the model by maximizing the edges of the decision boundaries. In this embodiment, a support vector machine network is used to process the performance evaluation input vector, providing an upgrade policy decision for another perspective. And processing the performance evaluation input vector through a random forest network to obtain a first decision result. This result reflects the upgrade strategy advice that the random forest comprehensively derives from its constructed decision tree. Each decision tree gives different suggestions, and the random forest synthesizes the suggestions by voting or averaging and the like so as to improve the accuracy and stability of the decision. And inputting the performance evaluation input vector into a support vector machine network to obtain a second decision result. The support vector machine network finds the optimal classification boundary based on the training data and then classifies or regresses the new input data based on this boundary, providing advice as to whether upgrades are needed and the way they are upgraded. And combining decision results of the two networks to obtain first network weight data of the random forest network and second network weight data of the support vector machine network. These weight data reflect the importance and contribution of the individual networks in the overall decision model. These weight data are then fed into the output layer of the model. And at the output layer, carrying out weighted fusion on the first decision result and the second decision result according to the first network weight data and the second network weight data. The advantages of the random forest network and the support vector machine network are integrated, and the specific application scene and the specific requirements are adapted by adjusting the weights. The result of the weighted fusion is a first firmware adaptive upgrade strategy that integrates the suggestions of both algorithms and adjusts to the upgrade strategy that is most appropriate for the current firmware state.

In a specific embodiment, the process of executing step S106 may specifically include the following steps:

(1) According to the first firmware self-adaptive upgrading strategy, carrying out upgrading stage analysis on the target firmware to obtain a corresponding target upgrading stage;

(2) Carrying out strategy optimization parameter initialization on the first firmware self-adaptive upgrading strategy according to the target upgrading stage by a preset genetic algorithm to generate a plurality of first strategy optimization parameters;

(3) Respectively calculating first matching degree data of each first strategy optimization parameter, and dividing parameter groups of a plurality of first strategy optimization parameters according to the first matching degree data to obtain a plurality of strategy optimization parameter groups;

(4) Propagating, intersecting and mutating the strategy optimization parameter groups to obtain a plurality of second strategy optimization parameters, and respectively calculating second matching degree data of each second strategy optimization parameter;

(5) According to the second matching degree data, performing non-dominant row ordering analysis on a plurality of second strategy optimization parameters to obtain a strategy optimization parameter sequence;

(6) Performing optimization analysis on a plurality of second strategy optimization parameters according to the strategy optimization parameter sequence to generate a second firmware self-adaptive upgrading strategy corresponding to the target firmware;

(7) And carrying out self-adaptive upgrade on the target firmware through a second firmware self-adaptive upgrade strategy, and outputting a firmware upgrade log.

Specifically, the target firmware is subjected to upgrade stage analysis according to the first firmware self-adaptive upgrade strategy. Understanding the current state of the firmware, the historical upgrade records and the potential upgrade requirements, determining the best timing and manner of firmware upgrade, and the expected upgrade effect. For example, based on the firmware's operational data and performance log, it can be analyzed whether the firmware is currently in a steady-state operational phase or in a performance degradation phase, or whether a problem has arisen that an emergency repair is required. From this information, a target upgrade stage of the firmware, such as performance optimization, functional enhancement, or emergency repair, etc., may be determined. And initializing strategy optimization parameters of the first firmware self-adaptive upgrading strategy through a preset genetic algorithm. The genetic algorithm is an optimization algorithm simulating natural selection and genetic mechanism, and iteratively optimizes candidate solutions through operations such as selection, propagation, crossover, mutation and the like. And initializing the strategy parameters according to the requirements of the target upgrading stage to generate a plurality of first strategy optimization parameters. These parameters represent different upgrade policy schemes, each of which is a path for firmware upgrades. First matching degree data of each first strategy optimization parameter are calculated respectively. The matching degree data reflects the matching degree of each strategy parameter and the target upgrading stage requirement. From these data, a number of first policy optimization parameters may be partitioned into groups of parameters, classifying parameters with similar characteristics and performance into the same group. This division facilitates subsequent breeding and crossover operations, ensuring that quality parameter characteristics can be preserved and inherited. And carrying out propagation, crossover and mutation operation on a plurality of strategy optimization parameter groups. These operations simulate the genetic mechanisms in natural evolution, creating new candidate solutions by combining and mutating superior features. In this process, a plurality of second policy optimization parameters are obtained, each representing an upgrade policy solution after a generation of evolution. And respectively calculating second matching degree data of each second strategy optimization parameter so as to evaluate the matching degree of the evolved schemes and the target upgrading stage requirements. Based on the second matching degree data, non-dominant line ordering analysis is performed on a plurality of second policy optimization parameters. The non-dominant line ordering is a multi-objective optimization method, which can identify solutions that are excellent in multiple evaluation dimensions, so that the selected strategy can meet the requirement of upgrading effects, and can also consider other factors such as cost, risk and the like. Through the sequencing analysis, a strategy optimization parameter sequence is obtained, wherein the strategy optimization parameter sequence comprises a plurality of optimal upgrading strategy schemes after comprehensive evaluation. And carrying out optimization analysis on the plurality of second strategy optimization parameters according to the strategy optimization parameter sequence. The best firmware adaptive upgrade strategy is selected from all the candidate solutions. The optimization analysis process involves complex computation and comparison, and needs to consider the comprehensive effect and adaptability of each policy parameter, and finally generates a second firmware self-adaptive upgrade policy corresponding to the target firmware. This strategy integrates the suggestions of various algorithms and models and adjusts to the upgrade scheme best suited to the current firmware state through an optimization algorithm. In the whole self-adaptive upgrading process, the actual upgrading operation of the firmware is guided according to the second firmware self-adaptive upgrading strategy, and the method comprises the steps of downloading new firmware, verifying the integrity of the firmware, installing and updating, restarting the equipment and the like. Meanwhile, the system can continuously monitor the upgrade state and the equipment performance, record all key events and results, and output a firmware upgrade log after the upgrade is completed. This log records not only each step of the upgrade process, but also performance comparisons of the firmware before and after the upgrade, problems that occur, solutions, and the like.

The method for adaptively upgrading firmware in the embodiment of the present application is described above, and the system for adaptively upgrading firmware in the embodiment of the present application is described below, referring to fig. 2, an embodiment of the system for adaptively upgrading firmware in the embodiment of the present application includes:

the monitoring module 201 is configured to monitor a device state of the target firmware to obtain processor load data, memory usage data, and network connection data;

the creating module 202 is configured to create a first firmware upgrade knowledge graph according to the processor load data, the memory usage data, and the network connection data, and perform knowledge graph expansion update on the first firmware upgrade knowledge graph to obtain a second firmware upgrade knowledge graph;

the computing module 203 is configured to perform performance evaluation index computation on the target firmware based on the second firmware upgrade knowledge graph to obtain a plurality of performance evaluation indexes;

the conversion module 204 is configured to perform index normalization and vector conversion on the multiple performance evaluation indexes, and generate a performance evaluation input vector;

the decision module 205 is configured to create a firmware upgrade policy decision model, and perform a firmware upgrade policy decision on the performance evaluation input vector through the firmware upgrade policy decision model to obtain a first firmware adaptive upgrade policy;

And the output module 206 is configured to generate a second firmware adaptive upgrade policy corresponding to the target firmware according to the first firmware adaptive upgrade policy, adaptively upgrade the target firmware according to the second firmware adaptive upgrade policy, and output a firmware upgrade log.

Through the cooperation of the components, the method has real-time performance and dynamic performance through the equipment state monitoring and the development and the update of the knowledge graph. The latest state of the target firmware can be obtained at any time, and the upgrade strategy can be formulated more accurately. By utilizing a firmware upgrading strategy decision model, the system can make more intelligent and comprehensive firmware upgrading decisions through comprehensive decisions of a random forest network and a support vector machine network and weighted fusion of an output layer. The device state monitoring data is clustered through the K-means clustering algorithm, and the first firmware self-adaptive upgrading strategy is optimized through the genetic algorithm, so that the complexity of the device state can be better understood and adapted, and the accuracy and the robustness of the algorithm are improved. By creating a multi-level knowledge graph, the system can more fully understand the structure and performance characteristics of the target firmware. This helps to improve the accuracy of firmware upgrade decisions, especially where multiple entities and relationships are considered. Through traversal of the knowledge graph, performance factor identification and sequencing, and mapping of performance evaluation indexes, the system can comprehensively evaluate the performance of the target firmware. This helps to get a finer understanding of the needs and direction of firmware upgrades. The genetic algorithm is utilized to optimize the self-adaptive upgrade strategy of the first firmware, so that the optimal upgrade strategy parameters can be searched, the self-adaptability and the performance of the system are improved, and the accuracy of the self-adaptive upgrade of the firmware is further improved.

The application also provides a firmware self-adaptive upgrading device, which comprises a memory and a processor, wherein the memory stores computer readable instructions, and the computer readable instructions, when executed by the processor, cause the processor to execute the steps of the firmware self-adaptive upgrading method in the above embodiments.

It will be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, systems and units may refer to the corresponding processes in the foregoing method embodiments, which are not repeated herein.

The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be embodied in essence or a part contributing to the prior art or all or part of the technical solution in the form of a software product stored in a storage medium, including several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a read-only memory (ROM), a random access memory (random acceS memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.

The above embodiments are merely for illustrating the technical solution of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the corresponding technical solutions.

Claims

1. The firmware self-adaptive upgrading method is characterized by comprising the following steps of:

2. The method for adaptively upgrading firmware according to claim 1, wherein the step of performing device status monitoring on the target firmware to obtain processor load data, memory usage data, and network connection data includes:

monitoring the equipment state of the target firmware through a preset firmware management platform to obtain equipment state monitoring data;

inputting the equipment state monitoring data into a preset K-means clustering algorithm, and performing cluster center calculation on the equipment state monitoring data through the K-means clustering algorithm to obtain a plurality of corresponding first cluster centers;

Generating a plurality of corresponding first cluster clusters according to the plurality of first cluster centers, performing distance calculation on the plurality of first cluster clusters and the plurality of first cluster centers to obtain a plurality of center point distances of each first cluster, and performing mean value operation on the plurality of center point distances of each first cluster to obtain a mean value of the point distances of each first cluster;

performing cluster center correction on the first cluster centers through the point distance average value to obtain a plurality of second cluster centers;

and performing secondary clustering on the equipment state monitoring data through the plurality of second clustering centers to obtain a plurality of second clustering clusters, and generating corresponding processor load data, memory use data and network connection data according to the plurality of second clustering clusters.

3. The method of claim 1, wherein creating a first firmware upgrade knowledge graph according to the processor load data, the memory usage data, and the network connection data, and performing knowledge graph expansion update on the first firmware upgrade knowledge graph to obtain a second firmware upgrade knowledge graph comprises:

determining a plurality of corresponding processor load entities according to the processor load data, determining a plurality of corresponding memory use entities according to the memory use data, and determining a plurality of corresponding network connection entities according to the network connection data;

Performing relationship analysis on the plurality of processor load entities and the plurality of memory use entities through a preset Bayesian algorithm to obtain a plurality of first entity relationships, and performing relationship analysis on the plurality of memory use entities and the plurality of network connection entities through the Bayesian algorithm to obtain a plurality of second entity relationships;

creating a first hierarchical structure between the plurality of processor load entities and the plurality of memory use entities according to the plurality of first entity relationships, and creating a second hierarchical structure between the plurality of memory use entities and the plurality of network connection entities according to the plurality of second entity relationships;

carrying out knowledge graph construction on the first hierarchical structure and the second hierarchical structure to generate a corresponding first firmware upgrading knowledge graph;

and carrying out knowledge graph expansion updating on the first firmware upgrading knowledge graph to obtain a second firmware upgrading knowledge graph.

4. The method for adaptively upgrading firmware according to claim 3, wherein said performing performance evaluation index calculation on the target firmware based on the second firmware upgrading knowledge graph to obtain a plurality of performance evaluation indexes includes:

Performing knowledge graph traversal on the target firmware in the second firmware upgrading knowledge graph to obtain a knowledge graph traversal result;

performing performance factor identification and sequencing on the target firmware according to the knowledge graph traversal result to obtain a plurality of performance factors;

and performing performance evaluation index mapping on the plurality of performance factors to obtain a plurality of performance evaluation indexes.

5. The method of claim 1, wherein performing index normalization and vector conversion on the plurality of performance evaluation indexes to generate a performance evaluation input vector comprises:

performing evaluation index normalization on each performance evaluation index to generate a plurality of normalized evaluation indexes;

performing serialization conversion on the plurality of normalized evaluation indexes to generate a target evaluation index sequence;

and carrying out vector conversion on the target evaluation index sequence to generate a performance evaluation input vector.

6. The method of claim 1, wherein creating a firmware upgrade policy decision model and performing a firmware upgrade policy decision on the performance evaluation input vector by the firmware upgrade policy decision model to obtain a first firmware adaptive upgrade policy comprises:

Creating a firmware upgrade policy decision model, wherein the firmware upgrade policy decision model comprises: a random forest network, a support vector machine network and an output layer;

carrying out firmware upgrading strategy decision on the performance evaluation input vector through the random forest network to obtain a first decision result;

carrying out firmware upgrading strategy decision on the performance evaluation input vector through the support vector machine network to obtain a second decision result;

acquiring first network weight data of the random forest network and second network weight data of the support vector machine network;

and carrying out result weighted fusion on the first decision result and the second decision result according to the first network weight data and the second network weight data through the output layer to obtain a first firmware self-adaptive upgrade strategy.

7. The method of claim 1, wherein generating a second firmware adaptive upgrade policy corresponding to the target firmware according to the first firmware adaptive upgrade policy, and adaptively upgrade the target firmware by the second firmware adaptive upgrade policy, and outputting a firmware upgrade log, includes:

According to the first firmware self-adaptive upgrading strategy, carrying out upgrading stage analysis on the target firmware to obtain a corresponding target upgrading stage;

carrying out strategy optimization parameter initialization on the first firmware self-adaptive upgrading strategy according to the target upgrading stage through a preset genetic algorithm to generate a plurality of first strategy optimization parameters;

respectively calculating first matching degree data of each first strategy optimization parameter, and carrying out parameter group division on the plurality of first strategy optimization parameters according to the first matching degree data to obtain a plurality of strategy optimization parameter groups;

propagating, intersecting and mutating the strategy optimization parameter groups to obtain a plurality of second strategy optimization parameters, and respectively calculating second matching degree data of each second strategy optimization parameter;

according to the second matching degree data, performing non-dominant row ordering analysis on the plurality of second strategy optimization parameters to obtain a strategy optimization parameter sequence;

performing optimization analysis on the plurality of second strategy optimization parameters according to the strategy optimization parameter sequence to generate a second firmware self-adaptive upgrading strategy corresponding to the target firmware;

and carrying out self-adaptive upgrade on the target firmware through the second firmware self-adaptive upgrade strategy, and outputting a firmware upgrade log.

8. A firmware adaptive upgrade system, the firmware adaptive upgrade system comprising:

9. A firmware adaptive upgrade apparatus, the firmware adaptive upgrade apparatus comprising: a memory and at least one processor, the memory having instructions stored therein;

the at least one processor invoking the instructions in the memory to cause the firmware adaptive upgrade apparatus to perform the firmware adaptive upgrade method of any of claims 1-7.