Distributed network resource dynamic allocation method and system based on multiple strategies

Abstract

The invention relates to the technical field of network resource allocation and management, and discloses a distributed network resource dynamic allocation method and system based on multiple strategies, and the method comprises the steps: collecting node demand related data of each network node of a distributed network and equipment health state data of an edge node, and carrying out the preprocessing; constructing an improved prediction network model and inputting the preprocessed data into the prediction network model for training; predicting the future resource demand of each network node and the equipment alarm level of each edge node through the trained prediction network model; according to the predicted future resource demand of each network node, adopting a dynamic allocation strategy to optimize allocated resources; and the actual alarm level of the edge node is updated according to the predicted alarm level, and corresponding management measures are taken according to different alarm levels. According to the method, the utilization efficiency of distributed network resources and the overall performance of the network are remarkably improved, and the stability and reliability of the network are enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of network resource allocation and management, and in particular to a distributed network resource dynamic allocation method and system based on multiple strategies. Background Art

[0002] With the rapid development of information technology, distributed networks are increasingly widely used in all walks of life, the scale of networks is constantly expanding, and the application scenarios are becoming more and more complex. Traditional network resource allocation methods are mostly based on static rules or simple prediction models. These methods are unable to cope with the dynamic changes in node requirements and uncertain factors in complex network environments. When the node business volume suddenly increases or network failures cause resource fluctuations, traditional methods often lead to uneven resource allocation, delayed response to key services, and the coexistence of idle and overloaded resources in some nodes. In addition, as an important part of the network, the edge node lacks an effective intelligent early warning mechanism for its equipment management, and usually relies on manual inspections or fixed threshold judgments, which makes the discovery of potential equipment failures delayed and maintenance untimely, thereby affecting the overall performance and reliability of the network, increasing operation and maintenance costs, and increasing the risk of business interruption. Therefore, the existing network resource allocation and management methods can no longer meet the urgent needs of modern distributed networks for efficiency, stability and intelligent management. Summary of the invention

[0003] In order to overcome the shortcomings of the prior art, the present invention aims to provide a method and system for dynamic allocation of distributed network resources based on multiple strategies, which can achieve optimal configuration of resources and reliable operation of equipment by accurately predicting node requirements and intelligently managing edge nodes, thereby improving the overall performance and operation and maintenance efficiency of the network, and solving the technical problem that the existing network resource allocation and management methods can no longer meet the requirements of modern distributed networks for high efficiency, stability and intelligent management.

[0004] The purpose of the present invention can be achieved through the following technical solutions:

[0005] In a first aspect, the present invention provides a distributed network resource dynamic allocation method based on multiple strategies, comprising:

[0006] Collect node demand-related data of each network node in the distributed network and device health status data of edge nodes and perform preprocessing;

[0007] An improved prediction network model is constructed and the preprocessed data is input into the prediction network model for training. The improvement and training process of the prediction network model includes: capturing the long-range dependency relationship between nodes through an approximate attention mechanism, and processing input features through a number of hidden layers to predict the future resource demand of network nodes and the alarm level of edge node devices; wherein a loss function composed of a weighted combination of data fitting loss, physical constraint loss and alarm level constraint loss is constructed for training the model;

[0008] Through the trained prediction network model, the future resource demand of each network node and the alarm level of each edge node device are predicted;

[0009] Based on the predicted future resource requirements of each network node, a dynamic allocation strategy is used to optimize the allocated resources; and based on the predicted alarm level of each edge node device and the preset alarm threshold, the actual alarm level of the edge node is updated, and corresponding management measures are taken according to different alarm levels.

[0010] Preferably, the node demand-related data includes historical resource usage, network traffic characteristics, time characteristics, changes in the number of users or devices connected to the node, the type of running applications and load conditions; the health status data of the edge node device includes the hardware parameters of the edge node device, software operation logs, and network connection quality data; and the preprocessing includes cleaning, standardization and time series alignment of the data.

[0011] Preferably, the step of optimizing the allocated resources using a dynamic allocation strategy based on the predicted future resource requirements of each network node includes:

[0012] Based on the future resource requirements of each network node predicted by the trained model, a dynamic allocation strategy is adopted to minimize the difference between the predicted resource requirements and the actual allocated resources, and resources are allocated by weighted summation of node priorities. In the allocation process, the limited total resources and node operation constraints are considered, and feedback adjustments are made based on the difference between actual resource usage and allocated resources to optimize the resource allocation strategy.

[0013] Preferably, the actual alarm level of the edge node is updated according to the predicted alarm level of each edge node device and the preset alarm threshold, and corresponding management measures are taken according to different alarm levels, including:

[0014] According to the predicted alarm level of each edge node device and the preset alarm threshold, the actual alarm level of the edge node is updated, and corresponding management measures are taken according to different alarm levels, including automatic inspection and remote diagnosis.

[0015] Preferably, the approximate attention mechanism adopts a Nyström-based method to replace the traditional attention mechanism, and forms new approximate queries and approximate keys through subsampling and mean aggregation to capture long-range dependencies between nodes; then the allocation and alarm levels are processed through the approximate attention mechanism.

[0016] Preferably, the data fitting loss in the loss function uses mean square error to measure the error of resource demand prediction, and uses absolute value error to measure the error of alarm level prediction; the physical constraint loss is constructed according to physical laws such as the law of conservation of resources; the alarm level constraint loss is constructed according to the correlation between the change of equipment alarm level and the equipment status indicator.

[0017] Preferably, the dynamic allocation strategy is optimized using the following formula:

[0018] ;

[0019] in, is the allocated resource vector, which represents the resource allocation result for each network node i; is the total number of nodes; The resources allocated to network node i; is the predicted future resource demand of network node i; is the priority weight of network node i; The role of , find a set of resource allocation schemes so that the predicted future resource demand of network node i and the corresponding actual allocated resources The weighted sum of squared errors between them is minimized.

[0020] Preferably, performing feedback adjustment based on the difference between actual resource usage and allocated resources to optimize the resource allocation strategy includes:

[0021] The difference between the actual resource usage of the computing node and the allocated resources is triggered when the difference is greater than the set tolerance threshold, which triggers the re-optimization of the resource allocation strategy.

[0022] Preferably, the step of updating the device alarm level includes: dividing the actual alarm level of the edge node into normal, minor alarm and moderate alarm levels according to the predicted alarm level and the preset alarm threshold, and taking corresponding management measures, including automatic inspection and remote diagnosis.

[0023] In a second aspect, the present invention further provides a distributed network resource dynamic allocation system based on multiple strategies, comprising:

[0024] The data collection module is used to collect and pre-process the node demand-related data of each network node in the distributed network and the health status data of the edge devices;

[0025] A model building module is used to build an improved prediction network model and input the preprocessed data into the prediction network model for training. The improvement and training process of the prediction network model includes: capturing the long-range dependency relationship between nodes through an approximate attention mechanism, and processing input features through a number of hidden layers to predict the future resource demand of the node and the alarm level of the edge node device; wherein a loss function composed of a weighted combination of data fitting loss, physical constraint loss and alarm level constraint loss is constructed for training the model;

[0026] The prediction output module is used to predict the future resource requirements of each network node and the alarm level of each edge node device through the trained prediction network model;

[0027] The optimization and update module is used to optimize the allocated resources by adopting a dynamic allocation strategy based on the predicted future resource requirements of each network node; and to update the actual alarm level of the edge node based on the predicted alarm level of each edge node device and the preset alarm threshold, and to take corresponding management measures according to different alarm levels.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] (1) Accurate prediction and dynamic allocation: By constructing an improved prediction network model, the present invention can accurately predict the future resource requirements of each node and the alarm level of the edge node. Based on these prediction results, a dynamic allocation strategy is adopted to optimize the allocation of resources, ensure the efficient use of resources, avoid resource waste and overload, and thus significantly improve the overall utilization of network resources.

[0030] (2) Improve network performance and stability: By adjusting resource allocation strategies in real time, it can ensure that key businesses receive sufficient support during peak resource demand, reduce response delays, improve user experience, and enhance the network's adaptability. At the same time, through intelligent management of edge nodes, potential faults can be discovered and handled in advance, thereby enhancing network stability and reliability and reducing the risk of business interruption.

[0031] (3) Intelligent management and reduced operation and maintenance costs: The intelligent early warning and management mechanism of the present invention can automatically inspect and remotely diagnose edge nodes, reduce the need for manual intervention, and reduce operation and maintenance costs. At the same time, through real-time feedback and adjustment of resource allocation strategies, operation and maintenance efficiency is improved, making network management more intelligent and automated.

[0032] In summary, the present invention significantly improves the utilization efficiency of distributed network resources and the overall performance of the network, reduces operation and maintenance costs, enhances the stability and reliability of the network, and provides strong support for efficient, stable and intelligent management of modern distributed networks. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which: Figure 1 It is a flow chart of a method for dynamic allocation of distributed network resources based on multiple strategies provided by an embodiment of the present invention; Figure 2 It is a data processing flow chart of a distributed network resource dynamic allocation method based on multiple strategies provided by an embodiment of the present invention; Figure 3 It is a unit module diagram of a distributed network resource dynamic allocation system based on multiple strategies provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0034] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0035] In the description of the present invention, it should be understood that descriptions involving orientations, such as up, down, front, back, left, right, etc., and orientations or positional relationships indicated are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present invention.

[0036] In the description of the present invention, "several" means one or more, "more" means more than two, "greater than", "less than", "exceed" etc. are understood as not including the number itself, and "above", "below", "within" etc. are understood as including the number itself. If there is a description of "first" or "second", it is only used for the purpose of distinguishing the technical features, and cannot be understood as indicating or implying the relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0037] The present invention aims to construct a set of distributed network resource dynamic allocation methods and systems based on multiple strategies. The core goal is to achieve optimal resource configuration and reliable operation of equipment by accurately predicting node requirements and intelligently managing edge nodes, thereby improving the overall network performance and operation and maintenance efficiency.

[0038] Specifically, with the help of data collection and preprocessing, the characteristics and laws of node demand and edge node device status data are mined. Using the improved prediction model, the future resource demand of the node is predicted and the alarm level of the edge node is evaluated in real time, which provides a key basis for resource allocation and equipment management. In the resource allocation link, according to the prediction results and priority strategy, combined with the total resource constraints and node operating conditions, the allocation strategy is dynamically optimized, and the actual resource usage deviation is responded to in real time through the adjustment mechanism to ensure that resource allocation is accurately adapted to business needs, improve resource utilization, reduce energy consumption and costs, and ensure network service quality. In terms of edge node management, according to the predicted alarm level and threshold system, the alarm level is updated in a timely and accurate manner, and the corresponding management measures are automatically triggered, such as automatic inspection for minor alarms and remote diagnosis for moderate alarms, etc., to achieve active prevention before failure and rapid response in case of failure, enhance the reliability and stability of edge nodes, reduce the workload of operation and maintenance and the duration of business interruption, and ultimately improve the adaptive ability, operation efficiency and service continuity of distributed networks in complex environments.

[0039] In the present invention, network nodes refer to general nodes in the network, which can be servers, client devices or other network devices, which are responsible for processing data transmission, storage and computing tasks in the network. The demand and resource usage of network nodes reflect the basic operating status of the entire network. Edge nodes specifically refer to devices located at the edge of the network, which are usually closer to the user end and are used to provide low-latency, high-bandwidth services. The health status and management of edge nodes are crucial to ensuring the stability and reliability of the network.

[0040] See also Figure 1-Figure 3 The embodiment of the present invention provides a distributed network resource dynamic allocation method 100 based on multiple strategies, comprising:

[0041] S110: collecting node demand-related data of each network node in the distributed network and device health status data of edge nodes and performing preprocessing;

[0042] In this step S110, a data collection module is deployed on the devices of each network node and edge node of the distributed network to collect multi-dimensional data:

[0043] (1) Node demand-related data: including historical resource usage (such as CPU usage, memory usage, network bandwidth consumption, etc.), network traffic characteristics (such as packet size, latency, packet loss rate), time characteristics (such as timestamp, weekend / weekday), changes in the number of users or devices connected to the node, the type of application running and load conditions, etc. The timestamp is recorded to facilitate the analysis of time series data.

[0044] (2) Device health status data of edge nodes: collect hardware parameters (such as temperature, voltage, fan speed, load, etc.), software operation logs (such as system error information, application crash records, etc.), and network connection quality data (such as latency, packet loss rate, etc.).

[0045] Preprocessing includes data cleaning, standardization, and time series alignment. Data cleaning is used to remove data points that are obviously erroneous or abnormal, such as extreme values ​​or minimum values ​​caused by sensor failure. Data standardization is used to convert different types of data to a unified numerical range, such as converting data to a value between 0 and 1 to facilitate subsequent model processing. Time series alignment is used to ensure that node demand data and edge node device status data are synchronized in time for joint analysis.

[0046] S120: constructing an improved prediction network model and inputting the preprocessed data into the prediction network model for training, wherein the improvement and training process of the prediction network model includes: capturing the long-range dependency relationship between nodes through an approximate attention mechanism, and processing input features through a plurality of hidden layers to predict the future resource demand of the node and the alarm level of the edge node device; wherein a loss function composed of a weighted combination of a data fitting loss, a physical constraint loss, and an alarm level constraint loss is constructed for training the model;

[0047] This step S120 is used to realize the construction and improvement of the network model, and input the data obtained in step S110 into the improved prediction network model for training. Taking the following process as an example, this step S120 includes the following steps:

[0048] S121: define various variables;

[0049] (1) Resource demand vector : Network Node At the moment The resource demand vector ,in Can represent network nodes At the moment For The demand for various resources (such as CPU, memory, bandwidth, etc.), The number of resource types.

[0050] (2) Device state vector : Edge Node At the moment The device state vector ,in Can represent edge nodes At the moment No. device status indicators (such as temperature, voltage, packet loss rate, etc.), is the number of device status indicators.

[0051] (3) Input variables :Change time and network nodes or edge node The identifier (such as node number, edge node number, etc., which can be converted into numerical form through appropriate encoding methods) is used as the input variable of the neural network, denoted as

[0052] (4) Output variables: There are two main output variables: one is the predicted future resource demand vector of network nodes, denoted as ; The other is the predicted edge node device alarm level (different levels can be represented by appropriate numerical codes, such as 0 for normal, 1 for minor alarm, 2 for moderate alarm, etc.), recorded as .

[0053] S122: Construction and improvement of prediction network model;

[0054] Step S1221: approximate attention mechanism;

[0055] The traditional attention mechanism has high computational complexity when processing a large amount of node data and is difficult to apply in real time. Therefore, the present invention adopts an approximate attention mechanism based on the Nyström method to replace the traditional attention mechanism to capture the long-range dependencies between nodes and reduce the computational complexity. The Nyström method is a prior art proposed in 1977 and will not be described here.

[0056] The model adopts an approximate attention mechanism to form new approximate queries and approximate keys through subsampling and mean aggregation to capture the long-range dependencies between nodes.

[0057] Specifically, it includes:

[0058] (1) Input feature quantization;

[0059] The input feature vector The query takes the node requirements and device status data in as input and key( ), capturing long-range dependencies in these data.

[0060] ;

[0061] in, is the number of nodes, is the dimension of the feature, Represents the field of real numbers.

[0062] (2) Subsampling and mean aggregation;

[0063] For Query and key( ) is sub-sampled to obtain:

[0064] ;

[0065] in, is the dimension after subsampling (larger than the original matrix dimension small), and form a new approximate query by average aggregation and approximate keys ( ).

[0066] (3) Approximate attention calculation:

[0067] A. Use approximate query and approximate keys ( ) Calculate the attention weight , the formula is:

[0068] ;

[0069] Here, the softmax function is used to normalize the attention weights so that their sum is 1.

[0070] B. The calculation of the dependency relationship between nodes is adjusted by the pseudo-inverse matrix. The formula is:

[0071] ;

[0072] The pseudo-inverse matrix is ​​used to handle possible singular matrix situations and ensure the stability of the calculation.

[0073] Finally, the processing results of resource allocation and alarm level through the approximate attention mechanism are:

[0074] ;

[0075] In the above formula, Represents device states or other decision variables; is the output of the approximate attention mechanism, which contains the long-range dependencies between the input features; softmax normalizes each element of the matrix through an exponential function so that the output value is The sum of each row is 1. represents the dot product of the approximate query matrix and the approximate key matrix, and represents the relationship matrix between nodes after subsampling; The relationship matrix is ​​scaled and normalized to prevent the gradient vanishing or exploding problem in the attention mechanism.

[0076] Step S1222: passing input features to the hidden layer;

[0077] Next, the model will be processed through several hidden layers , the model is hidden layers, where The forward propagation process of the layer is expressed as follows:

[0078] ;

[0079] Among them, the initial input features are , that is, the approximate attention output obtained by step S1221. It is The output vector of the hidden layer, is the activation function (such as ReLU, wait), and They are The weight matrix and bias vector of the layer, is the total number of hidden layers; is the predicted future resource demand vector of the node; It is the predicted alarm level of the edge node device (different levels can be represented by appropriate numerical codes, such as 0 for normal, 1 for minor alarm, 2 for moderate alarm, etc.).

[0080] Step S1223: construct a loss function;

[0081] Loss Function It consists of three weighted parts: data fitting loss , Physical constraint loss (defined according to the physical laws related to distributed networks) and alarm level constraint loss :

[0082] ;

[0083] in, is the corresponding weight coefficient, which is used to balance the importance of different loss parts in training.

[0084] Preferably, the data fitting loss in the loss function is The mean square error is used to measure the error of resource demand prediction, and the absolute value error is used to measure the error of alarm level prediction; physical constraint loss Constructed according to physical laws such as the law of conservation of resources; alarm level constraints on losses It is constructed based on the correlation between the change of equipment alarm level and equipment status indicators. Specifically,

[0085] (1) Data fitting loss :

[0086] ;

[0087] in, is the number of training data points, and They are and input data points, and are the corresponding real resource demand vector and real alarm level respectively. Here, the mean square error is used to measure the error of resource demand prediction, and the absolute value error is used to measure the error of alarm level prediction. They are the corresponding future resource demand vector of the node predicted by the model and the alarm level of the edge node device; This term is the mean square error, which is used to measure the difference between the resource demand predicted by the model and the actual resource demand. By minimizing this term, the model will adjust its parameters to reduce the error in resource demand prediction. This item is the absolute error, which is used to measure the difference between the alarm level predicted by the model and the actual alarm level. Since the alarm level is usually a discrete level classification, it is more appropriate to use the absolute value error.

[0088] (2) Physical constraint loss :

[0089] In a distributed network, physical constraints are constructed based on physical laws such as the law of conservation of resources. Assuming that the total amount of resources in the entire distributed network is fixed, Indicates that the resource demand prediction of all nodes is converted into the form of loss function, and the The resource conservation constraint loss is then accumulated and averaged at different times during the training process to obtain the total physical constraint loss :

[0090] ;

[0091] Law of conservation of resources (constraints): ;

[0092] in, is the rate of change of resource demand; is the network node i at time For The predicted demand for various resources (such as CPU, memory, bandwidth, etc.); Representative The total amount of the resource (in the entire distributed network); Indicates the number of resource types and is used to index different types of resources. is the total number of nodes; is the length of the time horizon considered.

[0093] (3) Alarm level constraint loss :

[0094] Constraints are constructed based on the correlation between the change in device alarm level and device status indicators. If a device status indicator exceeds the normal range, the alarm level will increase, which can be expressed as:

[0095] ;

[0096] in, is the number of data points associated with the alarm level, the device state vector ,in Represents an edge node At the moment No. device status indicators (such as temperature, voltage, packet loss rate, etc.).

[0097] in, is a function that describes the relationship between the change in device alarm level and device status indicators. First, each device status indicator is judged to see whether it exceeds the normal range. Represents an edge node No. The threshold corresponding to the device status indicator is based on the different edge nodes. The characteristics of different equipment status indicators are set separately to accurately determine whether the indicator is in an abnormal state at the current node, thereby affecting the alarm level. Used to indicate whether the indicator is in an abnormal state at the current node:

[0098] ;

[0099] The device alarm level is calculated based on the triggering conditions of the above status indicators and their weights. The formula is as follows:

[0100] ;

[0101] In the above formula, Represents an edge node No. The weights corresponding to the device status indicators reflect the differential importance of different device status indicators on different edge nodes k on the alarm level. This is to map the calculation results to discrete alarm level values ​​(for example, we can set When the alarm level is 0, it means normal; in the interval When the alarm level is 1, it means a minor alarm; When the alarm level is 2, it indicates a moderate alarm).

[0102] S130: Predicting the future resource requirements of each network node and the alarm level of each edge node device through the trained prediction network model;

[0103] In step S130: two main output variables are output through the trained prediction network model, that is, the future resource demand of each node is predicted and predicted edge node device alarm levels .

[0104] S140: Based on the predicted future resource requirements of each node, a dynamic allocation strategy is used to optimize the allocated resources; and based on the predicted alarm level of each edge node device and the preset alarm threshold, the actual alarm level of the edge node is updated, and corresponding management measures are taken according to different alarm levels.

[0105] In the embodiment of the present invention, the network node optimizes its resource usage through a dynamic resource allocation strategy to ensure efficient use of network resources. In addition to resource allocation, edge nodes need to perform intelligent early warning and management based on the health status data of the equipment, such as automatic inspection and remote diagnosis, to detect and solve potential faults in advance.

[0106] The step S140 includes the following two steps, namely, dynamic allocation of distributed network resources and updating of device alarm levels, and specifically includes the following steps:

[0107] S141: According to the predicted future resource requirements of each node, a dynamic allocation strategy is used to optimize the allocated resources, including:

[0108] Based on the future resource requirements of each network node predicted by the trained model, a dynamic allocation strategy is adopted to minimize the difference between the predicted resource requirements and the actual allocated resources, and resources are allocated by weighted sum of the node priorities; in the allocation process, the limited total resources and node operation constraints are considered, and feedback adjustments are made based on the difference between the actual resource usage and the allocated resources to optimize the resource allocation strategy. More specifically, it includes:

[0109] S1411: Resource allocation;

[0110] Through the trained model, the predicted future resource demand vector of each network node is obtained , a dynamic allocation strategy is adopted, the goal is to minimize the difference between the predicted resource requirements and the actual allocated resources, and to sum them weighted by the priority of the nodes in order to optimize the allocated resources:

[0111] ;

[0112] in, is the allocated resource vector, which represents the allocated resource vector for each network node i (in total Nodes) resource allocation results; Indicates the resources allocated to network node i; is the predicted resource demand of network node i, which is the predicted value obtained by the model in step S130; is the priority weight of network node i, which is used to measure the importance of the node in resource allocation. Nodes with larger weights will receive more resource allocation; The role of (the amount of resources allocated to each node), find a set of resource allocation schemes that make the predicted demand and actual allocation The weighted sum of squared errors between them is minimized.

[0113] S1412: Add constraints;

[0114] In the process of resource allocation, some constraints need to be considered to ensure that resource allocation is reasonable and feasible.

[0115] 1) Limited total resources: the sum of resources allocated to all nodes cannot exceed the total resources.

[0116] ;

[0117] 2) Node operation constraints: that is, the resource allocation of each node must be within its minimum and maximum requirements.

[0118] ;

[0119] in, is the minimum resource requirement of network node i, indicating the minimum resource allocation that the node can accept; is the maximum resource requirement of network node i, indicating the maximum resource allocation that the node can accept.

[0120] S1413: Feedback adjustment;

[0121] In order to dynamically optimize according to the actual resource utilization, a feedback adjustment mechanism is set: including the difference between the actual resource usage of the computing node and the allocated resources. When there is a large difference, specifically when the difference is greater than the set tolerance threshold, the resource allocation strategy is re-optimized and updated. Specifically:

[0122] ;

[0123] in, is the network node i at time The actual resource usage (obtained through monitoring data); is the resource adjustment amount of network node i, which represents the difference between the allocated resources and the actually used resources.

[0124] Update conditions:

[0125] if , then it is necessary to update the resource allocation strategy, that is, re-optimize resource allocation. The tolerance threshold is set. If the resource difference is greater than the threshold, re-optimization is triggered.

[0126] Through this feedback adjustment mechanism, the model can optimize the allocation strategy in real time according to the actual resource usage and reduce resource waste or shortage.

[0127] S142: updating the actual alarm level of the edge node according to the predicted alarm level of each edge node device and the preset alarm threshold, and taking corresponding management measures according to different alarm levels, including:

[0128] In this step S142, the actual alarm level of the edge node is updated according to the predicted alarm level of each edge node device and the preset alarm threshold, and corresponding management measures are taken according to different alarm levels, including automatic inspection and remote diagnosis. The step of updating the device alarm level includes: according to the predicted alarm level of each edge node device and the preset alarm threshold, the actual alarm level of the edge node is divided into normal, slight alarm and moderate alarm levels, and corresponding management measures are taken respectively, including automatic inspection and remote diagnosis.

[0129] Specifically, according to the predicted alarm level of each edge node device and preset alarm thresholds (setting thresholds Distinguish between normal and minor alarms, thresholds Distinguish between minor and moderate alarms, etc.), update the actual alarm level of edge node k, and take corresponding management measures. Take different measures to , start the automatic inspection program; for , arrange for operation and maintenance personnel to perform remote diagnosis and operation:

[0130] ;

[0131] It is an edge node The actual alarm level, which is based on the predicted alarm level and alarm thresholds ( , ) is used to decide what management measures to take for edge node k.

[0132] According to the above-mentioned embodiments of the present invention, through accurate prediction, dynamic allocation, intelligent management and adaptive adjustment, the utilization efficiency of distributed network resources and the overall performance of the network are significantly improved, the operation and maintenance costs are reduced, the stability and reliability of the network are enhanced, and strong support is provided for the efficient, stable and intelligent management of modern distributed networks.

[0133] The above is an introduction to a method embodiment. The following is a further explanation of the solution of the present invention through an apparatus embodiment.

[0134] Figure 3 FIG. 2 shows a module schematic diagram of a distributed network resource dynamic allocation system 200 based on multiple strategies according to an embodiment of the present invention. Figure 3 As shown, a distributed network resource dynamic allocation system 200 based on multiple strategies of the present invention includes:

[0135] The data collection module 201 is used to collect node demand related data of each network node in the distributed network and device health status data of the edge node and perform pre-processing;

[0136] The model building module 202 is used to build an improved prediction network model and input the preprocessed data into the prediction network model for training. The improvement and training process of the prediction network model includes: capturing the long-range dependency relationship between nodes through an approximate attention mechanism, and processing input features through a number of hidden layers to predict the future resource demand of network nodes and the alarm level of edge node devices; wherein a loss function composed of a weighted combination of data fitting loss, physical constraint loss and alarm level constraint loss is constructed for training the model;

[0137] The prediction output module 203 is used to predict the future resource demand of each network node and the alarm level of each edge node device through the trained prediction network model;

[0138] The optimization and update module 204 is used to optimize the allocated resources by adopting a dynamic allocation strategy according to the predicted future resource requirements of each node; and to update the actual alarm level of the edge node according to the predicted alarm level of each edge node device and the preset alarm threshold, and to take corresponding management measures according to different alarm levels.

[0139] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the described module can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0140] The method of this embodiment and the above-mentioned reaction device are not described in detail here.

[0141] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made according to the scope of application of the present invention should still fall within the scope of the claims of the present invention.

Claims

1. A distributed network resource dynamic allocation method based on multiple strategies, characterized in that: include: Collect node demand-related data of each network node in the distributed network and device health status data of edge nodes and perform preprocessing; An improved prediction network model is constructed and the preprocessed data is input into the prediction network model for training. The improvement and training process of the prediction network model includes: capturing the long-range dependency relationship between nodes through an approximate attention mechanism, and processing input features through a number of hidden layers to predict the future resource demand of network nodes and the alarm level of edge node devices; wherein a loss function composed of a weighted combination of data fitting loss, physical constraint loss and alarm level constraint loss is constructed for training the model; Through the trained prediction network model, the future resource demand of each network node and the alarm level of each edge node device are predicted; Based on the predicted future resource requirements of each network node, a dynamic allocation strategy is used to optimize the allocated resources; and based on the predicted alarm level of each edge node device and the preset alarm threshold, the actual alarm level of the edge node is updated, and corresponding management measures are taken according to different alarm levels.

2. The method for dynamic allocation of distributed network resources based on multiple strategies according to claim 1, characterized in that: in, The node demand-related data includes historical resource usage, network traffic characteristics, time characteristics, changes in the number of users or devices connected to the node, the type of running applications and load conditions; the device health status data of the edge node includes the hardware parameters of the edge node device, software operation logs, and network connection quality data; preprocessing includes data cleaning, standardization and time series alignment processing.

3. The method for dynamic allocation of distributed network resources based on multiple strategies according to claim 2 is characterized in that: in, The method of optimizing the allocated resources by using a dynamic allocation strategy based on the predicted future resource requirements of each node includes: Based on the future resource requirements of each network node predicted by the trained model, a dynamic allocation strategy is adopted to minimize the difference between the predicted resource requirements and the actual allocated resources, and resources are allocated by weighted summation of the priorities of network nodes. The limited total resources and node operation constraints are taken into account during the allocation process, and feedback adjustments are made based on the difference between the actual resource usage and the allocated resources to optimize the resource allocation strategy.

4. The method for dynamic allocation of distributed network resources based on multiple strategies according to claim 3 is characterized in that: in, The actual alarm level of the edge node is updated according to the predicted alarm level and the preset alarm threshold, and corresponding management measures are taken according to different alarm levels, including: According to the predicted alarm level of each edge node device and the preset alarm threshold, the actual alarm level of the edge node is updated, and corresponding management measures are taken according to different alarm levels, including automatic inspection and remote diagnosis.

5. The method for dynamic allocation of distributed network resources based on multiple strategies according to claim 1 is characterized in that: in, The approximate attention mechanism adopts the Nyström method to replace the traditional attention mechanism, and forms new approximate queries and approximate keys through subsampling and mean aggregation to capture the long-range dependencies between nodes; then the resource allocation and alarm level are processed through the approximate attention mechanism.

6. The method for dynamic allocation of distributed network resources based on multiple strategies according to claim 1 is characterized in that: The data fitting loss in the loss function uses mean square error to measure the error of resource demand prediction, and uses absolute value error to measure the error of alarm level prediction; the physical constraint loss is constructed according to physical laws such as the law of conservation of resources; the alarm level constraint loss is constructed according to the correlation between the change of equipment alarm level and equipment status indicators.

7. The method for dynamic allocation of distributed network resources based on multiple strategies according to claim 3 is characterized in that: The dynamic allocation strategy is optimized using the following formula: ； in, is the allocated resource vector, which represents the resource allocation result for each network node i; is the total number of network nodes; The resources allocated to network node i; is the predicted future resource demand of network node i; is the priority weight of network node i; The role of , find a set of resource allocation schemes so that the predicted future resource demand of network node i and the corresponding actual allocated resources The weighted sum of squared errors between them is minimized.

8. The method for dynamic allocation of distributed network resources based on multiple strategies according to claim 4, characterized in that: The feedback adjustment based on the difference between the actual resource usage and the allocated resources to optimize the resource allocation strategy includes: The difference between the actual resource usage of the computing node and the allocated resources is triggered when the difference is greater than the set tolerance threshold, which triggers the re-optimization of the resource allocation strategy.

9. The method for dynamic allocation of distributed network resources based on multiple strategies according to claim 1, characterized in that: The step of updating the device alarm level includes: dividing the actual alarm level of the edge node into normal, slight alarm and moderate alarm levels according to the predicted alarm level of each edge node device and the preset alarm threshold, and taking corresponding management measures respectively, including automatic inspection and remote diagnosis.

10. A distributed network resource dynamic allocation system based on multiple strategies, characterized in that: include: The data collection module is used to collect and pre-process the node demand-related data of each network node in the distributed network and the health status data of the edge devices; A model building module, used to build an improved prediction network model and input the preprocessed data into the prediction network model for training, wherein the improvement and training process of the prediction network model includes: capturing the long-range dependency relationship between nodes through an approximate attention mechanism, and processing input features through a number of hidden layers to predict the future resource demand of network nodes and the alarm level of edge node devices; wherein a loss function composed of a weighted combination of data fitting loss, physical constraint loss and alarm level constraint loss is constructed for training the model; The prediction output module is used to predict the future resource requirements of each network node and the alarm level of each edge node device through the trained prediction network model; The optimization and update module is used to optimize the allocated resources by adopting a dynamic allocation strategy based on the predicted future resource requirements of each network node; and to update the actual alarm level of the edge node based on the predicted alarm level of each edge node device and the preset alarm threshold, and to take corresponding management measures according to different alarm levels.

Images