Resource allocation and flow control methods, apparatus and equipment based on multi-container networks

Abstract

This application relates to a method, apparatus, and device for resource allocation and traffic control based on a multi-container network. The method includes: acquiring network state information of the multi-container network; sequentially performing feature extraction and feature enhancement processing on the network state information using a pre-built multi-information fusion model to obtain multi-information fusion features; and sequentially performing decoding and mapping processing on the multi-information fusion features using a pre-built multi-information decoding model to obtain resource allocation results and traffic control results for each container. The embodiments of this application can meet the ever-changing network service quality requirements.

Description

Technical Field

[0001] This application relates to the field of cloud computing technology, and in particular to a method, apparatus and device for resource allocation and traffic control based on a multi-container network. Background Technology

[0002] In today's digital age, with the rapid development of cloud computing, enterprises and service providers rely on multi-container networks to support complex applications and services. Multi-container network architectures, by breaking down applications into multiple containers, enable greater flexibility, scalability, and fault tolerance.

[0003] However, with the increase in the number of containers and the expansion of application scale, traditional resource allocation and traffic control methods are struggling to meet the ever-changing network service quality requirements. Summary of the Invention

[0004] Therefore, it is necessary to provide a resource allocation and traffic control method, apparatus, and device based on multi-container networks that can meet the ever-changing network service quality requirements, addressing the aforementioned technical problems.

[0005] Firstly, this application provides a resource allocation and traffic control method based on a multi-container network, including:

[0006] Obtain network status information for a multi-container network;

[0007] By using a pre-built multi-information fusion model, feature extraction and feature enhancement processes are sequentially performed on network state information to obtain multi-information fusion features.

[0008] By using a pre-built multi-information decoding model, the multi-information fusion features are sequentially decoded and mapped to obtain the resource allocation results and flow control results for each container.

[0009] Secondly, this application also provides a resource allocation and flow control device based on a multi-container network, comprising:

[0010] The information acquisition module is used to acquire network status information of the multi-container network;

[0011] The enhancement processing module is used to sequentially perform feature extraction and feature enhancement processing on the network state information using a pre-built multi-information fusion model to obtain multi-information fusion features.

[0012] The mapping processing module is used to sequentially decode and map the multi-information fusion features using a pre-built multi-information decoding model to obtain the resource allocation results and flow control results for each container.

[0013] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method of any one of the first aspects. Fourthly, this application also provides a computer-readable storage medium storing a computer program thereon, wherein the computer program, when executed by a processor, implements the method of any one of the first aspects.

[0014] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the method of any one of the first aspects. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is an internal structural diagram of a computer device in one embodiment;

[0017] Figure 2 This is a flowchart illustrating a resource allocation and flow control method based on a multi-container network in one embodiment.

[0018] Figure 3 This is a flowchart illustrating a resource allocation and flow control method based on a multi-container network in another embodiment.

[0019] Figure 4 This is a structural diagram of a multi-information fusion model in one embodiment;

[0020] Figure 5 This is a flowchart illustrating a resource allocation and flow control method based on a multi-container network in another embodiment.

[0021] Figure 6 Here is a diagram of the feature enhancement model structure in one embodiment;

[0022] Figure 7 This is a flowchart illustrating a resource allocation and flow control method based on a multi-container network in another embodiment.

[0023] Figure 8 This is a flowchart illustrating a resource allocation and flow control method based on a multi-container network in another embodiment.

[0024] Figure 9 This is a flowchart illustrating a resource allocation and flow control method based on a multi-container network in another embodiment.

[0025] Figure 10 This is a structural block diagram of a resource allocation and flow control device based on a multi-container network in one embodiment. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0027] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 1 As shown, this computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores data related to resource allocation and flow control in a multi-container network. The I / O interfaces facilitate information exchange between the processor and external devices. The communication interface allows communication with external terminals via a network connection. When executed by the processor, the computer program implements a resource allocation and flow control method based on a multi-container network.

[0028] Those skilled in the art will understand that Figure 1 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0029] In one exemplary embodiment, such as Figure 2 As shown, a resource allocation and flow control method based on a multi-container network is provided, which is then applied to... Figure 1 The following steps, 201 to 203, are used as an example of computer equipment.

[0030] Step 201: Obtain network status information for the multi-container network.

[0031] In this context, multi-container networking typically refers to deploying applications using multiple containers within a distributed system. These containers may span multiple hosts and communicate over a network. Containers are lightweight virtualization environments used to encapsulate applications and their dependencies. Multiple containers are usually managed through a container orchestration platform (such as Kubernetes) to work together to provide efficient services.

[0032] Network status information includes communication status between various containers in the network, bandwidth utilization, latency, packet loss rate, and other information.

[0033] In this embodiment of the application, the computer device monitors the health status of the multi-container network in real time through the container management platform and obtains the network status information of the multi-container network.

[0034] In some embodiments, resource usage data for each container, such as CPU, memory, storage, and network bandwidth, can be collected to ensure that the status of each container in the network is effectively monitored. Simultaneously, network traffic analysis tools are used to monitor communication traffic between containers in real time, obtaining information such as traffic volume, protocol type, bandwidth usage, latency, and packet loss rate.

[0035] Step 202: The network state information is sequentially processed by feature extraction and feature enhancement using a pre-built multi-information fusion model to obtain multi-information fusion features.

[0036] Feature extraction processing refers to extracting important and effective features (such as traffic volume, response time, resource consumption, etc.) from network state information.

[0037] Feature enhancement processing enhances the expressive power of extracted features through algorithms to better reflect the state and load of container networks.

[0038] In this embodiment of the application, when processing network state information using a pre-built multi-information fusion model, the network state information first needs to be pre-processed. This includes removing redundant information, filling in missing values, standardizing the format, and performing normalization operations to ensure that data from different sources can be input into the subsequent multi-information fusion model with a consistent standard.

[0039] The next step is feature extraction. Feature extraction identifies key, meaningful data features from network state information. For example, relevant features can be extracted from container resource usage (such as CPU, memory, and bandwidth), network performance metrics (such as latency, packet loss rate, and bandwidth utilization), and container health status (such as error rate and load). The core purpose of feature extraction is to transform raw, complex data into a feature set that effectively describes the network state; these features will provide the foundation for subsequent processing.

[0040] After feature extraction, feature enhancement processing is performed to obtain multi-information fusion features. The purpose of feature enhancement is to improve the expressive power of the extracted features, making them more conducive to subsequent analysis and modeling. Feature enhancement methods can include data smoothing and denoising to reduce random fluctuations in the data; feature crossing and combination, which captures the relationships between features by combining multiple features into new composite features; and nonlinear transformations, such as logarithmic or power transformations, to address differences in features of different magnitudes.

[0041] In some embodiments, information fusion processing can also be performed to organically combine information from different data sources to form a multimodal feature representation. There are various information fusion methods, including weighted fusion, concatenation fusion, and hierarchical fusion. Weighted fusion assigns weights to different features, combining them into a single feature; concatenation fusion directly connects the feature vectors to form a higher-dimensional feature vector; and hierarchical fusion processes local features first and then merges them with global features. Information fusion can not only compensate for the shortcomings of a single data source but also capture more dimensions of network state information.

[0042] In the fused feature space, redundant or irrelevant features often exist, which can affect the performance and efficiency of the multi-information fusion model. Therefore, feature selection and optimization processes can be performed. Feature selection removes redundant features through correlation analysis and evaluates the importance of each feature, retaining those that contribute significantly to the decision-making of the multi-information fusion model. Furthermore, dimensionality reduction methods can further reduce the dimensionality of features, thereby improving the computational efficiency of the multi-information fusion model while avoiding overfitting.

[0043] Step 203: The multi-information fusion features are decoded and mapped sequentially using a pre-built multi-information decoding model to obtain the resource allocation results and flow control results for each container.

[0044] The resource allocation result refers to determining how to allocate network resources (such as CPU, memory, and network bandwidth) of multiple containers to different containers based on factors such as the current load of the containers, network bandwidth, and computing power, in order to ensure the efficient operation of multiple containers.

[0045] Flow control refers to how different traffic flows are managed and adjusted based on network conditions during network transmission. Flow control includes traffic scheduling, bandwidth allocation, and latency optimization to avoid network congestion, increase throughput, and reduce packet loss.

[0046] The aforementioned resource allocation and traffic control method based on multi-container networks first acquires the network status information of the multi-container network. Then, a pre-built multi-information fusion model is used to sequentially perform feature extraction and feature enhancement processing on the network status information, resulting in multi-information fusion features. Finally, a pre-built multi-information decoding model is used to sequentially decode and map the multi-information fusion features, yielding the resource allocation and traffic control results for each container. This method allows for precise monitoring of the current network environment by acquiring the network status information of the multi-container network. The use of the multi-information fusion model enables comprehensive processing of network status information, avoiding the limitations of single-dimensional information. By extracting and enhancing features from the network status information, effective and high-quality features can be extracted, allowing the network management system to better understand the changing trends of the network status. Through the decoding and mapping of the fusion features using the multi-information decoding model, the resource allocation and traffic control results for each container can be obtained. This allows for dynamic adjustment of resource allocation strategies and traffic control rules based on real-time changes in network demand, ensuring that each container receives the necessary resources under different load conditions, reducing resource waste or bottlenecks.

[0047] In one exemplary embodiment, such as Figure 3 As shown, the multi-information fusion model includes a multi-level sub-fusion model. Based on this, the network state information is sequentially processed by feature extraction and feature enhancement using the pre-constructed multi-information fusion model to obtain multi-information fusion features, including steps 301 to 303. Wherein:

[0048] Step 301: Iteratively calculate the network state information using a multi-level sub-fusion model. In the first iteration, the network state information is input into the first-level sub-fusion model. The first-level sub-fusion model is used to sequentially perform feature extraction and feature enhancement processing on the network state information to obtain the first fused feature.

[0049] In this embodiment of the application, during the initial iteration, network state information is input into the first-level sub-fusion model. For example... Figure 4 As shown, Figure 4 This is a structural diagram of a multi-information fusion model. Figure 4In this context, transformer, encoder, and encoder represent encoders; MLP represents multilayer perceptron; LSTM represents long short-term memory network; QoS represents the feature enhancement model; secondary input features represent network traffic information; primary input features represent resource demand information; and temporal input features represent temporal information.

[0050] The purpose of this process is to extract and enhance fundamental features from network state information to provide initial fusion features for subsequent iterations. In this process, the first-level sub-fusion model first performs feature extraction on the input network state information, identifying important fundamental features such as resource usage, network performance, and container health status from the raw data. Next, after feature enhancement processing, these features are further transformed and optimized, for example, through feature crossing, data smoothing, or nonlinear transformations, to enhance their expressive power and make them more suitable for subsequent model processing. Finally, after processing by the first-level sub-fusion model, the first fusion feature is obtained, which will be passed as input to the next stage of processing.

[0051] Step 302: In the non-first iteration process, the previous fusion feature obtained from the previous sub-fusion model is input into the current sub-fusion model. The current sub-fusion model is used to perform feature extraction and feature enhancement processing on the previous fusion feature in sequence to obtain the current fusion feature.

[0052] In the embodiments of this application, see also Figure 4 In subsequent iterations, the current-level sub-fusion model processes the data layer by layer based on the results of previous iterations. The output of the previous-level sub-fusion model, i.e., the previous fused feature, is input into the current-level sub-fusion model. This current-level sub-fusion model then performs feature extraction and enhancement on these fused features output from the previous level. Feature extraction at this stage focuses more on discovering deeper patterns from the already fused features; for example, by combining fused features from multiple dimensions, more complex relationships and trends can be extracted. After enhancement, the current-level sub-fusion model outputs its current-level fused feature, which is then passed as input to the next-level sub-fusion model.

[0053] Through this iterative process, each sub-fusion model continuously deepens and refines its features, gradually enhancing their expressiveness and information richness. Each iteration extracts more potential key information based on the fused features of the previous level, and further improves the effectiveness of the features through enhancement processing. This progressively deepening feature extraction and enhancement process makes the final fused features more comprehensive and better able to describe the complexity of network state information.

[0054] Step 303: Determine the fusion features output by the final-level sub-fusion model as multi-information fusion features.

[0055] In this embodiment, after processing through all iterative levels, the final sub-fusion model outputs the final multi-information fusion feature. This multi-information fusion feature is the cumulative result of all iterative processes, integrating the feature extraction and enhancement performed by each level of the model. Therefore, the final multi-information fusion feature not only contains the key information in the original network state information, but also captures more complex and deeper network state features through multiple iterations and enhancements, providing strong data support for subsequent decision-making and optimization tasks.

[0056] In the above embodiments, a multi-level sub-fusion model is used for iterative calculation of network state information. Through phased feature extraction and enhancement, features are gradually mined and optimized from simple to complex. Through the initial processing of the first-level sub-fusion model and the progression of each level of sub-fusion model, high-quality multi-information fusion features are finally obtained, which can effectively improve the perception of complex network states, thereby improving the accuracy and efficiency of subsequent decision-making.

[0057] In an exemplary embodiment, network state information includes resource demand information, network traffic information, and time series information. The sub-fusion model includes an encoder, a multilayer perceptron, and a long short-term memory network. Based on this, a pre-built multi-information fusion model is used to sequentially perform feature extraction and feature enhancement processing on the network state information to obtain multi-information fusion features, including:

[0058] In the first scenario, the resource requirement information is input into the encoder for feature extraction processing, resulting in the first resource requirement feature output by the encoder.

[0059] The encoder, a type of feedforward neural network, consists of multiple layers of neurons and is capable of effectively extracting high-dimensional abstract features from the raw input data. For resource demand information, the encoder performs a nonlinear transformation through a multi-layered structure, thereby capturing complex patterns and relationships.

[0060] In the embodiments of this application, see also Figure 4 As shown, resource requirement information is input into the encoder for feature extraction processing, extracting meaningful features from the resource requirement data, such as the CPU, memory, and bandwidth requirements of each container. After processing by the encoder, the first resource requirement feature is output.

[0061] In some embodiments, since network status information includes resource demand information, network traffic information, and time series information, resource demand information datasets, network traffic information datasets, and time series information datasets can be constructed first in practical applications.

[0062] (1) Construction of resource demand information dataset

[0063] This application embodiment considers the resource requirements of containers, such as CPU utilization and memory usage, to be the main characteristics of multi-container network management. Based on these characteristics, a resource requirement information dataset is constructed, and the data can be collected in the following ways:

[0064] Use the monitoring interface provided by the container orchestration tool to periodically obtain CPU and memory usage data for each container. For example, obtain the data every 5 seconds and record it.

[0065] Deploy a monitoring agent inside the container to collect and report resource usage in real time.

[0066] (2) Construction of network traffic information dataset

[0067] In this application embodiment, packet size in network traffic characteristics and idle or faulty state in container operation status are considered as secondary characteristics of multi-container network management, and a dataset is constructed based on these characteristics. Specifically, data can be collected in the following ways:

[0068] Regarding packet size in network traffic characteristics, traffic monitoring tools can be deployed on network nodes to capture packets and analyze their size.

[0069] For the idle or faulty state of a container during its operation, the status information can be obtained and recorded through the container's health check mechanism.

[0070] (3) Construction of time series information dataset

[0071] In this application embodiment, time-series data is considered as time-series characteristics of multi-container network management, such as CPU utilization, memory usage, network bandwidth requirements, network traffic periodicity, and the duration of container busy states over time. A dataset is constructed based on these characteristics. Specifically, it can be obtained through the following methods:

[0072] Timestamps are set for metrics such as CPU utilization, memory usage, and network bandwidth requirements. Samples are taken and recorded at fixed time intervals (such as every minute) to form time series data.

[0073] To determine the periodicity of network traffic, we analyze traffic data over a period of time (such as a day or a week) to identify its periodic patterns and record the relevant information in a time series dataset.

[0074] For the duration of a container being in a busy state, the start and end times of each busy state are calculated and recorded by monitoring the container's activity, thus obtaining time series data of the duration.

[0075] When building a dataset, it is also necessary to pay attention to the accuracy, completeness and consistency of the data. At the same time, the data should be cleaned and preprocessed to remove outliers and erroneous data to ensure the quality of the dataset.

[0076] Examine the entire dataset to identify and remove completely duplicate rows. This can be achieved through database deduplication or using a data processing library.

[0077] Outliers in numerical characteristics (such as CPU utilization, memory usage, etc.) can be identified using box plots or the 3σ rule. If a container's CPU utilization exceeds three times the standard deviation of the average, it can be considered an outlier. Outliers can be deleted, corrected, or retained depending on the specific circumstances. Reasonable outliers caused by temporary system spikes can be retained; those due to data acquisition errors should be corrected or deleted.

[0078] The data preprocessing process can be as follows:

[0079] (1) Data standardization / normalization: Min-Max normalization is performed on feature data of different magnitudes and ranges to facilitate subsequent analysis and model training.

[0080] (2) Feature encoding: For the classification features (such as application type, container running status, etc.) present in the dataset, one-hot encoding is used for encoding.

[0081] (3) Data smoothing: For noisy data, such as network traffic data, methods such as moving average and exponential smoothing can be used to smooth the data and reduce the impact of noise.

[0082] In the second scenario, network traffic information is input into a multilayer perceptron for feature extraction processing, resulting in the first network traffic feature output by the multilayer perceptron.

[0083] Among them, a multilayer perceptron usually refers to a processing module specifically used to parse and transform input data, which can analyze information such as network traffic patterns, behaviors and trends.

[0084] In the embodiments of this application, see also Figure 4As shown, network traffic information is input into a multilayer perceptron for feature extraction. The multilayer perceptron analyzes the network traffic data to extract key features, such as time-varying traffic patterns, peak periods, and bandwidth utilization. At this stage, the multilayer perceptron's task is to transform the network traffic information into high-dimensional data that reflects the traffic characteristics, obtaining the first network traffic feature. This feature provides the basis for subsequent operations such as traffic control and bandwidth allocation.

[0085] In the third scenario, the temporal information is input into the Long Short-Term Memory (LSTM) network for feature extraction, resulting in the first temporal feature output by the LTM network.

[0086] Long Short-Term Memory (LSTM) networks are a special type of recurrent neural network that can effectively process time-series data and capture long-term dependencies in the data.

[0087] In the embodiments of this application, see also Figure 4 As shown, temporal information is input into a Long Short-Term Memory (LSTM) network for feature extraction. In this stage, temporal information (such as historical data on network load, traffic changes, and container resource changes) is fed into the LTM network. The LTM network, through its memory units, analyzes the temporal dependencies of the data and extracts temporal features, such as trend predictions, periodic changes, and sudden events. These temporal features help the system understand the evolution of network states and future trends, ultimately outputting the first temporal feature.

[0088] In the above embodiments, by inputting resource demand information, network traffic information, and time series information into different neural network models for processing, the core features of various types of information are gradually extracted. Through the collaborative work of modules such as multilayer perceptrons, compilers, and long short-term memory networks, key features of the network state can be mined from multiple dimensions, providing comprehensive data support for subsequent decision-making.

[0089] In one exemplary embodiment, such as Figure 5 As shown, the sub-fusion model also includes a feature enhancement model. Based on this, the first-level sub-fusion model is used to perform feature enhancement processing on the network state information to obtain the first fused feature, including steps 401 to 403. Wherein:

[0090] Step 401: The first network traffic feature is subjected to max pooling and average pooling using the feature enhancement model. The processed first network traffic feature is then subjected to concatenation, convolution, and mapping processes to obtain the first initial information feature.

[0091] The goal of the feature enhancement model is to enhance the expressiveness of network traffic features through different processing methods, so as to provide richer information for subsequent calculations.

[0092] In this embodiment, the first network traffic feature is processed by max pooling and average pooling using a feature enhancement model. Max pooling helps extract the most salient information from the feature, thus preserving the strongest network traffic pattern; while average pooling smooths the feature by taking the average value, eliminating noise and preserving global features. These two pooling methods help the feature enhancement model understand the details and overall trends of network traffic at different levels.

[0093] After max pooling and average pooling, the pooling results undergo concatenation, convolution, and mapping. The concatenation process combines the feature vectors from max and average pooling to generate a high-dimensional feature vector that integrates information from both pooling methods. Next, the convolution process performs deep learning on the concatenated features to extract meaningful spatial features. Convolution effectively captures local patterns and complex relationships within the features. Finally, the mapping process maps the convolution results to the target feature space through a fully connected layer, obtaining the first initial information feature. This initial feature is a reinforced representation of the network traffic features, preparing to support subsequent computations.

[0094] In some embodiments, such as Figure 6 As shown, Figure 6 This is a diagram of the feature enhancement model structure. Figure 6 Secondary features This represents the first network traffic feature; `max pool` indicates max pooling; `avg pool` indicates average pooling; and `cat` indicates concatenation. This represents the first network traffic characteristic after max pooling. This represents the first network traffic characteristic after average pooling. This represents the first network traffic feature after concatenation; BatchNorm represents batch normalization. This represents the first network traffic feature after convolution processing. This represents the first network traffic feature after mapping, i.e., the first initial information feature. This indicates the first characteristic of resource demand. This represents the first intermediate information feature. Indicates the first fusion feature, This represents the first temporal feature.

[0095] This application embodiment utilizes a secondary feature extraction attention mechanism, inspired by the channel attention mechanism, to effectively extract features with minimal parameters, and uses a sigmoid function to transform the features into a weight map. Max pooling and average pooling are then used to obtain the features. and ,Will and Features obtained by splicing , will be through A convolutional layer consisting of 1x1 convolutions, BatchNorm, and 3x3 convolutions is obtained. Then, the secondary feature weight vector map is obtained through the sigmoid function. .

[0096] Step 402: Perform a dot product on the first initial information feature and the first resource demand feature to obtain the first intermediate information feature.

[0097] The dot product operation performs element-wise multiplication on two feature vectors, thereby fusing their information.

[0098] In this embodiment, the first initial information feature and the first resource demand feature are multiplied to obtain the first intermediate information feature. The multiplication of the resource demand feature and the network traffic feature can capture the relationship between resource demand and traffic status, providing more accurate input for subsequent resource optimization and traffic control.

[0099] In some embodiments, see Figure 6 As shown, the obtained and Dot product yields secondary feature weighted features And set the weight parameter to 0.3. and Adding them together gives The specific formula is as follows:

[0100]

[0101] for Feature extraction is performed using a temporal attention mechanism to obtain... .

[0102] Step 403: The first intermediate information feature and the first temporal feature are sequentially concatenated and convolved to obtain the first fused feature.

[0103] In this embodiment of the application, the first intermediate information feature and the first temporal feature are sequentially spliced ​​and convolved to obtain the first fused feature.

[0104] First, intermediate information features and temporal features are concatenated to fuse multidimensional features. Then, convolutional processing is used to further learn the concatenated features, extracting the deep-seated correlation between temporal changes and network state. After these processes, the first fused feature is finally obtained.

[0105] In some embodiments, see Figure 6 As shown, and The components are concatenated and then processed using a 1x1 convolution to obtain... The first fusion feature of the same size .

[0106] In the above embodiments, the expressive power of network traffic features is enhanced by max pooling and average pooling, and resource demand and traffic features are fused through dot product operations. Then, temporal information is combined with intermediate features through concatenation and convolution processing to finally obtain a comprehensive and expressive first fused feature. This feature not only integrates multi-dimensional information, but also captures the complex interrelationships between them through deep processing, providing a more accurate basis for subsequent optimization, prediction, or decision-making.

[0107] In an exemplary embodiment, the previous fusion feature includes the previous enhancement feature, the previous network traffic feature, and the previous time-series feature. Feature extraction processing of the previous fusion feature is performed using the current-level sub-fusion model, including:

[0108] In the first scenario, the previous enhanced feature is input into the encoder for feature extraction, resulting in the current level enhanced feature output by the encoder.

[0109] In this embodiment, the previous enhanced feature is input into the encoder for feature extraction. The encoder performs in-depth processing on the previous enhanced feature, extracting a more refined and meaningful feature representation. By performing nonlinear transformations on the features, the encoder captures more structural relationships or latent patterns, thereby generating local enhanced features.

[0110] In the second scenario, the network traffic features from the previous layer are input into the multilayer perceptron for feature extraction, resulting in the network traffic features of the current layer output by the multilayer perceptron.

[0111] In this embodiment, the previous network traffic features are input into a multilayer perceptron for feature extraction. The multilayer perceptron performs deep learning on the network traffic features to extract important patterns and features from the traffic, thereby generating current-level network traffic features.

[0112] Scenario 3: Input the previous temporal feature into the Long Short-Term Memory (LSTM) network for feature extraction to obtain the current temporal feature output by the LSTM network.

[0113] In this embodiment, the previous temporal feature is input into a Long Short-Term Memory (LSTM) network for feature extraction. The LTM network analyzes the dynamic changes in temporal information and extracts time-related trends, periodic changes, and sudden events. After processing by the LTM network, the current-level temporal feature is obtained.

[0114] In the above embodiments, the processing results from the previous stage are input into the encoder, multilayer perceptron, and long short-term memory network to further extract features from the augmented features, network traffic features, and temporal features, respectively. Each of these processing modules leverages its strengths: the encoder uses deep learning to abstract the complex relationships of the augmented features; the multilayer perceptron focuses on the nonlinear mapping of traffic features; and the long short-term memory network extracts long-term trends and dynamic dependencies from the temporal data. The resulting augmented features, network traffic features, and temporal features provide richer and more profound feature representations for subsequent decision-making and optimization tasks.

[0115] In one exemplary embodiment, such as Figure 7 As shown, the current-level sub-fusion model is used to perform feature enhancement processing on the previous fusion feature to obtain the current-level fusion feature, including steps 501 to 503. Wherein:

[0116] Step 501: Use the feature enhancement model to perform max pooling and average pooling on the network traffic features of this level, and then perform concatenation, convolution and mapping on the processed network traffic features of this level to obtain the initial information features of this level.

[0117] In this embodiment, the local network traffic features are input into a feature enhancement model for processing. The goal of the feature enhancement model is to extract more expressive information from the input network traffic features. The feature enhancement model first performs max pooling and average pooling on the network traffic features. Max pooling extracts the most salient parts of the features, preserving high-intensity signals in the network traffic features, while average pooling helps smooth the features, remove noise, and maintain global information. Through these two pooling processes, the feature enhancement model can understand the traffic features from different perspectives, thereby obtaining more comprehensive information.

[0118] Step 502: Perform a dot product on the initial information features and resource demand features of this level to obtain the intermediate information features of this level.

[0119] In this embodiment, the pooled network traffic features are further transformed through concatenation, convolution, and mapping. Concatenation merges the feature vectors from max pooling and average pooling to generate a high-dimensional feature representation, combining the advantages of both pooling methods. Then, convolution extracts spatial information and local patterns from the features, further enhancing their expressive power. Finally, mapping maps the convolutional features to the target feature space, generating initial information features at this level.

[0120] Step 503: Perform concatenation and convolution processing on the intermediate information features and temporal features of this level in sequence to obtain the fused features of this level.

[0121] In this embodiment, the initial information features and resource demand features at this level are multiplied together to obtain the fused features at this level. The multiplication operation fuses these two types of features by multiplying the elements of the two vectors one by one and then adding the results. The multiplication of resource demand features and network traffic features captures the correlation between them.

[0122] Finally, the intermediate information features and temporal features at this level are sequentially concatenated and convolved. The concatenation operation merges the temporal and intermediate information features to form a multi-dimensional comprehensive feature. The temporal features capture time-related trends, while the intermediate information features provide interactive information about network traffic and resource demand; concatenation integrates this information. Next, the convolution process extracts the complex patterns between temporal and resource information by convolving the concatenated features, further enhancing the feature's expressiveness. Ultimately, after these processes, the fused feature at this level is obtained.

[0123] In the above embodiments, network traffic features are enhanced by max pooling and average pooling, and different features are gradually fused by combining dot multiplication and convolution processing to finally obtain a comprehensive local fusion feature. This feature integrates the deep correlation between traffic, resource demand and time series information, and can provide accurate basis for subsequent decision-making, resource optimization and traffic control.

[0124] In one exemplary embodiment, such as Figure 8 As shown, the multi-information decoding model includes a decoder, a first fully connected neural network, and a second fully connected neural network. Based on this, the pre-constructed multi-information decoding model is used to sequentially decode and map the multi-information fusion features to obtain the resource allocation results and flow control results for each container, including steps 601 to 603. Wherein:

[0125] Step 601: Input the multi-information fusion features into the decoder to obtain the decoded features output by the decoder.

[0126] Among them, the decoded features are processed representations with higher semantic information, which can reflect key patterns of network and resource status.

[0127] In this embodiment, the multi-information fusion features are input to the decoder for processing. The decoder's role is to extract more expressive decoding features from the fusion features. The decoder performs feature extraction and mapping on the input multi-information fusion features to generate decoding features.

[0128] Step 602: Input the decoded features into the first fully connected neural network, and map the decoded features into the parameter space related to resource allocation through the first fully connected neural network to obtain the resource allocation result for each container.

[0129] In deep learning, the first fully connected neural network is used to map input features to a specific output space.

[0130] In this embodiment, the decoded features are input into a first fully connected neural network. The first fully connected neural network maps the decoded features into a parameter space related to resource allocation. This mapping process can calculate the resource allocation required for each container based on network and resource state information, such as calculating the computing power, memory, and bandwidth required for each container.

[0131] Step 603: Input the decoded features into the second fully connected neural network, and map the decoded features into the parameter space related to flow control through the second fully connected neural network to obtain the flow control result for each container.

[0132] In this embodiment, the decoded features are further input into a second fully connected neural network. The task of the second fully connected neural network is to map the decoded features into a parameter space related to flow control. Based on the flow patterns and demands in the decoded features, the second fully connected neural network calculates the flow control strategy required for each container, such as bandwidth allocation and latency control.

[0133] Through independent processing by these two neural networks, two key results can be obtained for each container: resource allocation and flow control. These two results ensure that the system allocates resources reasonably among multiple containers, while guaranteeing the efficiency and stability of flow control.

[0134] In some embodiments, the main decoding part of the decoder directly uses the decoder structure of the transformer to obtain the decoding features. The decoded features are passed through two fully connected layers to output two results:

[0135] One of them The prediction of how resources are allocated, i.e. the resource allocation result, is obtained by mapping the decoded features to the relevant parameter space of resource allocation through a fully connected layer. The prediction result of resource allocation is mainly to determine how to optimally allocate computing resources such as CPU, memory, storage resources and network bandwidth.

[0136] Another one is The predicted results of the flow control strategy, i.e., the flow control results for each container. It uses another fully connected layer to obtain a specific flow control strategy. The strategy is mainly based on the decoding characteristics and the real-time state of the network to determine the data output flow of each container.

[0137] In the above embodiments, key information in the fusion features is extracted by the decoder, and then this information is mapped to the parameter space of resource allocation and flow control by two fully connected neural networks, respectively, and finally, accurate resource allocation and flow control strategies are generated for each container.

[0138] In one exemplary embodiment, such as Figure 9 As shown, the above method further includes steps 701 to 702. Wherein:

[0139] Step 701: Based on the resource allocation results and flow control results, determine the resource allocation loss value, service quality loss value, and flow control loss value.

[0140] The resource allocation loss value is used to measure the accuracy of the model in resource allocation. It is calculated by comparing the deviation between the resource allocation calculated by the model and the actual resources needed.

[0141] The quality of service (QoS) loss value reflects the model's ability to guarantee service quality, and is usually calculated by evaluating metrics such as system response time, throughput, and latency.

[0142] Flow control loss measures the accuracy of a system in traffic scheduling and bandwidth management. Flow control loss is determined by comparing the actual traffic allocation with the expected flow control strategy.

[0143] In this embodiment of the application, resource allocation loss value, service quality loss value, and flow control loss value are determined based on resource allocation results and flow control results.

[0144] Step 702: Based on the resource allocation loss value, service quality loss value, and flow control loss value, the parameters of the multi-information fusion model are adjusted to optimize the multi-information fusion model.

[0145] In this embodiment, the multi-information fusion model is optimized by adjusting the parameters based on the resource allocation loss value, service quality loss value, and traffic control loss value.

[0146] In some embodiments, the loss value may further include:

[0147] Resource allocation loss value: This application embodiment will consider the resource allocation loss value from four aspects: CPU, memory, storage resources, and network bandwidth.

[0148] CPU allocation loss:

[0149]

[0150] in This represents the actual value allocated by the CPU. The predicted value representing CPU allocation; This indicates CPU allocation loss.

[0151] Memory allocation penalty:

[0152]

[0153] in Represents the actual value of memory allocation. Represents the predicted value of memory allocation; This represents a loss in memory allocation.

[0154] Storage resource allocation loss:

[0155]

[0156] in Represents the actual value of memory allocation. Represents the predicted value of memory allocation. This indicates a loss in storage resource allocation.

[0157] Network bandwidth allocation loss:

[0158]

[0159] in Represents the actual value of memory allocation. Represents the predicted value of memory allocation. This indicates a loss in network bandwidth allocation.

[0160] The resource allocation loss value is:

[0161]

[0162] λ can be set according to the application's preferences; for example, if more emphasis is placed on CPU usage, λ can be set to 1. The coefficient was adjusted to be relatively large.

[0163] The service quality loss component is mainly based on the specific system's service quality requirements and is also an indicator for evaluating the overall network capability, as well as the model's evaluation metric. This application's embodiment will evaluate the predicted results based on simulation tools. This evaluation does not rely on actual labels and is more like a semi-supervised model evaluation and loss value. Specifically, regarding this evaluation system: the following are some evaluation metric rules specified for multi-container network management:

[0164] Resource utilization rate:

[0165] CPU utilization: Average CPU utilization should be maintained between 70% and 90% to avoid excessive idleness or overload. If CPU utilization is below 70%, then loss = If CPU utilization is greater than 90%, then loss = .

[0166] Memory utilization: Average memory utilization should be within a reasonable range of 60%-85%. If memory utilization is below 60%, then the loss = If CPU utilization is greater than 85%, then loss = .

[0167] Storage utilization: Storage usage should not exceed 80% of the total capacity to reserve some expansion space. If it exceeds 80%, then the loss = .

[0168] By combining the losses obtained above with appropriate weights, we obtain the resource utilization loss function. .

[0169] Network performance:

[0170] Network latency: Average network latency should be less than 100 milliseconds, and latency for critical applications should be less than 50 milliseconds. Network latency loss function:

[0171]

[0172] Packet loss rate: The overall packet loss rate should be below 1%, and for applications with high real-time requirements, it should be below 0.1%. Packet loss rate loss function:

[0173]

[0174] Bandwidth: Ensure each container receives the predetermined minimum bandwidth, with average bandwidth utilization between 60% and 80%. If bandwidth utilization falls below 60%, then loss = If the bandwidth utilization is greater than 80%, then loss = .

[0175] By concatenating the losses obtained above using appropriate weights, we obtain the network performance loss function. .

[0176] Service availability:

[0177] Service uptime should reach 99.9% or higher, meaning annual downtime should not exceed a threshold of T hours. Service uptime is often difficult to represent directly using a loss function, but it can be indirectly assessed by monitoring the difference between actual uptime and the target uptime.

[0178] .

[0179] Response time:

[0180] For common requests, the average response time should be less than 2 seconds. For urgent or critical requests, the response time should be less than 500 milliseconds. Let the ideal response time be... Then the loss function is:

[0181]

[0182] Flow control accuracy:

[0183] The deviation between the actual flow rate and the set flow rate should be within 10%. Let the ideal deviation be... Then the loss function is:

[0184] .

[0185] Fairness in resource allocation:

[0186] When resources are insufficient, containers of different priorities are allocated resources according to their priority, and low-priority containers should not be completely deprived of resources. In this application embodiment, the difference in resource allocation between containers of the same type is considered as a loss value; assuming the expected value for containers of the same priority is set to... Each container acquires resources as follows: The loss function is:

[0187]

[0188] Fault recovery capability:

[0189] The target time from the occurrence of a failure to the restoration of normal service should be within 5 minutes. The amount of data loss during the failure should be kept within an acceptable range; let the acceptable range be... The loss value is rounded up during the calculation process. Therefore, the loss function is:

[0190]

[0191] Scalability:

[0192] When adding new containers or services, resource allocation and traffic control can be performed without significantly impacting the QoS of existing services. Let the loss function be:

[0193]

[0194] Energy efficiency:

[0195] To minimize overall energy consumption, the target minimum energy consumption is set as follows: The loss function is then:

[0196]

[0197] Cost-effectiveness:

[0198] Taking into account both resource usage cost and service quality, this application aims to ensure the optimal QoS level is achieved within a given budget. Assuming resource usage cost is C, the service quality score can be calculated comprehensively based on specific QoS indicators. This embodiment uses principal component analysis to calculate QoS. By analyzing multiple QoS indicators, the principal components that explain the largest variance in the data are identified. These principal components are linear combinations of the original indicators and can be used as a comprehensive indicator. The calculated QoS indicator is Q, the budget is B, and the target service quality score threshold is [missing information]. The loss function is then:

[0199]

[0200] in, and The adjustment coefficients are used to weigh the losses caused by cost overruns and substandard service quality.

[0201] The above are relatively comprehensive service quality indicators established in the embodiments of this application. The loss function is calculated based on each indicator to obtain... arrive , set parameters as arrive The service quality loss function is:

[0202]

[0203] Considering that each deployment environment has different requirements for each metric of service quality, It can be fine-tuned according to the actual application scenario to better adapt to the application scenario.

[0204] By weighted and fused resource allocation loss, flow control loss, and service quality loss, the final loss function is obtained. This loss function can more comprehensively consider all aspects, thereby training a model that is more suitable for the application environment, and adjusting the model parameters accordingly.

[0205] In the above embodiments, parameter tuning and optimization based on resource allocation loss values, service quality loss values, and flow control loss values ​​is the core process for improving the performance of the multi-information fusion model. Through this process, the model can gradually reduce losses, improve the accuracy of resource allocation, optimize service quality, improve flow control strategies, and ultimately achieve a more efficient and stable system performance.

[0206] According to some embodiments of this application, a resource allocation and flow control method based on a multi-container network is provided. Taking the application of this method to a computer device as an example, it may include the following steps:

[0207] Step 1: Obtain network status information for the multi-container network. Network status information includes resource requirements, network traffic, and timing information.

[0208] Step 2 involves iteratively calculating the network state information using a multi-level sub-fusion model. In the first iteration, the network state information is input into the first-level sub-fusion model, which then sequentially performs feature extraction processing on the network state information. The sub-fusion model includes an encoder, a multilayer perceptron, a long short-term memory network, and a feature enhancement model.

[0209] In some embodiments, resource requirement information is input into the encoder for feature extraction processing to obtain the first resource requirement feature output by the encoder.

[0210] In some embodiments, network traffic information is input into a multilayer perceptron for feature extraction processing to obtain the first network traffic feature output by the multilayer perceptron.

[0211] In some embodiments, temporal information is input into a long short-term memory network for feature extraction to obtain the first temporal feature output by the long short-term memory network.

[0212] Step 3: Use the feature enhancement model to perform max pooling and average pooling on the first network traffic feature, and then perform concatenation, convolution and mapping on the processed first network traffic feature to obtain the first initial information feature.

[0213] Step 4: Perform a dot product on the first initial information feature and the first resource demand feature to obtain the first intermediate information feature.

[0214] Step 5: Perform concatenation and convolution on the first intermediate information feature and the first temporal feature in sequence to obtain the first fused feature.

[0215] In subsequent iterations, the previous fusion feature obtained from the previous sub-fusion model is input into the current sub-fusion model, and the current sub-fusion model is used to perform feature extraction processing on the previous fusion feature. The previous fusion feature includes the previous enhancement feature, the previous network traffic feature, and the previous time series feature.

[0216] In some embodiments, the previous enhanced feature is input into the encoder for feature extraction processing to obtain the current-level enhanced feature output by the encoder.

[0217] In some embodiments, the previous network traffic feature is input into a multilayer perceptron for feature extraction processing to obtain the current-level network traffic feature output by the multilayer perceptron.

[0218] In some embodiments, the previous temporal feature is input into a long short-term memory network for feature extraction, resulting in the current-level temporal feature output by the long short-term memory network.

[0219] Step 6: Use the feature enhancement model to perform max pooling and average pooling on the network traffic features of this level, and then perform concatenation, convolution and mapping on the processed network traffic features of this level to obtain the initial information features of this level.

[0220] Step 7: Perform a dot product on the initial information features and resource demand features of this level to obtain the intermediate information features of this level.

[0221] Step 8: Perform concatenation and convolution processing on the intermediate information features and temporal features of this level in sequence to obtain the fused features of this level.

[0222] Step 9: Determine the fusion features output by the final-level sub-fusion model as multi-information fusion features.

[0223] Step 10: Input the multi-information fusion features into the decoder to obtain the decoded features output by the decoder.

[0224] Step 11: Input the decoded features into the first fully connected neural network. The first fully connected neural network maps the decoded features into the parameter space related to resource allocation to obtain the resource allocation result for each container.

[0225] Step 12: Input the decoded features into the second fully connected neural network. The second fully connected neural network maps the decoded features into the parameter space related to flow control to obtain the flow control result for each container.

[0226] Step 13: Based on the resource allocation results and flow control results, determine the resource allocation loss value, service quality loss value, and flow control loss value.

[0227] Step 14: Based on the resource allocation loss value, service quality loss value, and flow control loss value, perform parameter tuning on the multi-information fusion model to optimize it.

[0228] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0229] Based on the same inventive concept, this application also provides a resource allocation and flow control device for implementing the resource allocation and flow control method for multi-container networks described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the resource allocation and flow control device for multi-container networks provided below can be found in the limitations of the resource allocation and flow control method for multi-container networks described above, and will not be repeated here.

[0230] In one exemplary embodiment, such as Figure 10 As shown, a resource allocation and traffic control device based on a multi-container network is provided, including: an information acquisition module 801, an enhancement processing module 802, and a mapping processing module 803, wherein:

[0231] Information acquisition module 801 is used to acquire network status information of multi-container network;

[0232] The enhancement processing module 802 is used to sequentially perform feature extraction and feature enhancement processing on the network state information using a pre-built multi-information fusion model to obtain multi-information fusion features.

[0233] The mapping processing module 803 is used to perform decoding and mapping processing on the multi-information fusion features in sequence using a pre-built multi-information decoding model to obtain the resource allocation results and flow control results for each container.

[0234] In an exemplary embodiment, the enhancement processing module 802 is specifically used to iteratively calculate network state information using a multi-level sub-fusion model. In the first iteration, the network state information is input into the first-level sub-fusion model, and the network state information is sequentially processed by feature extraction and feature enhancement to obtain the first fused feature. In subsequent iterations, the previous fused feature obtained from the previous-level sub-fusion model is input into the current-level sub-fusion model, and the previous fused feature is sequentially processed by feature extraction and feature enhancement to obtain the current-level fused feature. The fused feature output by the final-level sub-fusion model is determined as the multi-information fusion feature.

[0235] In an exemplary embodiment, the above-mentioned enhancement processing module 802 is specifically used to input resource demand information into the encoder for feature extraction processing to obtain the first resource demand feature output by the encoder; input network traffic information into the multilayer perceptron for feature extraction processing to obtain the first network traffic feature output by the multilayer perceptron; and input temporal information into the long short-term memory network for feature extraction to obtain the first temporal feature output by the long short-term memory network.

[0236] In an exemplary embodiment, the enhancement processing module 802 is specifically used to perform max pooling and average pooling on the first network traffic feature using the feature enhancement model, and then perform concatenation, convolution, and mapping on the processed first network traffic feature to obtain the first initial information feature; perform dot product on the first initial information feature and the first resource demand feature to obtain the first intermediate information feature; and perform concatenation and convolution on the first intermediate information feature and the first time series feature to obtain the first fused feature.

[0237] In one exemplary embodiment, the above-described apparatus may further include:

[0238] The loss value determination module 804 is used to determine the resource allocation loss value, service quality loss value, and flow control loss value based on the resource allocation result and the flow control result.

[0239] The optimization module 805 is used to perform parameter tuning on the multi-information fusion model based on resource allocation loss value, service quality loss value, and flow control loss value, so as to optimize the multi-information fusion model.

[0240] The modules in the aforementioned resource allocation and flow control device based on multi-container networks can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can invoke and execute the corresponding operations of each module.

[0241] According to some embodiments of this application, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory including instructions that can be executed by a processor of an electronic device to perform the above-described method. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0242] According to some embodiments of this application, a computer program product is also provided, which, when executed by a processor, can implement the above-described methods. The computer program product includes one or more computer instructions. When these computer instructions are loaded and executed on a computer, some or all of the above-described methods can be implemented, wholly or partially, according to the processes or functions described in the embodiments of this application.

[0243] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0244] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0245] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A resource allocation and flow control method based on multi-container networks, characterized in that, The method includes: Obtain network status information for a multi-container network; The network state information is sequentially processed by feature extraction and feature enhancement using a pre-built multi-information fusion model to obtain multi-information fusion features. The multi-information fusion features are sequentially decoded and mapped using a pre-built multi-information decoding model to obtain the resource allocation results and flow control results for each container. The multi-information decoding model includes a decoder, a first fully connected neural network, and a second fully connected neural network. The pre-constructed multi-information decoding model is used to sequentially decode and map the multi-information fusion features to obtain the resource allocation and flow control results for each container, including: The multi-information fusion features are input into the decoder to obtain the decoded features output by the decoder; The decoded features are input into the first fully connected neural network, and the first fully connected neural network maps the decoded features to the parameter space related to resource allocation to obtain the resource allocation result for each container; The decoded features are input into the second fully connected neural network, which maps the decoded features to a parameter space related to flow control, thereby obtaining the flow control result for each container.

2. The method according to claim 1, characterized in that, The multi-information fusion model includes a multi-level sub-fusion model. The pre-constructed multi-information fusion model is used to sequentially perform feature extraction and feature enhancement processing on the network state information to obtain multi-information fusion features, including: The network state information is iteratively calculated using the multi-level sub-fusion model. In the first iteration, the network state information is input into the first-level sub-fusion model. The first-level sub-fusion model is then used to sequentially perform feature extraction and feature enhancement processing on the network state information to obtain the first fused feature. In non-first iterations, the previous fusion feature obtained from the previous sub-fusion model is input into the current sub-fusion model. The current sub-fusion model is then used to sequentially perform feature extraction and feature enhancement processing on the previous fusion feature to obtain the current fusion feature. The fusion features output by the final-level sub-fusion model are determined as the multi-information fusion features.

3. The method according to claim 2, characterized in that, The network state information includes resource demand information, network traffic information, and time series information. The sub-fusion model includes an encoder, a multilayer perceptron, and a long short-term memory network. The feature extraction processing of the network state information using the first-level sub-fusion model includes: The resource requirement information is input into the encoder for feature extraction processing to obtain the first resource requirement feature output by the encoder; The network traffic information is input into the multilayer perceptron for feature extraction processing to obtain the first network traffic feature output by the multilayer perceptron; The time-series information is input into the Long Short-Term Memory (LSTM) network for feature extraction, resulting in the first time-series feature output by the LTM network.

4. The method according to claim 3, characterized in that, The sub-fusion model further includes a feature enhancement model, wherein the network state information is enhanced using the first-level sub-fusion model to obtain the first fused feature, including: The first network traffic feature is subjected to max pooling and average pooling using the feature enhancement model, and then the processed first network traffic feature is sequentially concatenated, convolved, and mapped to obtain the first initial information feature. Perform a dot product operation on the first initial information feature and the first resource requirement feature to obtain the first intermediate information feature; The first intermediate information feature and the first temporal feature are sequentially concatenated and convolved to obtain the first fused feature.

5. The method according to claim 4, characterized in that, The previous fusion feature includes the previous enhanced feature, the previous network traffic feature, and the previous time-series feature. The feature extraction process for the previous fusion feature using the current-level sub-fusion model includes: The previous enhanced feature is input into the encoder for feature extraction processing to obtain the current-level enhanced feature output by the encoder; The previous network traffic feature is input into the multilayer perceptron for feature extraction processing to obtain the current level network traffic feature output by the multilayer perceptron; The previous temporal feature is input into the Long Short-Term Memory (LSTM) network for feature extraction, resulting in the current temporal feature output by the LSTM network.

6. The method according to claim 5, characterized in that, The step of using the current-level sub-fusion model to perform feature enhancement processing on the previous fusion feature to obtain the current-level fusion feature includes: The feature enhancement model is used to perform max pooling and average pooling on the network traffic features of this level, and then the processed network traffic features of this level are sequentially concatenated, convolved and mapped to obtain the initial information features of this level. The initial information features and resource demand features of this level are multiplied by a dot product to obtain the intermediate information features of this level. The intermediate information features and the temporal features of this level are sequentially concatenated and convolved to obtain the fused features of this level.

7. The method according to any one of claims 1 to 6, characterized in that, The method further includes: Based on the resource allocation results and the flow control results, determine the resource allocation loss value, the service quality loss value, and the flow control loss value; The multi-information fusion model is optimized by adjusting the parameters based on the resource allocation loss value, service quality loss value, and flow control loss value.

8. A resource allocation and flow control device based on a multi-container network, characterized in that, The device includes: The information acquisition module is used to acquire network status information of the multi-container network; An enhancement processing module is used to sequentially perform feature extraction and feature enhancement processing on the network state information using a pre-built multi-information fusion model to obtain multi-information fusion features. The mapping processing module is used to sequentially decode and map the multi-information fusion features using a pre-built multi-information decoding model to obtain the resource allocation results and flow control results for each container. The multi-information decoding model includes a decoder, a first fully connected neural network, and a second fully connected neural network. The mapping processing module is specifically used to input the multi-information fusion features into the decoder to obtain the decoded features output by the decoder. The decoded features are input into the first fully connected neural network, and the first fully connected neural network maps the decoded features to the parameter space related to resource allocation to obtain the resource allocation result for each container; The decoded features are input into the second fully connected neural network, which maps the decoded features to a parameter space related to flow control, thereby obtaining the flow control result for each container.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

Images