Cloud server virtual machine management method and system

Abstract

The invention relates to a cloud server virtual machine management method and system, and belongs to the technical field of telecommunication. The cloud server virtual machine management method comprises the following steps: acquiring historical user behavior data; acquiring current business demand data; obtaining cloud server resource state data; inputting the user behavior pattern feature vector, the business demand feature vector and the resource state vector into a prediction model, and predicting the resource quantity required by each virtual machine in a future period of time; outputting a resource allocation scheme by using a resource allocation algorithm according to the predicted resource quantity and the current actual allocable resources of the cloud server; according to the method, the resource utilization rate can be remarkably increased, the service response speed is increased, the service stability is enhanced, green and energy-saving effects are achieved, the adaptive capacity is achieved, diversified scene requirements are met, and the management effect of the virtual machine of the cloud server is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of electrical communication technology, and more specifically, to a cloud server virtual machine management method and system. Background Art

[0002] With the development of cloud computing, cloud server virtual machines are widely used, but there are problems with current management: Disadvantages of static resource allocation: Traditionally, static allocation strategies are used to pre-allocate fixed resources for virtual machines based on experience or estimation. However, business loads and user behaviors change dynamically. For example, when e-commerce promotions are underway, resource demand increases significantly, and static allocation cannot meet the demand, resulting in slow business response or even interruption. In addition, different business types have different characteristics for resource demand, and static allocation can easily lead to waste or insufficient resources.

[0003] Insufficient risk assessment and early warning: The existing management system lacks effective prediction and response mechanisms for business risks caused by resource allocation, such as system crashes and business inability to operate caused by excessive or insufficient resource allocation. It can only handle them passively, causing losses to cloud service providers and users. Summary of the invention

[0004] The present invention provides a cloud server virtual machine management method and system to solve the problem that the traditional statically allocated resources in the related art cannot meet such sudden high demand, resulting in slow business response, even service interruption and resource waste or insufficient resources.

[0005] The present invention provides a cloud server virtual machine management method, comprising: Obtain historical user behavior data and use the time series processing model to extract user behavior pattern feature vectors; Obtain current business demand data and use feature extraction models to extract business demand feature vectors; Obtain cloud server resource status data and perform standardization to obtain a resource status vector; Input the user behavior pattern feature vector, business demand feature vector and resource state vector into the prediction model to predict the amount of resources required by each virtual machine in the future period of time; According to the predicted resource quantity and the actual allocatable resources of the current cloud server, the resource allocation plan is output through the resource allocation algorithm.

[0006] Furthermore, the time series processing model is a long short-term memory network model; the feature extraction model is a convolutional neural network model; the resource state vector is obtained by a standardized function The prediction model is constructed by processing the cloud server resource status data; the prediction model is constructed by taking the user behavior pattern feature vector, the business demand feature vector and the resource status vector as input and inputting them into a multi-layer perceptron prediction model. In the example above, the resource requirements of each virtual machine in the future are predicted.

[0007] Furthermore, a cloud server virtual machine management method further includes: Obtain historical resource allocation failure case data and use decision tree algorithm to extract risk feature vectors; The risk feature vector is input into the improved prediction model together with the user behavior pattern feature vector, the business demand feature vector and the resource state vector. , predict the amount of resources and risk assessment value required by each virtual machine in the future based on historical resource allocation failure case data; Set a risk threshold. When the risk assessment value is greater than the risk threshold, a risk warning is issued and the resource allocation plan that has not undergone risk assessment is adjusted according to the risk type.

[0008] Furthermore, a cloud server virtual machine management method further includes: Obtain energy consumption data of each hardware component of the cloud server and use the energy consumption model Predict energy consumption under different resource allocation schemes; With the goal of satisfying business needs with resource requirements and minimizing energy consumption, an optimization function is constructed to solve the optimal resource allocation based on the energy consumption model; Substitute the optimal resource allocation amount based on the energy consumption model into the resource allocation algorithm and output the final resource allocation plan based on the energy consumption model.

[0009] Furthermore, a cloud server virtual machine management method further includes: Obtain historical resource allocation plans and their business operation effect feedback data, and use deep reinforcement learning algorithm analysis to obtain the adjustment direction and amplitude information of resource allocation strategy; The parameters of the improved prediction model are adjusted according to the adjustment direction and amplitude information to obtain an adjusted prediction model. ; The relevant feature vectors are input into the adjusted prediction model to predict the amount of resources based on historical data required for each virtual machine in the future period, and are substituted into the resource allocation algorithm to output the final resource allocation plan based on historical data. The resource allocation plan based on historical data and the corresponding business operation effect feedback data are then updated to the feedback data set for cyclic optimization.

[0010] Furthermore, a cloud server virtual machine management method further includes: Obtain detailed information about big data analysis tasks and use custom feature extraction algorithms to extract feature vectors of big data tasks; The big data task feature vector is merged with the user behavior pattern feature vector, the business demand feature vector and the resource state vector to obtain a merged feature vector; The fused feature vector is input into the adjusted prediction model adapted to the big data scenario , predict the amount of resources required by each virtual machine suitable for big data analysis tasks based on big data scenarios, and substitute them into the resource allocation algorithm to output the final resource allocation plan based on big data scenarios.

[0011] Furthermore, a cloud server virtual machine management method further includes: Obtain relevant information about containerized applications and use graph neural networks to extract container feature vectors; The container feature vector, the user behavior pattern feature vector, the business demand feature vector and the resource state vector are input into the adjusted prediction model combined with time series prediction to predict the amount of resources required for each container at different time points; According to the predicted container resource requirements at different time points and the dependencies between containers, an improved heuristic algorithm is used to allocate resources and output a resource allocation plan based on containerized applications.

[0012] Furthermore, a cloud server virtual machine management method further includes: Obtain multi-tenant information and use deep autoencoders to extract tenant feature vectors; The tenant feature vector, user behavior pattern feature vector, business demand feature vector and resource state vector are used as input to the personalized prediction model. , predicting the amount of resources needed to meet the personalized needs of each tenant; A resource allocation coordination algorithm based on game theory is adopted to fairly allocate resources while ensuring the basic gaming experience of each tenant, and a resource allocation plan based on game theory is output.

[0013] The present invention provides a cloud server virtual machine management system, comprising: Data acquisition module, used to obtain historical user behavior data, current business demand data, and cloud server resource status data; A feature extraction module is used to extract a user behavior pattern feature vector using a time series processing model, and to extract a business demand feature vector using a feature extraction model; A data processing module is used to perform standardization processing on the cloud server resource status data to obtain a resource status vector; A prediction module, used to input the user behavior pattern feature vector, the business demand feature vector and the resource state vector into the adjusted personalized prediction model to predict the amount of resources required by each virtual machine in the future; The resource allocation module is used to output a resource allocation plan using a resource allocation algorithm based on the predicted resource quantity and the actual allocatable resources of the current cloud server.

[0014] Furthermore, a cloud server virtual machine management system further includes: The risk assessment module is used to obtain historical resource allocation failure case data, extract risk feature vectors using a decision tree algorithm, and input the risk feature vectors together with other feature vectors into an improved prediction model to predict a risk assessment value. When the risk assessment value is greater than a set threshold, a risk warning is issued and a resource allocation plan based on the risk assessment is adjusted.

[0015] The beneficial effects of the present invention are: Resource management optimization: Adaptive learning and prediction based on user behavior and business needs, accurately matching resources to avoid over- or under-allocation. Timely and accurate allocation of resources to avoid slow business response due to insufficient resources, improve user experience, and ensure efficient business operation.

[0016] Enhanced risk prevention and control: Introduce risk assessment and early warning mechanisms, extract risk feature vectors by analyzing historical failure cases, assess risks in combination with prediction models, issue warnings when thresholds are exceeded and adjust plans to enhance business stability.

[0017] Green energy saving: Resource allocation takes energy consumption into consideration, builds an energy consumption model combined with an optimization algorithm, allocates resources with the goal of minimizing energy consumption, and reduces the overall energy consumption of cloud servers.

[0018] Improved system adaptability: Use deep reinforcement learning to automatically optimize and tune resource allocation plans, automatically adapt as business and user behavior evolve, and improve system adaptability and robustness.

[0019] Adaptation to special scenarios: For special scenarios such as big data analysis, containerized application deployment, and multi-tenant cloud games, targeted resource allocation optimization methods are proposed to meet the needs of special scenarios and enhance service competitiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a flow chart of a cloud server virtual machine management method of the present invention; Figure 2 is an example diagram of user behavior data record of the present invention; Figure 3 is the user behavior feature vector of the present invention; Figure 4 is the business demand data graph of the present invention; Figure 5 is the business demand feature vector diagram of the present invention; Figure 6is a resource status data graph of the present invention; Figure 7 is a resource demand prediction result diagram of the present invention; Figure 8 is a diagram of a resource allocation scheme of the present invention; Fig. 9 It is a comparison diagram of the technical effects of the present invention. DETAILED DESCRIPTION

[0021] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the present specification. Various examples may omit, replace, or add various processes or components as needed. In addition, the features described in some examples may also be combined in other examples.

[0022] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings 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.

[0023] A cloud server virtual machine management method, such as Figure 1 As shown, the following steps are included: Step 101, behavior pattern extraction: obtaining historical user behavior data , using the Long Short-Term Memory (LSTM) model Process it and output the user behavior pattern feature vector LSTM is a recurrent neural network that can effectively process and predict time series data. In this embodiment, it is used to extract the time series features of user behavior.

[0024] In step 101, the LSTM model The specific structure is as follows: Input layer: receives user behavior time series data, including login times, CPU usage, memory usage, storage read and write volume, and other features; LSTM layer: contains multiple LSTM units, each unit has a dimension of 128 and 2 layers; Fully connected layer: converts the output of the LSTM layer into a feature vector of fixed dimension; Output layer: outputs 3D user behavior feature vector; Step 102, business demand analysis: obtain current business demand data , using the convolutional neural network (CNN) model Extract features and output business demand feature vector Among them, CNN can automatically extract the spatial features of data through convolution and pooling operations, and is suitable for feature extraction of business demand data (such as business type, business scale, etc.). The mathematical expression of CNN is: ; ; in, , , is the number of layers, rows, and columns of the feature map, For the Tier feature map, For the Tier The feature map and The convolution kernel between feature maps, For the Tier The bias term of the feature map, For the The feature map of the previous layer connected to the feature map, represents the convolution operation, is the activation function (such as ReLU), is the receptive field of the pooling operation, For the Tier feature map, For the Tier Finally, the output layer result of CNN is used as the business requirement feature vector .

[0025] In step 102, the CNN model The specific structure is as follows: Input layer: receives business demand data and organizes the data into a two-dimensional matrix form; Convolutional layer 1: uses 32 3×3 convolution kernels, stride 1, and padding mode 'same'; Pooling layer 1: Use 2×2 maximum pooling with a stride of 2; Convolutional layer 2: uses 64 3×3 convolution kernels, stride 1, and padding mode 'same'; Pooling layer 2: Use 2×2 maximum pooling with a step size of 2. Fully connected layer: Flatten the feature map and connect it to a fully connected layer with 512 neurons. Output layer: outputs 3D business demand feature vector; Step 103, resource status assessment: obtain cloud server resource status data , using the standardization function Process it to get the standardized resource state vector The normalization function can map resource status data of different dimensions (such as CPU usage, memory usage, etc.) to the same scale, which is convenient for subsequent prediction model processing. Commonly used normalization functions include maximum and minimum value normalization and Z-score normalization, and their mathematical expressions are: ; ; in, and are the minimum and maximum values ​​of the data, respectively. For data, and are the mean and standard deviation of the data respectively.

[0026] Step 104, prediction model construction: , and As input, it is fed into a multi-layer perceptron (MLP) prediction model In the future, the amount of resources required by each virtual machine is predicted MLP is a feedforward neural network consisting of an input layer, a hidden layer, and an output layer, which can fit complex nonlinear mapping relationships. The mathematical expression of MLP is: ; ; ; in, , , is the input feature, For the The hidden state of the layer, and For the The weight matrix and bias vector of the layer, For the the activation function of the layer (such as ReLU), is the number of layers in the network. For the The hidden state of the layer. is the input layer, is the output layer. is the weight matrix from the input layer to the hidden layer, is the weight matrix from the hidden layer to the output layer, is the bias vector from the input layer to the hidden layer, is the bias vector from the hidden layer to the output layer.

[0027] In step 104, the MLP prediction model The specific structure is as follows: Input layer: receives 9-dimensional input features (3-dimensional user behavior features, 3-dimensional business demand features, and 3-dimensional resource status features); Hidden layer 1: 256 neurons, using RelU activation function; Hidden layer 2: 128 neurons, using ReLU activation function; Hidden layer 3: 64 neurons, using ReLU activation function; Output layer: 3 neurons, corresponding to the predicted resource amounts of CPU, memory, and storage respectively; Step 105, resource allocation decision: based on the predicted resource quantity , combined with the actual allocatable resources of the current cloud server , use the resource allocation algorithm to allocate resources and output the resource allocation plan The basic idea of ​​the resource allocation algorithm is to take the best decision under the current state in each step of selection, so that the final result is optimal or close to optimal. In this embodiment, the resource allocation algorithm is sorted according to the resource demand of the virtual machine, and the virtual machine with large demand is given priority until the cloud server resources are allocated.

[0028] In one embodiment of the present invention, the resource allocation scheme includes: CPU allocation, memory allocation and storage allocation.

[0029] In step 105, the resource allocation algorithm The specific implementation steps are as follows: sort the virtual machines in descending order according to the resource demand, and allocate resources to each virtual machine in turn. The allocation rule is: if the remaining resources are sufficient, the predicted resource demand is fully met. If the remaining resources are insufficient, the actual allocation amount is calculated according to the following formula: ; in is the resource allocation coefficient, with a value range of [0,1], which is used to control the conservativeness of resource allocation. The value of is obtained by statistical analysis based on historical resource allocation data. The remaining available resources are updated, and steps 102 to 103 are repeated until all virtual machines have completed resource allocation.

[0030] The specific parameter settings of these models and algorithms can be adjusted according to the actual application scenarios to obtain the best resource allocation effect. For example, the number of neural network layers, number of neurons, learning rate and other parameters can be optimized through cross-validation or the resource allocation coefficient in the resource allocation algorithm can be adjusted. .

[0031] Through the above steps, this embodiment provides a cloud server virtual machine resource intelligent allocation method based on user behavior and business demand adaptive learning and prediction, which can dynamically adapt to complex and changing scenarios and improve resource utilization and user experience.

[0032] In order to more intuitively illustrate the specific implementation process of this embodiment, a practical application example is given below: There are three virtual machine instances (VM1, VM2, VM3) on a cloud service platform, and resources need to be allocated to them.

[0033] Example of historical user behavior data ( ),like Figure 2 User behavior data record example is shown in the figure; After being processed by the LSTM model, the user behavior pattern feature vector is obtained ,like Figure 3 User behavior feature vector is shown; Example of current business demand data ( ),like Figure 4 As shown in the business demand data diagram; After being processed by the CNN model, the business demand feature vector is obtained ,like Figure 5 It is shown in the business demand feature vector diagram; Example of cloud server resource status data ( ),like Figure 6 As shown in the resource status data diagram; after standardization, the resource status vector is obtained : [0.375, 0.375, 0.4]; The resource demand forecast results output by the forecast model are as follows: Figure 7 The resource demand forecast results are shown in the figure; Finally, we get the resource allocation scheme, such as Figure 8 The resource allocation scheme is shown in the figure; Technical effect verification: The resource allocation scheme of this embodiment is compared with the traditional static allocation scheme. The test period is 7 days. The results are as follows: Fig. 9 This is shown in the technical effect comparison chart; Depend on Figures 2 to 9 It can be seen that the resource allocation scheme of this embodiment has significant advantages over the traditional static allocation scheme: resource utilization is increased by 72.1%, which significantly improves resource utilization efficiency, reduces business response time by 27.8%, improves user experience, increases SLA achievement rate by 6.7%, and enhances service quality assurance. Through dynamic adjustment, it can respond to business changes more flexibly, and energy consumption is reduced by 24.1%, achieving the effect of energy conservation and emission reduction.

[0034] These data fully demonstrate the effectiveness and advancement of the cloud server virtual machine resource intelligent allocation method based on adaptive learning and prediction of user behavior and business needs proposed in this embodiment.

[0035] In one embodiment of the present invention, in order to introduce a risk assessment and early warning mechanism for resource allocation and improve business stability, a cloud server virtual machine management method further includes the following steps: Step 201 to step 203 are the same as step 101 to step 103 of embodiment 1.

[0036] Step 204, risk feature extraction: obtaining historical resource allocation failure case data , using the decision tree algorithm Process it and extract the risk feature vector . The decision tree divides the data set into different subsets by recursively selecting the optimal partitioning features until all subsets belong to the same category or the predetermined termination condition is reached. The generation algorithms of decision trees mainly include ID3, C4.5 and CART, among which the Gini index calculation formula of the CART algorithm is: ; ; in, For the dataset The Gini index, Features The Gini index, For the data set, For the categories, is the total number of categories, To divide the features, is the number of values ​​of this feature, Features Take the first The feature with the smallest Gini index is selected as the optimal partition feature, and a decision tree is generated recursively until the termination condition is met. Finally, the output of the decision tree is used as the risk feature vector .

[0037] Step 205, prediction model expansion: , , and As input, it is fed into an improved MLP prediction model In the future, the amount of resources required by each virtual machine is predicted and risk assessment value The improved MLP model adds risk assessment output on the basis of the original one, and its mathematical expression is: ; ; ; ; in, , and , are the output layer parameters for resource prediction and risk assessment, is the activation function for risk assessment (such as the sigmoid function, which maps the output to between 0 and 1, indicating the risk probability), For the The hidden state of the layer, is the dimension of the input features.

[0038] Step 206, Risk Warning and Processing: Setting Risk Thresholds ,when When a risk occurs, a risk warning is issued, and the old resource allocation plan is adjusted according to the risk type to obtain the adjusted resource amount. .

[0039] Step 207: Output resource allocation plan: Substitute into the resource allocation algorithm , output the final resource allocation plan to avoid risks .

[0040] Compare and get the resource allocation plan In this way, this embodiment introduces a risk assessment and early warning mechanism to discover in advance the business risks that may be caused by resource allocation, adjust the old resource allocation plan in time, change the CPU allocation, memory allocation and storage allocation, and effectively improve business stability.

[0041] In one embodiment of the present invention, the green energy-saving factor of resource allocation is taken into consideration to reduce the overall energy consumption of the cloud server; a cloud server virtual machine management method further includes the following steps: Step 301 to step 303 are the same as step 101 to step 103 of embodiment 1.

[0042] Step 304, energy consumption model construction: obtain energy consumption data of each hardware component (CPU, memory, storage, etc.) of the cloud server , using the linear regression model Model it and predict the energy consumption under different resource allocation schemes The linear regression model fits the data by minimizing the sum of squared errors, and its mathematical expression is: ; ; in, is the predicted value, is the input feature, is the feature weight, is the bias term, and Respectively The true value and predicted value of samples, is the total number of samples. The optimal weight and bias are solved by gradient descent and other optimization algorithms to obtain the energy consumption prediction model. .

[0043] Step 305, comprehensive optimization model: , and As input, it is fed into an improved MLP prediction model In the future, the amount of resources required by each virtual machine is predicted At the same time, combined with the energy consumption prediction model , with the goal of satisfying business needs with resource requirements and minimizing energy consumption, build an optimization function , solve for the optimal resource allocation The mathematical expression of the optimization function is: ; ; ; ; in, For the The predicted resource volume for each virtual machine, is the total number of virtual machines, is the total amount of resources that can be allocated to the cloud server. For the The goal of the optimization function is to minimize the predicted energy consumption while satisfying resource constraints and business requirements. . You can use heuristic optimization algorithms (such as genetic algorithms) to solve the optimization problem and get the optimal resource allocation , and its calculation formula is: ; in is the energy consumption prediction model, is the feature weight, is the bias term.

[0044] Step 306: Output resource allocation plan: Substitute into the resource allocation algorithm , output the final resource allocation plan .

[0045] Compare and get the resource allocation plan In this way, this embodiment comprehensively considers energy consumption factors when allocating resources, and can change CPU allocation, memory allocation and storage allocation while meeting business needs, effectively reducing the overall energy consumption of the cloud server and achieving the goal of green energy saving.

[0046] In one embodiment of the present invention, the resource allocation scheme is automatically optimized and tuned to adapt to long-term business changes and user behavior evolution; a cloud server virtual machine management method also includes the following steps: Step 401 to step 403 are the same as step 101 to step 103 of embodiment 1.

[0047] Step 404, feedback analysis: obtaining historical resource allocation plans and their business operation effect feedback data , using deep reinforcement learning algorithms (such as deep Q network DQN) Analyze it to obtain the direction and amplitude information of resource allocation strategy adjustment DQN fits the Q function (state-action value function) through a deep neural network and uses techniques such as experience replay and target network to improve training stability and convergence speed. The mathematical expression of DQN is: ; ; ; in, is the true Q function, is the Q function fitted by the neural network, are network parameters, is the learning rate, For TD target, For instant rewards, is the discount factor, is the target network parameter (periodically copied from the current network). The current network parameters are updated by minimizing the TD error, and the estimation of the Q function is continuously improved. In this embodiment, the resource allocation plan is used as the action of the intelligent agent, and the business operation effect is used as the reward of environmental feedback. The optimal resource allocation strategy is learned through the DQN algorithm, and the adjustment direction and amplitude information are given according to the strategy. .

[0048] Step 405, forecast model adjustment: according to MLP prediction model The parameters are adjusted to obtain the adjusted prediction model .Will , and Input into the adjusted forecasting model to predict the amount of resources required for each virtual machine in the future .

[0049] Step 406: Output resource allocation plan: Substitute into the resource allocation algorithm , output the final resource allocation plan Then, the new resource allocation plan and the corresponding business operation effect feedback data are updated to In the collection, loop optimization.

[0050] Compare and get the resource allocation plan In this embodiment, deep reinforcement learning is introduced to automatically optimize and tune the resource allocation plan, which can continuously improve the resource allocation strategy according to historical feedback, making the system adaptive, so as to better adapt to long-term business changes and user behavior evolution.

[0051] In one embodiment of the present invention, for cloud server virtual machine resource allocation in a big data analysis scenario, the real-time performance and efficiency of data analysis are improved; a cloud server virtual machine management method further includes the following steps: Step 501 is the same as step 101 in Example 1.

[0052] Step 502, big data task feature extraction: obtain detailed information of the big data analysis task , using a custom feature extraction algorithm Process it and extract the feature vector of big data task Custom feature extraction algorithms can design targeted feature engineering methods such as statistical features, text features, and graph features based on factors such as task type, data scale, and computational complexity to extract key features that can characterize the resource requirements of big data tasks.

[0053] Step 503 is the same as step 103 of embodiment 1.

[0054] Step 504 is the same as step 104 in Example 1.

[0055] Step 505, joint feature fusion: , , and Fusion is performed to obtain the fused feature vector The fusion method can be simple feature concatenation or weighted fusion, and the calculation method is: ; in , , , is the weight coefficient of each feature, which can be optimized by cross-validation and other methods. is the feature vector of the big data task, .

[0056] Step 506, prediction model adaptation: fusion feature vector Input into an MLP prediction model adapted to big data scenarios In the prediction, the amount of resources required for each virtual machine suitable for big data analysis tasks is predicted. Compared with general scenarios, the prediction model in the big data scenario makes targeted improvements and adaptations in terms of network structure, loss function, optimization algorithm, etc. to better fit the resource demand pattern of big data tasks.

[0057] Step 507, same as step 105 in Example 1, output the final resource allocation plan .

[0058] Compare and get the resource allocation plan In this way, this embodiment extracts big data task features and integrates them into resource prediction, so as to optimize resource allocation according to the characteristics of big data analysis scenarios, change CPU allocation, memory allocation and storage allocation, accelerate the data analysis process, improve the real-time performance of analysis results, and meet special needs in big data scenarios.

[0059] In one embodiment of the present invention, for cloud server virtual machine resource allocation in a containerized application deployment scenario, the flexibility and efficiency of resource allocation are improved; a cloud server virtual machine management method further includes the following steps: Step 601 is the same as step 101 in Example 1.

[0060] Step 602: Container feature extraction: Obtaining relevant information of containerized applications , using graph neural network (GNN) Process it and extract the container feature vector . GNN can effectively learn the feature representation of graph structured data through message passing and aggregation operations. In this embodiment, the dependency relationship, resource requirements and other information between containers are constructed into a graph, each node represents a container, and the edge represents the relationship between containers. Then GNN is used to encode the graph to obtain the feature vector of each container. The forward propagation formula of GNN is: ; ; in, For the Tier The feature vector of a node, For the Tier The feature vector of a node, For the Tier The feature vector of each node. and They are node update and message passing functions (generally implemented by multi-layer perceptrons), For Node The set of neighbor nodes of For Node To Node The edge features of is the eigenvector of the entire graph, is a graph-level pooling function (such as summation, maximum pooling, etc.). Through multi-layer message passing and node updates, GNN is able to capture complex relationships and interaction patterns between containers.

[0061] Step 603 is the same as step 103 of Example 1.

[0062] Step 604 is the same as step 104 of embodiment 1.

[0063] Step 605, dynamic resource prediction: , , and As input, it is fed into an MLP forecasting model combined with time series forecasting In the process, the dynamic changes in container resource requirements are considered to predict the amount of resources required for each container at different time points. Time series prediction can be achieved by introducing modules such as recurrent neural networks (RNNs) into MLP to capture the temporal dependencies of container resource requirements.

[0064] Step 606, resource allocation optimization: according to the predicted container resource requirements at different time points As well as the dependencies between containers, an improved resource allocation algorithm is used Perform resource allocation and output resource allocation plan The improved resource allocation algorithm needs to give priority to meeting the resource requirements of key containers while satisfying the container dependency constraints, while trying to improve resource utilization and allocation efficiency.

[0065] Compare and get the resource allocation plan In this way, this embodiment extracts container features and considers the dynamic characteristics and dependencies of containerized applications to change CPU allocation, memory allocation, and storage allocation, so as to better adapt to containerized deployment scenarios, improve the flexibility and efficiency of resource allocation, and ensure stable and efficient operation of containerized applications.

[0066] In one embodiment of the present invention, for cloud server virtual machine resource allocation in a multi-tenant cloud gaming scenario, the personalization and fairness of resource allocation are improved; a cloud server virtual machine management method further includes the following steps: Step 701 is the same as step 101 in Example 1.

[0067] Step 702: Tenant feature extraction: obtaining multi-tenant information , using deep autoencoder (DAE) Process it and extract tenant feature vector DAE learns the high-level feature representation of data by reconstructing the input data. Its encoder and decoder are generally implemented by multi-layer neural networks. The mathematical expression of DAE is: ; ; ; in, For input data, is the encoded feature vector, is the reconstructed data, and are encoder and decoder respectively, For the samples, For the The reconstructed data of samples, and is the activation function, , , , are learnable parameters, is the reconstruction loss function (such as mean square error), is the total number of samples. The parameters of DAE are optimized by minimizing the reconstruction error so that it can learn the essential characteristics of tenant data.

[0068] Step 703 is the same as step 103 of Example 1.

[0069] Step 704 is the same as step 104 in Example 1.

[0070] Step 705: Personalized resource prediction: , , and As input, it is fed into a personalized MLP prediction model In the game, the amount of resources required to meet the personalized needs of each tenant is predicted. Based on the original MLP structure, the personalized prediction model introduces technologies such as attention mechanism and user profiling to generate personalized resource prediction results according to the characteristics and preferences of different tenants.

[0071] Step 706, resource allocation coordination: Considering that multiple tenants share cloud server resources, a resource allocation coordination algorithm based on game theory is adopted , while ensuring the basic gaming experience of each tenant, try to allocate resources fairly and output a resource allocation plan The game theory resource allocation algorithm seeks fair resource allocation solutions such as Nash equilibrium by modeling the strategic interactions and conflicts of interest among multiple tenants, while taking into account the maximization of overall benefits.

[0072] Compare and get the resource allocation plan In this way, this embodiment can better meet the differentiated needs of different tenants in multi-tenant cloud gaming scenarios by extracting multi-tenant personalized features and performing personalized resource prediction and allocation coordination, improve the personalization and fairness of resource allocation, and enhance user experience and service competitiveness.

[0073] In at least one embodiment of the present invention, a cloud server virtual machine management system is provided, comprising: Data acquisition module, used to obtain historical user behavior data, current business demand data, and cloud server resource status data; A feature extraction module is used to extract a user behavior pattern feature vector using a time series processing model, and to extract a business demand feature vector using a feature extraction model; A data processing module is used to perform standardization processing on the cloud server resource status data to obtain a resource status vector; A prediction module, used to input the user behavior pattern feature vector, the business demand feature vector and the resource state vector into a prediction model to predict the amount of resources required by each virtual machine in the future; The resource allocation module is used to output a resource allocation plan using a resource allocation algorithm based on the predicted resource quantity and the actual allocatable resources of the current cloud server.

[0074] It also includes a risk assessment module, which is used to obtain historical resource allocation failure case data, extract risk feature vectors using a decision tree algorithm, and input the risk feature vectors together with other feature vectors into an improved prediction model to predict a risk assessment value. When the risk assessment value is greater than a set threshold, a risk warning is issued and the resource allocation plan is adjusted.

[0075] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A cloud server virtual machine management method, characterized in that: include: Obtain historical user behavior data and use the time series processing model to extract user behavior pattern feature vectors; Obtain current business demand data and use feature extraction models to extract business demand feature vectors; Obtain cloud server resource status data and perform standardization to obtain a resource status vector; Input the user behavior pattern feature vector, business demand feature vector and resource state vector into the prediction model to predict the amount of resources required by each virtual machine in the future period of time; According to the predicted resource quantity and the actual allocatable resources of the current cloud server, the resource allocation plan is output through the resource allocation algorithm.

2. A cloud server virtual machine management method according to claim 1, characterized in that: The time series processing model is a long short-term memory network model; the feature extraction model is a convolutional neural network model; the resource state vector is obtained by a standardized function The prediction model is constructed by processing the cloud server resource status data; the prediction model is constructed by taking the user behavior pattern feature vector, the business demand feature vector and the resource status vector as input and inputting them into a multi-layer perceptron prediction model. In the example above, the resource requirements of each virtual machine in the future are predicted.

3. A cloud server virtual machine management method according to claim 1, characterized in that: Also includes: Obtain historical resource allocation failure case data and use decision tree algorithm to extract risk feature vectors; The risk feature vector is input into the improved prediction model together with the user behavior pattern feature vector, the business demand feature vector and the resource state vector. , predict the amount of resources and risk assessment value required by each virtual machine in the future based on historical resource allocation failure case data; Set a risk threshold. When the risk assessment value is greater than the risk threshold, a risk warning is issued and the resource allocation plan that has not undergone risk assessment is adjusted according to the risk type.

4. A cloud server virtual machine management method according to claim 1, characterized in that: Also includes: Obtain energy consumption data of each hardware component of the cloud server and use the energy consumption model Predict energy consumption under different resource allocation schemes; With the goal of satisfying business needs with resource requirements and minimizing energy consumption, an optimization function is constructed to solve the optimal resource allocation based on the energy consumption model; Substitute the optimal resource allocation amount based on the energy consumption model into the resource allocation algorithm and output the final resource allocation plan based on the energy consumption model.

5. A cloud server virtual machine management method according to claim 1, characterized in that: Also includes: Obtain historical resource allocation plans and their business operation effect feedback data, and use deep reinforcement learning algorithm analysis to obtain the adjustment direction and amplitude information of resource allocation strategy; The parameters of the improved prediction model are adjusted according to the adjustment direction and amplitude information to obtain an adjusted prediction model. ; The relevant feature vectors are input into the adjusted prediction model to predict the amount of resources based on historical data required for each virtual machine in the future period, and are substituted into the resource allocation algorithm to output the final resource allocation plan based on historical data. The resource allocation plan based on historical data and the corresponding business operation effect feedback data are then updated to the feedback data set for cyclic optimization.

6. A cloud server virtual machine management method according to claim 1, characterized in that: Also includes: Obtain detailed information about big data analysis tasks and use custom feature extraction algorithms to extract feature vectors of big data tasks; The big data task feature vector is merged with the user behavior pattern feature vector, the business demand feature vector and the resource state vector to obtain a merged feature vector; The fused feature vector is input into the adjusted prediction model adapted to the big data scenario , predict the amount of resources required by each virtual machine suitable for big data analysis tasks based on big data scenarios, and substitute them into the resource allocation algorithm to output the final resource allocation plan based on big data scenarios.

7. A cloud server virtual machine management method according to claim 1, characterized in that: Also includes: Obtain relevant information about containerized applications and use graph neural networks to extract container feature vectors; The container feature vector, the user behavior pattern feature vector, the business demand feature vector and the resource state vector are input into the adjusted prediction model combined with time series prediction to predict the amount of resources required for each container at different time points; According to the predicted container resource requirements at different time points and the dependencies between containers, an improved heuristic algorithm is used to allocate resources and output a resource allocation plan based on containerized applications.

8. A cloud server virtual machine management method according to claim 1, characterized in that: Also includes: Obtain multi-tenant information and use deep autoencoders to extract tenant feature vectors; The tenant feature vector, user behavior pattern feature vector, business demand feature vector and resource state vector are used as input to the personalized prediction model. , predicting the amount of resources needed to meet the personalized needs of each tenant; A resource allocation coordination algorithm based on game theory is adopted to fairly allocate resources while ensuring the basic gaming experience of each tenant, and a resource allocation plan based on game theory is output.

9. A cloud server virtual machine management system, used to execute a cloud server virtual machine management method according to any one of claims 1 to 8, characterized in that: include: Data acquisition module, used to obtain historical user behavior data, current business demand data, and cloud server resource status data; A feature extraction module is used to extract a user behavior pattern feature vector using a time series processing model, and to extract a business demand feature vector using a feature extraction model; A data processing module is used to perform standardization processing on the cloud server resource status data to obtain a resource status vector; A prediction module, used to input the user behavior pattern feature vector, the business demand feature vector and the resource state vector into the adjusted personalized prediction model to predict the amount of resources required by each virtual machine in the future; The resource allocation module is used to output a resource allocation plan using a resource allocation algorithm based on the predicted resource quantity and the actual allocatable resources of the current cloud server.

10. A cloud server virtual machine management system according to claim 9, characterized in that: It also includes a risk assessment module, which is used to obtain historical resource allocation failure case data, extract risk feature vectors using a decision tree algorithm, and input the risk feature vectors together with other feature vectors into an improved prediction model to predict a risk assessment value. When the risk assessment value is greater than a set threshold, a risk warning is issued and a resource allocation plan based on the risk assessment is adjusted.

Images