Virtual machine intelligent deployment method and system based on dynamic resource prediction

Abstract

The invention provides a virtual machine intelligent deployment method and system based on dynamic resource prediction in the technical field of cloud computing and artificial intelligence crossing. The method comprises the following steps: S1, creating a resource demand prediction model; s2, acquiring a large number of historical operation data of the virtual machines to construct a data set; s3, training the resource demand prediction model based on the data set; s4, verifying and analyzing the input virtual machine deployment instruction to obtain virtual machine deployment parameters; s5, automatically executing deployment operation of the virtual machines based on the deployment parameters of the virtual machines, and setting priorities of the deployed virtual machines; s6, collecting real-time operation data, inputting the data into the resource demand prediction model, and obtaining a resource demand prediction result; and S7, dynamically adjusting virtual machine deployment parameters based on the resource demand prediction result, the priority and a Nash equalization algorithm. The method has the advantages that the resource utilization rate, efficiency, reliability, flexibility and safety of virtual machine deployment are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intersection of cloud computing and artificial intelligence, and in particular to a method and system for intelligent deployment of virtual machines based on dynamic resource prediction. Background Art

[0002] A virtual machine is a computer system simulated through software, creating one or more virtual computer environments (virtual machines) on a physical computer (host). Developers can create different development environments within virtual machines for software development and testing; students and learners can install different operating systems and software within virtual machines for learning and experimentation; businesses can use virtual machines to consolidate server resources, improve resource utilization, and reduce hardware costs; and users can install different operating systems within virtual machines for entertainment, office work, and other purposes. Virtual machines have a wide range of applications.

[0003] Commercial virtual machines are generally deployed on servers. Services are purchased from corresponding service providers to create virtual machines, and the virtual machines call on the server's computing resources to perform related computing tasks. When deploying a virtual machine, hardware resources, including at least processors, memory, and storage, need to be allocated to the virtual machine. Traditionally, these hardware resources are allocated manually in fixed proportions, resulting in severe resource fragmentation. Manual allocation is also inefficient and untimely, prone to operational errors, and unable to be flexibly adjusted based on current load conditions. While there are existing methods for dynamically allocating hardware resources through predictive models, these models often use linear regression to predict loads, which cannot adapt to sudden traffic bursts and often cause virtual machines to crash due to insufficient resources. Furthermore, traditionally, corresponding security measures are not taken when allocating resources to virtual machines, leading to malicious modifications to virtual machine configurations (malicious adjustments to resource allocation), which poses a hidden danger to the stable operation of virtual machines.

[0004] Therefore, how to provide a virtual machine intelligent deployment method and system based on dynamic resource prediction to improve the resource utilization, efficiency, reliability, flexibility and security of virtual machine deployment has become a technical problem that needs to be solved urgently. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method and system for intelligent deployment of virtual machines based on dynamic resource prediction, so as to improve the resource utilization, efficiency, reliability, flexibility and security of virtual machine deployment.

[0006] In a first aspect, the present invention provides a method for intelligent deployment of virtual machines based on dynamic resource prediction, comprising the following steps: Step S1: The server creates a resource demand prediction model based on the input module, the local feature extraction module, the temporal dependency feature extraction module, the semantic feature extraction module, the feature fusion module, and the output module, and sets a loss function of the resource demand prediction model based on a mean square error function, a mean absolute error function, and a mean absolute percentage error function; Step S2: The server obtains historical operation data of a large number of virtual machines, including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data, and news data, and constructs a data set after preprocessing and annotating each of the historical operation data; Step S3: The server trains the resource demand prediction model based on the data set and the loss function, and deploys the trained resource demand prediction model; Step S4: The server obtains the input virtual machine deployment instruction, verifies the virtual machine deployment instruction, and parses it to obtain virtual machine deployment parameters; Step S5: The server automatically executes the virtual machine deployment operation based on the virtual machine deployment parameters and sets the priority of each deployed virtual machine; Step S6: The server collects real-time operation data and inputs the real-time operation data into the deployed resource demand prediction model to obtain a resource demand prediction result; Step S7: The server dynamically adjusts the virtual machine deployment parameters of the deployed virtual machines based on the resource demand prediction result, priority, and Nash equilibrium algorithm; Step S8: The server records the virtual machine deployment log in real time and encrypts and stores the virtual machine deployment log; In step S1, the input module is constructed based on a normalization layer and is used to perform a normalization operation on the input operation data; the local feature extraction module is constructed based on a convolutional neural network and is used to extract local features from the operation data; the temporal dependency feature extraction module is constructed based on a bidirectional long short-term memory network and a self-attention network and is used to extract temporal dependency features from the operation data; the semantic feature extraction module is constructed based on a BERT network and is used to extract semantic features from the operation data; the feature fusion module is constructed based on a weighted bidirectional feature pyramid network and is used to fuse local features, temporal dependency features and semantic features to obtain fused features; the output module is constructed based on a fully connected layer and is used to output a resource demand forecast result within a preset time range in the future based on the fused features; The formula of the loss function is: L total =a*L MSE +b*L MAE +c*L MAPE ; Among them, L totalRepresents the loss value of the loss function; L MSE represents the mean square error function; L MAE represents the mean absolute error function; L MAPE represents the mean absolute percentage error function; a, b, and c represent weight coefficients.

[0007] Furthermore, the step S2 is specifically as follows: The server obtains historical operation data of a large number of virtual machines, including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data and news data; the virtual machine configuration data includes at least CPU configuration, memory configuration, storage configuration, network configuration, and operating system configuration; the virtual machine business data includes at least business type, business process and transaction data; the virtual machine performance data includes at least CPU usage rate, CPU utilization rate, CPU waiting time, CPU context switch times, CPU instruction execution times, real-time memory usage rate, average memory usage rate, memory utilization rate, memory swap rate, memory page fault rate, disk read and write speed, disk response time, network bandwidth usage rate, network throughput, network packet loss rate and network delay; the virtual machine status data includes operation status and availability status; the operation status is running, paused or closed; the availability status is normal, warning or fault; Each of the historical operation data is preprocessed, including at least text cleaning, text standardization, word segmentation, part-of-speech tagging, and sequence filling, and resource requirements are annotated on each of the preprocessed historical operation data, and a data set is constructed based on the annotated historical operation data.

[0008] Furthermore, the step S3 is specifically as follows: The dataset is divided into a training set, a validation set, and a test set based on a k-fold cross-validation method. The resource demand prediction model is trained using the training set. During the training process, adversarial samples are generated and added to the training set. The learning rate, batch size, optimizer, regularization parameter, and network structure of the resource demand prediction model are continuously optimized until the loss value of the loss function is less than a preset loss threshold. The trained resource demand prediction model is verified using the validation set to determine whether the prediction accuracy is greater than a preset accuracy threshold. If not, the training set is expanded to continue training. If so, then: The resource demand prediction model that has passed the verification is tested using the test set to determine whether the mean absolute percentage error is less than a preset error threshold. If not, the training set is expanded to continue training; if so, the training is terminated and the trained resource demand prediction model is deployed using a Docker container.

[0009] Furthermore, the step S4 is specifically as follows: The server obtains an input virtual machine deployment instruction carrying a parameter ciphertext, an authentication ciphertext, and a hash value; the parameter ciphertext is obtained by encrypting the virtual machine deployment parameters, and the encryption process is specifically as follows: encrypting the virtual machine deployment parameters using the AES algorithm to obtain first-level encrypted data, converting the first-level encrypted data into binary data, swapping the 1s and 0s in the binary data to obtain second-level encrypted data, and encrypting the second-level encrypted data into parameter ciphertext using the RC6 algorithm; the virtual machine deployment parameters include at least the number of CPU cores, CPU frequency, memory size, disk capacity, disk type, storage location, network interface, network bandwidth, and operating system; the authentication ciphertext is obtained by encrypting the username, password, device serial number, and timestamp, and the encryption process is specifically as follows: encrypting the username, password, device serial number, and timestamp into first-level encrypted data using the RSA algorithm, mapping a string of a specified position and specified length in the first-level encrypted data using a preset mapping rule to obtain second-level encrypted data, and encrypting the second-level encrypted data into authentication ciphertext using the ChaCha20 algorithm; the hash value is obtained by hashing the parameter ciphertext and the authentication ciphertext using the BLAKE2 algorithm; The server parses the received virtual machine deployment instruction to obtain a parameter ciphertext, an authentication ciphertext, and a hash value; after performing an integrity check on the parameter ciphertext and the authentication ciphertext using the hash value, the server decrypts the authentication ciphertext into a second-layer encrypted data using the ChaCha20 algorithm; maps a string of a specified position and a specified length in the second-layer encrypted data using the mapping rule to obtain a first-layer encrypted data; and decrypts the first-layer encrypted data into a user name and password, a device serial number, and a timestamp using the RSA algorithm; The server performs a validity check based on the timestamp, performs a validity check based on the device serial number, and then performs a matching check on the username and password based on the preset user management table. If the check passes: The parameter ciphertext is decrypted into secondary encrypted data using the RC6 algorithm, the 1s and 0s in the secondary encrypted data are swapped, and the binary data is inversely converted to obtain primary encrypted data, and the primary encrypted data is decrypted using the AES algorithm to obtain the virtual machine deployment parameters.

[0010] Furthermore, the step S5 is specifically as follows: The server automatically executes the virtual machine deployment operation through RPA based on the virtual machine deployment parameters, and sets the priority of each virtual machine based on the business type of each deployed virtual machine; The step S6 is specifically as follows: The server collects real-time operation data including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data and news data, pre-processes the real-time operation data and inputs it into the deployed resource demand prediction model to obtain resource demand prediction results; The step S7 is specifically as follows: The server analyzes the resource demand prediction result, determines whether the resource demand of each deployed virtual machine changes within a preset time range in the future, and if so, determines whether the server can meet the changed resource demand. If so, dynamically adjusts the virtual machine deployment parameters of the deployed virtual machine based on the Nash equilibrium algorithm; if not, screens the deployed virtual machines based on the priority, and dynamically adjusts the virtual machine deployment parameters of the screened virtual machines based on the Nash equilibrium algorithm; The step S8 is specifically as follows: The server records virtual machine deployment logs in real time through Elastic Observability, creates a pair of public and private keys using the ECDSA algorithm, encrypts the virtual machine deployment logs using the private key to obtain first encrypted data, cuts the last three characters of the private key to the middle to obtain the key, encrypts the first encrypted data and the key into second encrypted data using the Serpent algorithm, shifts each character of the second encrypted data left by 8 bits to obtain an encrypted log, stores the encrypted log in a specified path, and backs up the encrypted log to the blockchain.

[0011] In a second aspect, the present invention provides a virtual machine intelligent deployment system based on dynamic resource prediction, comprising the following modules: A resource demand prediction model creation module is used for the server to create a resource demand prediction model based on the input module, the local feature extraction module, the temporal dependency feature extraction module, the semantic feature extraction module, the feature fusion module, and the output module, and to set the loss function of the resource demand prediction model based on the mean square error function, the mean absolute error function, and the mean absolute percentage error function; A data set construction module is used for the server to obtain historical operation data of a large number of virtual machines, including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data, and news data, and to construct a data set after preprocessing and annotating each of the historical operation data; A resource demand prediction model training module is used for the server to train the resource demand prediction model based on the data set and the loss function, and deploy the trained resource demand prediction model; A virtual machine deployment instruction verification module is used for the server to obtain input virtual machine deployment instructions, verify the virtual machine deployment instructions and parse them to obtain virtual machine deployment parameters; A virtual machine automatic deployment module, configured to cause the server to automatically execute a virtual machine deployment operation based on the virtual machine deployment parameters and set a priority for each virtual machine to be deployed; The resource demand prediction module is used to collect real-time operation data from the server and input the real-time operation data into the deployed resource demand prediction model to obtain resource demand prediction results; A deployment parameter dynamic adjustment module is used for the server to dynamically adjust the virtual machine deployment parameters of the deployed virtual machines based on the resource demand prediction results, priority and Nash equilibrium algorithm; A deployment log management module is used for the server to record virtual machine deployment logs in real time and encrypt and store the virtual machine deployment logs; In the resource demand prediction model creation module, the input module is constructed based on the normalization layer, and is used to perform a normalization operation on the input operation data; the local feature extraction module is constructed based on the convolutional neural network, and is used to extract local features from the operation data; the temporal dependency feature extraction module is constructed based on the bidirectional long short-term memory network and the self-attention network, and is used to extract temporal dependency features from the operation data; the semantic feature extraction module is constructed based on the BERT network, and is used to extract semantic features from the operation data; the feature fusion module is constructed based on the weighted bidirectional feature pyramid network, and is used to fuse local features, temporal dependency features and semantic features to obtain fused features; the output module is constructed based on the fully connected layer, and is used to output the resource demand forecast results within a preset time range in the future based on the fused features; The formula of the loss function is: L total =a*L MSE +b*L MAE +c*L MAPE ; Among them, L total Represents the loss value of the loss function; L MSE represents the mean square error function; L MAE represents the mean absolute error function; L MAPE represents the mean absolute percentage error function; a, b, and c represent weight coefficients.

[0012] Furthermore, the dataset construction module is specifically used to: The server obtains historical operation data of a large number of virtual machines, including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data and news data; the virtual machine configuration data includes at least CPU configuration, memory configuration, storage configuration, network configuration, and operating system configuration; the virtual machine business data includes at least business type, business process and transaction data; the virtual machine performance data includes at least CPU usage rate, CPU utilization rate, CPU waiting time, CPU context switch times, CPU instruction execution times, real-time memory usage rate, average memory usage rate, memory utilization rate, memory swap rate, memory page fault rate, disk read and write speed, disk response time, network bandwidth usage rate, network throughput, network packet loss rate and network delay; the virtual machine status data includes operation status and availability status; the operation status is running, paused or closed; the availability status is normal, warning or fault; Each of the historical operation data is preprocessed, including at least text cleaning, text standardization, word segmentation, part-of-speech tagging, and sequence filling, and resource requirements are annotated on each of the preprocessed historical operation data, and a data set is constructed based on the annotated historical operation data.

[0013] Furthermore, the resource demand prediction model training module is specifically used to: The dataset is divided into a training set, a validation set, and a test set based on a k-fold cross-validation method. The resource demand prediction model is trained using the training set. During the training process, adversarial samples are generated and added to the training set. The learning rate, batch size, optimizer, regularization parameter, and network structure of the resource demand prediction model are continuously optimized until the loss value of the loss function is less than a preset loss threshold. The trained resource demand prediction model is verified using the validation set to determine whether the prediction accuracy is greater than a preset accuracy threshold. If not, the training set is expanded to continue training. If so, then: The resource demand prediction model that has passed the verification is tested using the test set to determine whether the mean absolute percentage error is less than a preset error threshold. If not, the training set is expanded to continue training; if so, the training is terminated and the trained resource demand prediction model is deployed using a Docker container.

[0014] Furthermore, the virtual machine deployment instruction verification module is specifically used to: The server obtains an input virtual machine deployment instruction carrying a parameter ciphertext, an authentication ciphertext, and a hash value; the parameter ciphertext is obtained by encrypting the virtual machine deployment parameters, and the encryption process is specifically as follows: encrypting the virtual machine deployment parameters using the AES algorithm to obtain first-level encrypted data, converting the first-level encrypted data into binary data, swapping the 1s and 0s in the binary data to obtain second-level encrypted data, and encrypting the second-level encrypted data into parameter ciphertext using the RC6 algorithm; the virtual machine deployment parameters include at least the number of CPU cores, CPU frequency, memory size, disk capacity, disk type, storage location, network interface, network bandwidth, and operating system; the authentication ciphertext is obtained by encrypting the username, password, device serial number, and timestamp, and the encryption process is specifically as follows: encrypting the username, password, device serial number, and timestamp into first-level encrypted data using the RSA algorithm, mapping a string of a specified position and specified length in the first-level encrypted data using a preset mapping rule to obtain second-level encrypted data, and encrypting the second-level encrypted data into authentication ciphertext using the ChaCha20 algorithm; the hash value is obtained by hashing the parameter ciphertext and the authentication ciphertext using the BLAKE2 algorithm; The server parses the received virtual machine deployment instruction to obtain a parameter ciphertext, an authentication ciphertext, and a hash value; after performing an integrity check on the parameter ciphertext and the authentication ciphertext using the hash value, the server decrypts the authentication ciphertext into a second-layer encrypted data using the ChaCha20 algorithm; maps a string of a specified position and a specified length in the second-layer encrypted data using the mapping rule to obtain a first-layer encrypted data; and decrypts the first-layer encrypted data into a user name and password, a device serial number, and a timestamp using the RSA algorithm; The server performs a validity check based on the timestamp, performs a validity check based on the device serial number, and then performs a matching check on the username and password based on the preset user management table. If the check passes: The parameter ciphertext is decrypted into secondary encrypted data using the RC6 algorithm, the 1s and 0s in the secondary encrypted data are swapped, and the binary data is inversely converted to obtain primary encrypted data, and the primary encrypted data is decrypted using the AES algorithm to obtain the virtual machine deployment parameters.

[0015] Furthermore, the virtual machine automatic deployment module is specifically used to: The server automatically executes the virtual machine deployment operation through RPA based on the virtual machine deployment parameters, and sets the priority of each virtual machine based on the business type of each deployed virtual machine; The resource demand prediction module is specifically used to: The server collects real-time operation data including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data and news data, pre-processes the real-time operation data and inputs it into the deployed resource demand prediction model to obtain resource demand prediction results; The deployment parameter dynamic adjustment module is specifically used to: The server analyzes the resource demand prediction result, determines whether the resource demand of each deployed virtual machine changes within a preset time range in the future, and if so, determines whether the server can meet the changed resource demand. If so, dynamically adjusts the virtual machine deployment parameters of the deployed virtual machine based on the Nash equilibrium algorithm; if not, screens the deployed virtual machines based on the priority, and dynamically adjusts the virtual machine deployment parameters of the screened virtual machines based on the Nash equilibrium algorithm; The deployment log management module is specifically used to: The server records virtual machine deployment logs in real time through Elastic Observability, creates a pair of public and private keys using the ECDSA algorithm, encrypts the virtual machine deployment logs using the private key to obtain first encrypted data, cuts the last three characters of the private key to the middle to obtain the key, encrypts the first encrypted data and the key into second encrypted data using the Serpent algorithm, shifts each character of the second encrypted data left by 8 bits to obtain an encrypted log, stores the encrypted log in a specified path, and backs up the encrypted log to the blockchain.

[0016] The advantages of the present invention are: 1. The server creates a resource demand prediction model based on the input module, local feature extraction module, temporal dependency feature extraction module, semantic feature extraction module, feature fusion module and output module, and sets the loss function of the resource demand prediction model based on the mean square error function, the mean absolute error function and the mean absolute percentage error function; then obtains a large number of historical operation data of virtual machines including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data and news data, pre-processes and annotates each historical operation data to construct a data set, trains the resource demand prediction model based on the data set and the loss function, and deploys the trained resource demand prediction model; then the server verifies and parses the obtained virtual machine deployment instructions to obtain virtual machine deployment parameters, automatically executes the virtual machine deployment operation based on the virtual machine deployment parameters, and sets the priority of each deployed virtual machine; then the server collects real-time operation data and inputs it into the resource demand prediction model to obtain resource demand prediction results, and The priority and Nash equilibrium algorithm dynamically adjust the virtual machine deployment parameters of the deployed virtual machine, record the virtual machine deployment log in real time, and encrypt and store the virtual machine deployment log; that is, the server automatically performs the initial deployment operation based on the input virtual machine deployment instruction. During the operation of the virtual machine, resource demand is predicted through the pre-trained resource demand prediction model, and the virtual machine deployment parameters are dynamically adjusted in combination with the priority and Nash equilibrium algorithm. Only the virtual machine deployment instruction needs to be sent during the initial deployment, and no manual intervention is required subsequently, avoiding the inefficiency and errors of manual operation. In addition, the virtual machine deployment parameters are dynamically adjusted through multi-dimensional analysis of virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data and news data, which can effectively respond to emergencies and alleviate resource fragmentation problems. Combined with the verification of virtual machine deployment instructions in the early stage of deployment and the encrypted storage of virtual machine deployment logs, illegal operations are avoided and it is convenient for later tracing, ultimately greatly improving the resource utilization, efficiency, reliability, flexibility and security of virtual machine deployment.

[0017] 2. By setting up a resource demand prediction model, local features, time-dependent features and semantic features are extracted through the local feature extraction module, time-dependent feature extraction module and semantic feature extraction module respectively, thereby greatly improving the feature extraction capability and being able to effectively mine potential patterns from the operating data; for example, the deployed virtual machine is used for e-commerce operations. The semantic feature extraction module extracts semantic features from news data to determine that a shopping festival for promotion will be held in the near future. This means that the load of the virtual machine will surge in the near future, and the corresponding virtual machine deployment parameters need to be adjusted to cope with it, ensuring the reliable operation of the virtual machine, thereby greatly improving the reliability of virtual machine deployment.

[0018] 3. By setting up a feature fusion module based on a weighted bidirectional feature pyramid network, the weighted bidirectional feature pyramid network extracts multi-scale features through convolutional layers of different scales (such as dilated convolutions with different dilation rates), and fuses high-level semantic features with low-level detail features through bottom-up and top-down paths. Different weights are assigned to features of each scale during the fusion process. By learning these weights to optimize the fusion effect, it is possible to capture both high-level semantic information and low-level detail information at the same time. Combined with the weighting mechanism, the contribution of features of each scale is dynamically adjusted to improve the fusion effect, thereby greatly improving the accuracy of resource demand prediction.

[0019] 4. The loss function of the resource demand prediction model is set by using the mean square error function, the mean absolute error function, and the mean absolute percentage error function. This allows the loss function to combine the advantages of the mean square error function, the mean absolute error function, and the mean absolute percentage error function, thereby greatly improving the training effect of the resource demand prediction model and thus greatly improving the accuracy of resource demand prediction.

[0020] 5. Construct a data set through historical operation data including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data and news data; virtual machine configuration data at least includes CPU configuration, memory configuration, storage configuration, network configuration, and operating system configuration; virtual machine business data at least includes business type, business process and transaction data; virtual machine performance data at least includes CPU usage rate, CPU utilization rate, CPU waiting time, CPU context switching times, CPU instruction execution times, real-time memory usage rate, average memory usage rate, memory utilization rate, memory swap rate, memory page fault rate, disk read and write speed, disk response time, network bandwidth utilization rate, network throughput, network packet loss rate and network delay; virtual machine status data includes operation status and availability status; covering internal operation status and external environmental factors, the data set can be used for multi-dimensional training and analysis in the future to effectively improve the generalization ability of the resource demand prediction model.

[0021] 6. By preprocessing the historical operation data, including at least text cleaning, text standardization, word segmentation, part-of-speech tagging, and sequence filling, the quality of the dataset can be effectively improved, the validity and consistency of the model input can be ensured, and the training effect of the resource demand prediction model can be effectively improved.

[0022] 7. The dataset is divided into training set, validation set and test set through the k-fold cross-validation method to reduce the accidental impact caused by single data segmentation and effectively reduce the risk of overfitting. Adversarial samples are generated and added to the training set during the training process to improve the model's anti-interference ability and adapt to sudden traffic scenarios. The learning rate, batch size, optimizer, regularization parameters and network structure of the resource demand prediction model are continuously optimized until the loss value of the loss function is less than the preset loss threshold; then the trained resource demand prediction model is verified through the validation set, and the verified resource demand prediction model is tested through the test set. That is, optimization, verification and testing are continuously performed during the training process, thereby greatly improving the accuracy of resource demand prediction.

[0023] 8. By setting the virtual machine deployment instruction to carry parameter ciphertext, authentication ciphertext and hash value, the parameter ciphertext is obtained by encrypting the virtual machine deployment parameters through the AES algorithm, base conversion adjustment, and RC6 algorithm. The authentication ciphertext is obtained by encrypting the username, password, device serial number and timestamp through the RSA algorithm, mapping rules, and ChaCha20 algorithm. Subsequently, multi-dimensional verification can be performed through the hash value, username, password, device serial number and timestamp, and the relevant data can be prevented from being stolen in plain text during transmission, thereby greatly improving the security of virtual machine deployment instruction transmission, and also avoiding the execution of virtual machine deployment operations based on illegal instructions, thereby greatly improving the security of virtual machine deployment.

[0024] 9. By using the Nash balancing algorithm to dynamically adjust virtual machine deployment parameters, efficient resource utilization, load balancing, performance optimization and cost control can be achieved, while improving the flexibility, adaptability and reliability of the system.

[0025] 10. Create a pair of public and private keys using the ECDSA algorithm. Use the private key to encrypt the virtual machine deployment log to obtain the first encrypted data. Cut the last three characters of the private key to the middle to obtain the key. Use the Serpent algorithm to encrypt the first encrypted data and the key into the second encrypted data. Shift each character of the second encrypted data 8 bits to the left to obtain the encrypted log. Store the encrypted log in the specified path and back it up to the blockchain. This combines at least five security measures (public and private keys, private key encryption, Serpent algorithm, character shifting, and blockchain), greatly improving the security of virtual machine deployment log storage.

[0026] 11. By extracting local features through CNN, capturing temporal dependencies through bidirectional LSTM+self-attention, and extracting semantic features through BERT, and combining the weighted bidirectional feature pyramid network to fuse multi-dimensional features, the prediction accuracy is significantly improved and it can adapt to complex scenarios. By fusing the three error functions of MSE, MAE, and MAPE, the model's sensitivity to different errors is dynamically adjusted through the weight coefficient, avoiding the bias of a single indicator and effectively enhancing the robustness of the model.

[0027] 12. By setting priorities based on business types, key business resources are ensured and system reliability and responsiveness are improved. Virtual machine deployment parameters are dynamically adjusted based on resource demand forecast results, priorities, and the Nash equilibrium algorithm. This uses game theory to optimize resource allocation strategies, balance resource competition between virtual machines, and maximize overall resource utilization.

[0028] 13. Through innovative multi-modal prediction, intelligent dynamic scheduling, multi-level security protection and automated deployment mechanisms, it effectively solves the pain points of rigid resource allocation, low prediction accuracy and weak security protection in traditional virtual machine deployment.

[0029] 14. In addition to traditional performance indicators, news data (such as industry trends and market events) are innovatively introduced as input features. The BERT model is used to capture the potential impact of external events on resource demand, enabling forward-looking predictions.

[0030] 15. AES-RC6 double encryption + binary inversion design avoids the mathematical vulnerability of traditional encryption algorithms; RSA mapping + ChaCha20 chain encryption prevents key guessing and man-in-the-middle attacks; ECDSA + Serpent + cyclic shift triple encryption, combined with blockchain evidence storage, meets financial-level security requirements.

[0031] 16. Through the Docker container deployment model, it supports orchestration tools such as Kubernetes to achieve elastic scaling of resources and cross-platform compatibility.

[0032] 17. Through interdisciplinary technology integration (cloud computing, AI, cryptography, game theory) and full life cycle security design, problems such as resource rigidity, prediction lag, and security risks in traditional virtual machine deployment have been solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0034] Figure 1 This is a flow chart of a method for intelligent deployment of virtual machines based on dynamic resource prediction according to the present invention.

[0035] Figure 2It is a structural diagram of a virtual machine intelligent deployment system based on dynamic resource prediction of the present invention. DETAILED DESCRIPTION

[0036] The technical solution in the embodiments of the present application has the following overall idea: the server automatically performs the initial deployment operation based on the input virtual machine deployment instruction, and during the operation of the virtual machine, the resource demand prediction model is used to predict the resource demand, and the virtual machine deployment parameters are dynamically adjusted in combination with the priority and Nash equilibrium algorithm. Only the virtual machine deployment instruction needs to be sent during the initial deployment, and no manual intervention is required subsequently. The virtual machine deployment parameters are dynamically adjusted through multi-dimensional analysis of virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data and news data, which can effectively respond to emergencies and alleviate resource fragmentation problems. Combined with the verification of virtual machine deployment instructions in the early stage of deployment and the encrypted storage of virtual machine deployment logs, illegal operations are avoided and it is convenient for later tracing, thereby improving the resource utilization, efficiency, reliability, flexibility and security of virtual machine deployment.

[0037] Please refer to Figures 1 to 2 As shown, a preferred embodiment of a method for intelligent deployment of virtual machines based on dynamic resource prediction of the present invention includes the following steps: Step S1: The server creates a resource demand prediction model based on the input module, the local feature extraction module, the temporal dependency feature extraction module, the semantic feature extraction module, the feature fusion module, and the output module, and sets a loss function of the resource demand prediction model based on a mean square error function, a mean absolute error function, and a mean absolute percentage error function; By setting up a resource demand prediction model, local features, temporal dependency features and semantic features are extracted through the local feature extraction module, temporal dependency feature extraction module and semantic feature extraction module respectively, thereby greatly improving the feature extraction capability and effectively mining potential patterns from the operating data; for example, the deployed virtual machine is used for e-commerce operations. The semantic feature extraction module extracts semantic features from news data and determines that a shopping festival for promotion will be held in the near future. This means that the load of the virtual machine will increase suddenly in the near future, and the corresponding virtual machine deployment parameters need to be adjusted to cope with it, so as to ensure the reliable operation of the virtual machine, thereby greatly improving the reliability of virtual machine deployment.

[0038] Step S2: The server obtains historical operation data of a large number of virtual machines, including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data, and news data, and constructs a data set after preprocessing and annotating each of the historical operation data; In addition to traditional performance indicators, by innovatively introducing news data (such as industry dynamics and market events) as input features, the BERT model is used to capture the potential impact of external events on resource demand, achieving forward-looking predictions.

[0039] Step S3: The server trains the resource demand prediction model based on the data set and the loss function, and deploys the trained resource demand prediction model; Step S4: The server obtains the input virtual machine deployment instruction, verifies the virtual machine deployment instruction, and parses it to obtain virtual machine deployment parameters; Step S5: The server automatically executes the virtual machine deployment operation based on the virtual machine deployment parameters and sets the priority of each deployed virtual machine; Step S6: The server collects real-time operation data and inputs the real-time operation data into the deployed resource demand prediction model to obtain a resource demand prediction result; Step S7: The server dynamically adjusts the virtual machine deployment parameters of the deployed virtual machines based on the resource demand prediction result, priority, and Nash equilibrium algorithm; Step S8: The server records the virtual machine deployment log in real time and encrypts and stores the virtual machine deployment log; Through innovative multi-modal prediction, intelligent dynamic scheduling, multi-level security protection and automated deployment mechanisms, it effectively solves the pain points of rigid resource allocation, low prediction accuracy and weak security protection in traditional virtual machine deployment.

[0040] Through interdisciplinary technology integration (cloud computing, AI, cryptography, game theory) and full life cycle security design, problems such as resource rigidity, prediction lag, and security risks in traditional virtual machine deployment are solved.

[0041] In step S1, the input module is constructed based on a normalization layer and is used to perform a normalization operation on the input operation data; the local feature extraction module is constructed based on a convolutional neural network and is used to extract local features from the operation data; the temporal dependency feature extraction module is constructed based on a bidirectional long short-term memory network and a self-attention network and is used to extract temporal dependency features from the operation data; the semantic feature extraction module is constructed based on a BERT network and is used to extract semantic features from the operation data; the feature fusion module is constructed based on a weighted bidirectional feature pyramid network and is used to fuse local features, temporal dependency features and semantic features to obtain fused features; the output module is constructed based on a fully connected layer and is used to output a resource demand forecast result within a preset time range in the future based on the fused features; By setting up a feature fusion module based on a weighted bidirectional feature pyramid network, the weighted bidirectional feature pyramid network extracts multi-scale features through convolutional layers of different scales (such as dilated convolutions with different dilation rates), and fuses high-level semantic features with low-level detail features through bottom-up and top-down paths. Different weights are assigned to features of each scale during the fusion process. By learning these weights to optimize the fusion effect, it is possible to capture both high-level semantic information and low-level detail information at the same time. Combined with the weighting mechanism, the contribution of features of each scale is dynamically adjusted to improve the fusion effect, thereby greatly improving the accuracy of resource demand prediction.

[0042] The formula of the loss function is: L total =a*L MSE +b*L MAE +c*L MAPE ; Among them, L total Represents the loss value of the loss function; L MSE represents the mean square error function; L MAE represents the mean absolute error function; L MAPE represents the mean absolute percentage error function; a, b, and c represent weight coefficients.

[0043] The loss function of the resource demand prediction model is set by the mean square error function, the mean absolute error function and the mean absolute percentage error function, so that the loss function combines the advantages of the mean square error function, the mean absolute error function and the mean absolute percentage error function, thereby greatly improving the training effect of the resource demand prediction model and greatly improving the accuracy of resource demand prediction.

[0044] By extracting local features through CNN, capturing temporal dependencies through bidirectional LSTM+self-attention, and extracting semantic features through BERT, and combining with the weighted bidirectional feature pyramid network to fuse multi-dimensional features, the prediction accuracy is significantly improved and it can adapt to complex scenarios. By fusing the three error functions of MSE, MAE, and MAPE, the sensitivity of the model to different errors is dynamically adjusted through the weight coefficient, avoiding the deviation of a single indicator and effectively enhancing the robustness of the model.

[0045] The step S2 is specifically as follows: The server obtains historical operation data of a large number of virtual machines, including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data and news data; the virtual machine configuration data includes at least CPU configuration, memory configuration, storage configuration, network configuration, and operating system configuration; the virtual machine business data includes at least business type, business process and transaction data; the virtual machine performance data includes at least CPU usage rate, CPU utilization rate, CPU waiting time, CPU context switch times, CPU instruction execution times, real-time memory usage rate, average memory usage rate, memory utilization rate, memory swap rate, memory page fault rate, disk read and write speed, disk response time, network bandwidth usage rate, network throughput, network packet loss rate and network delay; the virtual machine status data includes operation status and availability status; the operation status is running, paused or closed; the availability status is normal, warning or fault; A data set is constructed using historical operating data including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data, and news data; virtual machine configuration data includes at least CPU configuration, memory configuration, storage configuration, network configuration, and operating system configuration; virtual machine business data includes at least business type, business process, and transaction data; virtual machine performance data includes at least CPU usage, CPU utilization, CPU wait time, CPU context switch times, CPU instruction execution times, real-time memory usage, average memory usage, memory utilization, memory swap rate, memory page fault rate, disk read and write speed, disk response time, network bandwidth utilization, network throughput, network packet loss rate, and network latency; virtual machine status data includes operating status and availability status; covering internal operating status and external environmental factors, the data set can be used for multi-dimensional training and analysis in the future to effectively improve the generalization ability of the resource demand prediction model.

[0046] Each of the historical operation data is preprocessed, including at least text cleaning, text standardization, word segmentation, part-of-speech tagging, and sequence filling, and resource requirements are annotated on each of the preprocessed historical operation data, and a data set is constructed based on the annotated historical operation data.

[0047] By preprocessing each historical operation data, including at least text cleaning, text standardization, word segmentation, part-of-speech tagging, and sequence filling, the quality of the dataset can be effectively improved, the validity and consistency of the model input can be ensured, and the training effect of the resource demand prediction model can be effectively improved.

[0048] The step S3 is specifically as follows: The dataset is divided into a training set, a validation set, and a test set based on a k-fold cross-validation method. The resource demand prediction model is trained using the training set. During the training process, adversarial samples are generated and added to the training set. The learning rate, batch size, optimizer, regularization parameter, and network structure of the resource demand prediction model are continuously optimized until the loss value of the loss function is less than a preset loss threshold. The trained resource demand prediction model is verified using the validation set to determine whether the prediction accuracy is greater than a preset accuracy threshold. If not, the training set is expanded to continue training. If so, then: The resource demand prediction model that has passed the verification is tested using the test set to determine whether the mean absolute percentage error is less than a preset error threshold. If not, the training set is expanded to continue training; if so, the training is terminated and the trained resource demand prediction model is deployed using a Docker container.

[0049] The dataset is divided into training set, validation set and test set through the k-fold cross-validation method to reduce the accidental influence caused by single data segmentation and effectively reduce the risk of overfitting. Adversarial samples are generated and added to the training set during the training process to improve the model's anti-interference ability and adapt to sudden traffic scenarios. The learning rate, batch size, optimizer, regularization parameters and network structure of the resource demand prediction model are continuously optimized until the loss value of the loss function is less than the preset loss threshold; then the trained resource demand prediction model is verified through the validation set, and the verified resource demand prediction model is tested through the test set. That is, optimization, verification and testing are continuously carried out during the training process, which greatly improves the accuracy of resource demand prediction.

[0050] Through the Docker container deployment model, it supports orchestration tools such as Kubernetes to achieve elastic scaling of resources and cross-platform compatibility.

[0051] The step S4 is specifically as follows: The server obtains an input virtual machine deployment instruction carrying a parameter ciphertext, an authentication ciphertext, and a hash value; the parameter ciphertext is obtained by encrypting the virtual machine deployment parameters, and the encryption process is specifically as follows: encrypting the virtual machine deployment parameters using the AES algorithm to obtain first-level encrypted data, converting the first-level encrypted data into binary data, swapping the 1s and 0s in the binary data to obtain second-level encrypted data, and encrypting the second-level encrypted data into parameter ciphertext using the RC6 algorithm; the virtual machine deployment parameters include at least the number of CPU cores, CPU frequency, memory size, disk capacity, disk type, storage location, network interface, network bandwidth, and operating system; the authentication ciphertext is obtained by encrypting the username, password, device serial number, and timestamp, and the encryption process is specifically as follows: encrypting the username, password, device serial number, and timestamp into first-level encrypted data using the RSA algorithm, mapping a string of a specified position and specified length in the first-level encrypted data using a preset mapping rule to obtain second-level encrypted data, and encrypting the second-level encrypted data into authentication ciphertext using the ChaCha20 algorithm; the hash value is obtained by hashing the parameter ciphertext and the authentication ciphertext using the BLAKE2 algorithm; By setting the virtual machine deployment instruction to carry parameter ciphertext, authentication ciphertext and hash value, the parameter ciphertext is obtained by encrypting the virtual machine deployment parameters through the AES algorithm, base conversion adjustment, and RC6 algorithm. The authentication ciphertext is obtained by encrypting the username, password, device serial number and timestamp through the RSA algorithm, mapping rules, and ChaCha20 algorithm. Subsequently, multi-dimensional verification can be performed through the hash value, username, password, device serial number and timestamp, and the relevant data can be prevented from being stolen in plain text during transmission, thereby greatly improving the security of virtual machine deployment instruction transmission, and also avoiding the execution of virtual machine deployment operations based on illegal instructions, thereby greatly improving the security of virtual machine deployment.

[0052] The server parses the received virtual machine deployment instruction to obtain a parameter ciphertext, an authentication ciphertext, and a hash value; after performing an integrity check on the parameter ciphertext and the authentication ciphertext using the hash value, the server decrypts the authentication ciphertext into a second-layer encrypted data using the ChaCha20 algorithm; maps a string of a specified position and a specified length in the second-layer encrypted data using the mapping rule to obtain a first-layer encrypted data; and decrypts the first-layer encrypted data into a user name and password, a device serial number, and a timestamp using the RSA algorithm; The server performs a validity check based on the timestamp, performs a validity check based on the device serial number, and then performs a matching check on the username and password based on the preset user management table. If the check passes: The parameter ciphertext is decrypted into secondary encrypted data using the RC6 algorithm, the 1s and 0s in the secondary encrypted data are swapped, and the binary data is inversely converted to obtain primary encrypted data, and the primary encrypted data is decrypted using the AES algorithm to obtain the virtual machine deployment parameters.

[0053] The step S5 is specifically as follows: The server automatically executes the virtual machine deployment operation through RPA based on the virtual machine deployment parameters, and sets the priority of each virtual machine based on the business type of each deployed virtual machine; The step S6 is specifically as follows: The server collects real-time operation data including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data and news data, pre-processes the real-time operation data and inputs it into the deployed resource demand prediction model to obtain resource demand prediction results; The step S7 is specifically as follows: The server analyzes the resource demand prediction result, determines whether the resource demand of each deployed virtual machine changes within a preset time range in the future, and if so, determines whether the server can meet the changed resource demand. If so, dynamically adjusts the virtual machine deployment parameters of the deployed virtual machine based on the Nash equilibrium algorithm; if not, screens the deployed virtual machines based on the priority, and dynamically adjusts the virtual machine deployment parameters of the screened virtual machines based on the Nash equilibrium algorithm; By using the Nash balancing algorithm to dynamically adjust virtual machine deployment parameters, efficient resource utilization, load balancing, performance optimization, and cost control can be achieved, while improving the flexibility, adaptability, and reliability of the system.

[0054] By setting priorities based on business types, key business resources are ensured and system reliability and responsiveness are improved. Virtual machine deployment parameters are dynamically adjusted based on resource demand forecast results, priorities, and the Nash equilibrium algorithm. This uses game theory to optimize resource allocation strategies, balance resource competition between virtual machines, and maximize overall resource utilization.

[0055] The step S8 is specifically as follows: The server records virtual machine deployment logs in real time through Elastic Observability, creates a pair of public and private keys using the ECDSA algorithm, encrypts the virtual machine deployment logs using the private key to obtain first encrypted data, cuts the last three characters of the private key to the middle to obtain the key, encrypts the first encrypted data and the key into second encrypted data using the Serpent algorithm, shifts each character of the second encrypted data left by 8 bits to obtain an encrypted log, stores the encrypted log in a specified path, and backs up the encrypted log to the blockchain.

[0056] A pair of public and private keys is created using the ECDSA algorithm. The virtual machine deployment log is encrypted using the private key to obtain the first encrypted data. The last three characters of the private key are cut to the middle to obtain the key. The first encrypted data and the key are encrypted into the second encrypted data using the Serpent algorithm. Each character of the second encrypted data is circularly shifted 8 bits to the left to obtain the encrypted log. The encrypted log is stored in a specified path and backed up to the blockchain. At least five security measures are combined (public and private keys, private key encryption, Serpent algorithm, character shifting, and blockchain), greatly improving the security of virtual machine deployment log storage.

[0057] Through AES-RC6 double encryption + binary inversion design, the mathematical vulnerability of traditional encryption algorithms is avoided; through RSA mapping + ChaCha20 chain encryption, key guessing and man-in-the-middle attacks are prevented; through ECDSA + Serpent + cyclic shift triple encryption, combined with blockchain evidence storage, financial-level security requirements are met.

[0058] A preferred embodiment of a virtual machine intelligent deployment system based on dynamic resource prediction of the present invention includes the following modules: A resource demand prediction model creation module is used for the server to create a resource demand prediction model based on the input module, the local feature extraction module, the temporal dependency feature extraction module, the semantic feature extraction module, the feature fusion module, and the output module, and to set the loss function of the resource demand prediction model based on the mean square error function, the mean absolute error function, and the mean absolute percentage error function; By setting up a resource demand prediction model, local features, temporal dependency features and semantic features are extracted through the local feature extraction module, temporal dependency feature extraction module and semantic feature extraction module respectively, thereby greatly improving the feature extraction capability and effectively mining potential patterns from the operating data; for example, the deployed virtual machine is used for e-commerce operations. The semantic feature extraction module extracts semantic features from news data and determines that a shopping festival for promotion will be held in the near future. This means that the load of the virtual machine will increase suddenly in the near future, and the corresponding virtual machine deployment parameters need to be adjusted to cope with it, so as to ensure the reliable operation of the virtual machine, thereby greatly improving the reliability of virtual machine deployment.

[0059] A data set construction module is used for the server to obtain historical operation data of a large number of virtual machines, including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data, and news data, and to construct a data set after preprocessing and annotating each of the historical operation data; In addition to traditional performance indicators, by innovatively introducing news data (such as industry dynamics and market events) as input features, the BERT model is used to capture the potential impact of external events on resource demand, achieving forward-looking predictions.

[0060] A resource demand prediction model training module is used for the server to train the resource demand prediction model based on the data set and the loss function, and deploy the trained resource demand prediction model; A virtual machine deployment instruction verification module is used for the server to obtain input virtual machine deployment instructions, verify the virtual machine deployment instructions and parse them to obtain virtual machine deployment parameters; A virtual machine automatic deployment module, configured to cause the server to automatically execute a virtual machine deployment operation based on the virtual machine deployment parameters and set a priority for each virtual machine to be deployed; The resource demand prediction module is used to collect real-time operation data from the server and input the real-time operation data into the deployed resource demand prediction model to obtain resource demand prediction results; A deployment parameter dynamic adjustment module is used for the server to dynamically adjust the virtual machine deployment parameters of the deployed virtual machines based on the resource demand prediction results, priority and Nash equilibrium algorithm; A deployment log management module is used for the server to record virtual machine deployment logs in real time and encrypt and store the virtual machine deployment logs; Through innovative multi-modal prediction, intelligent dynamic scheduling, multi-level security protection and automated deployment mechanisms, it effectively solves the pain points of rigid resource allocation, low prediction accuracy and weak security protection in traditional virtual machine deployment.

[0061] Through interdisciplinary technology integration (cloud computing, AI, cryptography, game theory) and full life cycle security design, problems such as resource rigidity, prediction lag, and security risks in traditional virtual machine deployment are solved.

[0062] In the resource demand prediction model creation module, the input module is constructed based on the normalization layer, and is used to perform a normalization operation on the input operation data; the local feature extraction module is constructed based on the convolutional neural network, and is used to extract local features from the operation data; the temporal dependency feature extraction module is constructed based on the bidirectional long short-term memory network and the self-attention network, and is used to extract temporal dependency features from the operation data; the semantic feature extraction module is constructed based on the BERT network, and is used to extract semantic features from the operation data; the feature fusion module is constructed based on the weighted bidirectional feature pyramid network, and is used to fuse local features, temporal dependency features and semantic features to obtain fused features; the output module is constructed based on the fully connected layer, and is used to output the resource demand forecast results within a preset time range in the future based on the fused features; By setting up a feature fusion module based on a weighted bidirectional feature pyramid network, the weighted bidirectional feature pyramid network extracts multi-scale features through convolutional layers of different scales (such as dilated convolutions with different dilation rates), and fuses high-level semantic features with low-level detail features through bottom-up and top-down paths. Different weights are assigned to features of each scale during the fusion process. By learning these weights to optimize the fusion effect, it is possible to capture both high-level semantic information and low-level detail information at the same time. Combined with the weighting mechanism, the contribution of features of each scale is dynamically adjusted to improve the fusion effect, thereby greatly improving the accuracy of resource demand prediction.

[0063] The formula of the loss function is: L total =a*L MSE +b*L MAE +c*L MAPE ; Among them, L total Represents the loss value of the loss function; L MSE represents the mean square error function; L MAE represents the mean absolute error function; L MAPE represents the mean absolute percentage error function; a, b, and c represent weight coefficients.

[0064] The loss function of the resource demand prediction model is set by the mean square error function, the mean absolute error function and the mean absolute percentage error function, so that the loss function combines the advantages of the mean square error function, the mean absolute error function and the mean absolute percentage error function, thereby greatly improving the training effect of the resource demand prediction model and greatly improving the accuracy of resource demand prediction.

[0065] By extracting local features through CNN, capturing temporal dependencies through bidirectional LSTM+self-attention, and extracting semantic features through BERT, and combining with the weighted bidirectional feature pyramid network to fuse multi-dimensional features, the prediction accuracy is significantly improved and it can adapt to complex scenarios. By fusing the three error functions of MSE, MAE, and MAPE, the sensitivity of the model to different errors is dynamically adjusted through the weight coefficient, avoiding the deviation of a single indicator and effectively enhancing the robustness of the model.

[0066] The dataset construction module is specifically used for: The server obtains historical operation data of a large number of virtual machines, including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data and news data; the virtual machine configuration data includes at least CPU configuration, memory configuration, storage configuration, network configuration, and operating system configuration; the virtual machine business data includes at least business type, business process and transaction data; the virtual machine performance data includes at least CPU usage rate, CPU utilization rate, CPU waiting time, CPU context switch times, CPU instruction execution times, real-time memory usage rate, average memory usage rate, memory utilization rate, memory swap rate, memory page fault rate, disk read and write speed, disk response time, network bandwidth usage rate, network throughput, network packet loss rate and network delay; the virtual machine status data includes operation status and availability status; the operation status is running, paused or closed; the availability status is normal, warning or fault; A data set is constructed using historical operating data including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data, and news data; virtual machine configuration data includes at least CPU configuration, memory configuration, storage configuration, network configuration, and operating system configuration; virtual machine business data includes at least business type, business process, and transaction data; virtual machine performance data includes at least CPU usage, CPU utilization, CPU wait time, CPU context switch times, CPU instruction execution times, real-time memory usage, average memory usage, memory utilization, memory swap rate, memory page fault rate, disk read and write speed, disk response time, network bandwidth utilization, network throughput, network packet loss rate, and network latency; virtual machine status data includes operating status and availability status; covering internal operating status and external environmental factors, the data set can be used for multi-dimensional training and analysis in the future to effectively improve the generalization ability of the resource demand prediction model.

[0067] Each of the historical operation data is preprocessed, including at least text cleaning, text standardization, word segmentation, part-of-speech tagging, and sequence filling, and resource requirements are annotated on each of the preprocessed historical operation data, and a data set is constructed based on the annotated historical operation data.

[0068] By preprocessing each historical operation data, including at least text cleaning, text standardization, word segmentation, part-of-speech tagging, and sequence filling, the quality of the dataset can be effectively improved, the validity and consistency of the model input can be ensured, and the training effect of the resource demand prediction model can be effectively improved.

[0069] The resource demand prediction model training module is specifically used to: The dataset is divided into a training set, a validation set, and a test set based on a k-fold cross-validation method. The resource demand prediction model is trained using the training set. During the training process, adversarial samples are generated and added to the training set. The learning rate, batch size, optimizer, regularization parameter, and network structure of the resource demand prediction model are continuously optimized until the loss value of the loss function is less than a preset loss threshold. The trained resource demand prediction model is verified using the validation set to determine whether the prediction accuracy is greater than a preset accuracy threshold. If not, the training set is expanded to continue training. If so, then: The resource demand prediction model that has passed the verification is tested using the test set to determine whether the mean absolute percentage error is less than a preset error threshold. If not, the training set is expanded to continue training; if so, the training is terminated and the trained resource demand prediction model is deployed using a Docker container.

[0070] The dataset is divided into training set, validation set and test set through the k-fold cross-validation method to reduce the accidental influence caused by single data segmentation and effectively reduce the risk of overfitting. Adversarial samples are generated and added to the training set during the training process to improve the model's anti-interference ability and adapt to sudden traffic scenarios. The learning rate, batch size, optimizer, regularization parameters and network structure of the resource demand prediction model are continuously optimized until the loss value of the loss function is less than the preset loss threshold; then the trained resource demand prediction model is verified through the validation set, and the verified resource demand prediction model is tested through the test set. That is, optimization, verification and testing are continuously carried out during the training process, which greatly improves the accuracy of resource demand prediction.

[0071] Through the Docker container deployment model, it supports orchestration tools such as Kubernetes to achieve elastic scaling of resources and cross-platform compatibility.

[0072] The virtual machine deployment instruction verification module is specifically used to: The server obtains an input virtual machine deployment instruction carrying a parameter ciphertext, an authentication ciphertext, and a hash value; the parameter ciphertext is obtained by encrypting the virtual machine deployment parameters, and the encryption process is specifically as follows: encrypting the virtual machine deployment parameters using the AES algorithm to obtain first-level encrypted data, converting the first-level encrypted data into binary data, swapping the 1s and 0s in the binary data to obtain second-level encrypted data, and encrypting the second-level encrypted data into parameter ciphertext using the RC6 algorithm; the virtual machine deployment parameters include at least the number of CPU cores, CPU frequency, memory size, disk capacity, disk type, storage location, network interface, network bandwidth, and operating system; the authentication ciphertext is obtained by encrypting the username, password, device serial number, and timestamp, and the encryption process is specifically as follows: encrypting the username, password, device serial number, and timestamp into first-level encrypted data using the RSA algorithm, mapping a string of a specified position and specified length in the first-level encrypted data using a preset mapping rule to obtain second-level encrypted data, and encrypting the second-level encrypted data into authentication ciphertext using the ChaCha20 algorithm; the hash value is obtained by hashing the parameter ciphertext and the authentication ciphertext using the BLAKE2 algorithm; By setting the virtual machine deployment instruction to carry parameter ciphertext, authentication ciphertext and hash value, the parameter ciphertext is obtained by encrypting the virtual machine deployment parameters through the AES algorithm, base conversion adjustment, and RC6 algorithm. The authentication ciphertext is obtained by encrypting the username, password, device serial number and timestamp through the RSA algorithm, mapping rules, and ChaCha20 algorithm. Subsequently, multi-dimensional verification can be performed through the hash value, username, password, device serial number and timestamp, and the relevant data can be prevented from being stolen in plain text during transmission, thereby greatly improving the security of virtual machine deployment instruction transmission, and also avoiding the execution of virtual machine deployment operations based on illegal instructions, thereby greatly improving the security of virtual machine deployment.

[0073] The server parses the received virtual machine deployment instruction to obtain a parameter ciphertext, an authentication ciphertext, and a hash value; after performing an integrity check on the parameter ciphertext and the authentication ciphertext using the hash value, the server decrypts the authentication ciphertext into a second-layer encrypted data using the ChaCha20 algorithm; maps a string of a specified position and a specified length in the second-layer encrypted data using the mapping rule to obtain a first-layer encrypted data; and decrypts the first-layer encrypted data into a user name and password, a device serial number, and a timestamp using the RSA algorithm; The server performs a validity check based on the timestamp, performs a validity check based on the device serial number, and then performs a matching check on the username and password based on the preset user management table. If the check passes: The parameter ciphertext is decrypted into secondary encrypted data using the RC6 algorithm, the 1s and 0s in the secondary encrypted data are swapped, and the binary data is inversely converted to obtain primary encrypted data, and the primary encrypted data is decrypted using the AES algorithm to obtain the virtual machine deployment parameters.

[0074] The virtual machine automatic deployment module is specifically used to: The server automatically executes the virtual machine deployment operation through RPA based on the virtual machine deployment parameters, and sets the priority of each virtual machine based on the business type of each deployed virtual machine; The resource demand prediction module is specifically used to: The server collects real-time operation data including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data and news data, pre-processes the real-time operation data and inputs it into the deployed resource demand prediction model to obtain resource demand prediction results; The deployment parameter dynamic adjustment module is specifically used to: The server analyzes the resource demand prediction result, determines whether the resource demand of each deployed virtual machine changes within a preset time range in the future, and if so, determines whether the server can meet the changed resource demand. If so, dynamically adjusts the virtual machine deployment parameters of the deployed virtual machine based on the Nash equilibrium algorithm; if not, screens the deployed virtual machines based on the priority, and dynamically adjusts the virtual machine deployment parameters of the screened virtual machines based on the Nash equilibrium algorithm; By using the Nash balancing algorithm to dynamically adjust virtual machine deployment parameters, efficient resource utilization, load balancing, performance optimization, and cost control can be achieved, while improving the flexibility, adaptability, and reliability of the system.

[0075] By setting priorities based on business types, key business resources are ensured and system reliability and responsiveness are improved. Virtual machine deployment parameters are dynamically adjusted based on resource demand forecast results, priorities, and the Nash equilibrium algorithm. This uses game theory to optimize resource allocation strategies, balance resource competition between virtual machines, and maximize overall resource utilization.

[0076] The deployment log management module is specifically used to: The server records virtual machine deployment logs in real time through Elastic Observability, creates a pair of public and private keys using the ECDSA algorithm, encrypts the virtual machine deployment logs using the private key to obtain first encrypted data, cuts the last three characters of the private key to the middle to obtain the key, encrypts the first encrypted data and the key into second encrypted data using the Serpent algorithm, shifts each character of the second encrypted data left by 8 bits to obtain an encrypted log, stores the encrypted log in a specified path, and backs up the encrypted log to the blockchain.

[0077] A pair of public and private keys is created using the ECDSA algorithm. The virtual machine deployment log is encrypted using the private key to obtain the first encrypted data. The last three characters of the private key are cut to the middle to obtain the key. The first encrypted data and the key are encrypted into the second encrypted data using the Serpent algorithm. Each character of the second encrypted data is circularly shifted 8 bits to the left to obtain the encrypted log. The encrypted log is stored in a specified path and backed up to the blockchain. At least five security measures are combined (public and private keys, private key encryption, Serpent algorithm, character shifting, and blockchain), greatly improving the security of virtual machine deployment log storage.

[0078] Through AES-RC6 double encryption + binary inversion design, the mathematical vulnerability of traditional encryption algorithms is avoided; through RSA mapping + ChaCha20 chain encryption, key guessing and man-in-the-middle attacks are prevented; through ECDSA + Serpent + cyclic shift triple encryption, combined with blockchain evidence storage, financial-level security requirements are met.

[0079] Although the specific embodiments of the present invention are described above, those skilled in the art should understand that the specific embodiments described are merely illustrative and are not intended to limit the scope of the present invention. Equivalent modifications and changes made by those skilled in the art in accordance with the spirit of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A method for intelligent deployment of virtual machines based on dynamic resource prediction, characterized by: The steps include: Step S1: The server creates a resource demand prediction model based on the input module, the local feature extraction module, the temporal dependency feature extraction module, the semantic feature extraction module, the feature fusion module, and the output module, and sets a loss function of the resource demand prediction model based on a mean square error function, a mean absolute error function, and a mean absolute percentage error function; Step S2: The server obtains historical operation data of a large number of virtual machines, including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data, and news data, and constructs a data set after preprocessing and annotating each of the historical operation data; Step S3: The server trains the resource demand prediction model based on the data set and the loss function, and deploys the trained resource demand prediction model; Step S4: The server obtains the input virtual machine deployment instruction, verifies the virtual machine deployment instruction, and parses it to obtain virtual machine deployment parameters; Step S5: The server automatically executes the virtual machine deployment operation based on the virtual machine deployment parameters and sets the priority of each deployed virtual machine; Step S6: The server collects real-time operation data and inputs the real-time operation data into the deployed resource demand prediction model to obtain a resource demand prediction result; Step S7: The server dynamically adjusts the virtual machine deployment parameters of the deployed virtual machines based on the resource demand prediction result, priority, and Nash equilibrium algorithm; Step S8: The server records the virtual machine deployment log in real time and encrypts and stores the virtual machine deployment log; In step S1, the input module is constructed based on a normalization layer and is used to perform a normalization operation on the input operation data; the local feature extraction module is constructed based on a convolutional neural network and is used to extract local features from the operation data; the temporal dependency feature extraction module is constructed based on a bidirectional long short-term memory network and a self-attention network and is used to extract temporal dependency features from the operation data; the semantic feature extraction module is constructed based on a BERT network and is used to extract semantic features from the operation data; the feature fusion module is constructed based on a weighted bidirectional feature pyramid network and is used to fuse local features, temporal dependency features and semantic features to obtain fused features; the output module is constructed based on a fully connected layer and is used to output a resource demand forecast result within a preset time range in the future based on the fused features; The formula of the loss function is: L total =a*L MSE +b*L MAE +c*L MAPE ; Among them, L total Represents the loss value of the loss function; L MSE represents the mean square error function; L MAE represents the mean absolute error function; L MAPE represents the mean absolute percentage error function; a, b, and c represent weight coefficients.

2. The method for intelligent deployment of virtual machines based on dynamic resource prediction according to claim 1, characterized in that: The step S2 is specifically as follows: The server obtains historical operation data of a large number of virtual machines, including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data and news data; the virtual machine configuration data includes at least CPU configuration, memory configuration, storage configuration, network configuration, and operating system configuration; the virtual machine business data includes at least business type, business process and transaction data; the virtual machine performance data includes at least CPU usage rate, CPU utilization rate, CPU waiting time, CPU context switch times, CPU instruction execution times, real-time memory usage rate, average memory usage rate, memory utilization rate, memory swap rate, memory page fault rate, disk read and write speed, disk response time, network bandwidth usage rate, network throughput, network packet loss rate and network delay; the virtual machine status data includes operation status and availability status; the operation status is running, paused or closed; the availability status is normal, warning or fault; Each of the historical operation data is preprocessed, including at least text cleaning, text standardization, word segmentation, part-of-speech tagging, and sequence filling, and resource requirements are annotated on each of the preprocessed historical operation data, and a data set is constructed based on the annotated historical operation data.

3. The method for intelligent deployment of virtual machines based on dynamic resource prediction according to claim 1, characterized in that: The step S3 is specifically as follows: The dataset is divided into a training set, a validation set, and a test set based on a k-fold cross-validation method. The resource demand prediction model is trained using the training set. During the training process, adversarial samples are generated and added to the training set. The learning rate, batch size, optimizer, regularization parameter, and network structure of the resource demand prediction model are continuously optimized until the loss value of the loss function is less than a preset loss threshold. The trained resource demand prediction model is verified using the validation set to determine whether the prediction accuracy is greater than a preset accuracy threshold. If not, the training set is expanded to continue training. If so, then: The resource demand prediction model that has passed the verification is tested using the test set to determine whether the mean absolute percentage error is less than a preset error threshold. If not, the training set is expanded to continue training; if so, the training is terminated and the trained resource demand prediction model is deployed using a Docker container.

4. The method for intelligent deployment of virtual machines based on dynamic resource prediction according to claim 1, wherein: The step S4 is specifically as follows: The server obtains an input virtual machine deployment instruction carrying a parameter ciphertext, an authentication ciphertext, and a hash value; the parameter ciphertext is obtained by encrypting the virtual machine deployment parameters, and the encryption process is specifically as follows: encrypting the virtual machine deployment parameters using the AES algorithm to obtain first-level encrypted data, converting the first-level encrypted data into binary data, swapping the 1s and 0s in the binary data to obtain second-level encrypted data, and encrypting the second-level encrypted data into parameter ciphertext using the RC6 algorithm; the virtual machine deployment parameters include at least the number of CPU cores, CPU frequency, memory size, disk capacity, disk type, storage location, network interface, network bandwidth, and operating system; the authentication ciphertext is obtained by encrypting the username, password, device serial number, and timestamp, and the encryption process is specifically as follows: encrypting the username, password, device serial number, and timestamp into first-level encrypted data using the RSA algorithm, mapping a string of a specified position and specified length in the first-level encrypted data using a preset mapping rule to obtain second-level encrypted data, and encrypting the second-level encrypted data into authentication ciphertext using the ChaCha20 algorithm; the hash value is obtained by hashing the parameter ciphertext and the authentication ciphertext using the BLAKE2 algorithm; The server parses the received virtual machine deployment instruction to obtain a parameter ciphertext, an authentication ciphertext, and a hash value; after performing an integrity check on the parameter ciphertext and the authentication ciphertext using the hash value, the server decrypts the authentication ciphertext into a second-layer encrypted data using the ChaCha20 algorithm; maps a string of a specified position and a specified length in the second-layer encrypted data using the mapping rule to obtain a first-layer encrypted data; and decrypts the first-layer encrypted data into a user name and password, a device serial number, and a timestamp using the RSA algorithm; The server performs a validity check based on the timestamp, performs a validity check based on the device serial number, and then performs a matching check on the username and password based on the preset user management table. If the check passes: The parameter ciphertext is decrypted into secondary encrypted data using the RC6 algorithm, the 1s and 0s in the secondary encrypted data are swapped, and the binary data is inversely converted to obtain primary encrypted data, and the primary encrypted data is decrypted using the AES algorithm to obtain the virtual machine deployment parameters.

5. The method for intelligent deployment of virtual machines based on dynamic resource prediction according to claim 1, characterized in that: The step S5 is specifically as follows: The server automatically executes the virtual machine deployment operation through RPA based on the virtual machine deployment parameters, and sets the priority of each virtual machine based on the business type of each deployed virtual machine; The step S6 is specifically as follows: The server collects real-time operation data including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data and news data, pre-processes the real-time operation data and inputs it into the deployed resource demand prediction model to obtain resource demand prediction results; The step S7 is specifically as follows: The server analyzes the resource demand prediction result, determines whether the resource demand of each deployed virtual machine changes within a preset time range in the future, and if so, determines whether the server can meet the changed resource demand. If so, dynamically adjusts the virtual machine deployment parameters of the deployed virtual machine based on the Nash equilibrium algorithm; if not, screens the deployed virtual machines based on the priority, and dynamically adjusts the virtual machine deployment parameters of the screened virtual machines based on the Nash equilibrium algorithm; The step S8 is specifically as follows: The server records virtual machine deployment logs in real time through Elastic Observability, creates a pair of public and private keys using the ECDSA algorithm, encrypts the virtual machine deployment logs using the private key to obtain first encrypted data, cuts the last three characters of the private key to the middle to obtain the key, encrypts the first encrypted data and the key into second encrypted data using the Serpent algorithm, shifts each character of the second encrypted data left by 8 bits to obtain an encrypted log, stores the encrypted log in a specified path, and backs up the encrypted log to the blockchain.

6. A virtual machine intelligent deployment system based on dynamic resource prediction, characterized by: Includes the following modules: A resource demand prediction model creation module is used for the server to create a resource demand prediction model based on the input module, the local feature extraction module, the temporal dependency feature extraction module, the semantic feature extraction module, the feature fusion module, and the output module, and to set the loss function of the resource demand prediction model based on the mean square error function, the mean absolute error function, and the mean absolute percentage error function; A data set construction module is used for the server to obtain historical operation data of a large number of virtual machines, including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data, and news data, and to construct a data set after preprocessing and annotating each of the historical operation data; A resource demand prediction model training module is used for the server to train the resource demand prediction model based on the data set and the loss function, and deploy the trained resource demand prediction model; A virtual machine deployment instruction verification module is used for the server to obtain input virtual machine deployment instructions, verify the virtual machine deployment instructions and parse them to obtain virtual machine deployment parameters; A virtual machine automatic deployment module, configured to cause the server to automatically execute a virtual machine deployment operation based on the virtual machine deployment parameters and set a priority for each virtual machine to be deployed; The resource demand prediction module is used to collect real-time operation data from the server and input the real-time operation data into the deployed resource demand prediction model to obtain resource demand prediction results; A deployment parameter dynamic adjustment module is used for the server to dynamically adjust the virtual machine deployment parameters of the deployed virtual machines based on the resource demand prediction results, priority and Nash equilibrium algorithm; A deployment log management module is used for the server to record virtual machine deployment logs in real time and encrypt and store the virtual machine deployment logs; In the resource demand prediction model creation module, the input module is constructed based on the normalization layer, and is used to perform a normalization operation on the input operation data; the local feature extraction module is constructed based on the convolutional neural network, and is used to extract local features from the operation data; the temporal dependency feature extraction module is constructed based on the bidirectional long short-term memory network and the self-attention network, and is used to extract temporal dependency features from the operation data; the semantic feature extraction module is constructed based on the BERT network, and is used to extract semantic features from the operation data; the feature fusion module is constructed based on the weighted bidirectional feature pyramid network, and is used to fuse local features, temporal dependency features and semantic features to obtain fused features; the output module is constructed based on the fully connected layer, and is used to output the resource demand forecast results within a preset time range in the future based on the fused features; The formula of the loss function is: L total =a*L MSE +b*L MAE +c*L MAPE ; Among them, L total Represents the loss value of the loss function; L MSE represents the mean square error function; L MAE represents the mean absolute error function; L MAPE represents the mean absolute percentage error function; a, b, and c represent weight coefficients.

7. The virtual machine intelligent deployment system based on dynamic resource prediction according to claim 6, characterized in that: The dataset construction module is specifically used for: The server obtains historical operation data of a large number of virtual machines, including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data and news data; the virtual machine configuration data includes at least CPU configuration, memory configuration, storage configuration, network configuration, and operating system configuration; the virtual machine business data includes at least business type, business process and transaction data; the virtual machine performance data includes at least CPU usage rate, CPU utilization rate, CPU waiting time, CPU context switch times, CPU instruction execution times, real-time memory usage rate, average memory usage rate, memory utilization rate, memory swap rate, memory page fault rate, disk read and write speed, disk response time, network bandwidth usage rate, network throughput, network packet loss rate and network delay; the virtual machine status data includes operation status and availability status; the operation status is running, paused or closed; the availability status is normal, warning or fault; Each of the historical operation data is preprocessed, including at least text cleaning, text standardization, word segmentation, part-of-speech tagging, and sequence filling, and resource requirements are annotated on each of the preprocessed historical operation data, and a data set is constructed based on the annotated historical operation data.

8. The virtual machine intelligent deployment system based on dynamic resource prediction according to claim 6, characterized in that: The resource demand prediction model training module is specifically used to: The dataset is divided into a training set, a validation set, and a test set based on a k-fold cross-validation method. The resource demand prediction model is trained using the training set. During the training process, adversarial samples are generated and added to the training set. The learning rate, batch size, optimizer, regularization parameter, and network structure of the resource demand prediction model are continuously optimized until the loss value of the loss function is less than a preset loss threshold. The trained resource demand prediction model is verified using the validation set to determine whether the prediction accuracy is greater than a preset accuracy threshold. If not, the training set is expanded to continue training. If so, then: The resource demand prediction model that has passed the verification is tested using the test set to determine whether the mean absolute percentage error is less than a preset error threshold. If not, the training set is expanded to continue training; if so, the training is terminated and the trained resource demand prediction model is deployed using a Docker container.

9. The virtual machine intelligent deployment system based on dynamic resource prediction according to claim 6, characterized in that: The virtual machine deployment instruction verification module is specifically used to: The server obtains an input virtual machine deployment instruction carrying a parameter ciphertext, an authentication ciphertext, and a hash value; the parameter ciphertext is obtained by encrypting the virtual machine deployment parameters, and the encryption process is specifically as follows: encrypting the virtual machine deployment parameters using the AES algorithm to obtain first-level encrypted data, converting the first-level encrypted data into binary data, swapping the 1s and 0s in the binary data to obtain second-level encrypted data, and encrypting the second-level encrypted data into parameter ciphertext using the RC6 algorithm; the virtual machine deployment parameters include at least the number of CPU cores, CPU frequency, memory size, disk capacity, disk type, storage location, network interface, network bandwidth, and operating system; the authentication ciphertext is obtained by encrypting the username, password, device serial number, and timestamp, and the encryption process is specifically as follows: encrypting the username, password, device serial number, and timestamp into first-level encrypted data using the RSA algorithm, mapping a string of a specified position and specified length in the first-level encrypted data using a preset mapping rule to obtain second-level encrypted data, and encrypting the second-level encrypted data into authentication ciphertext using the ChaCha20 algorithm; the hash value is obtained by hashing the parameter ciphertext and the authentication ciphertext using the BLAKE2 algorithm; The server parses the received virtual machine deployment instruction to obtain a parameter ciphertext, an authentication ciphertext, and a hash value; after performing an integrity check on the parameter ciphertext and the authentication ciphertext using the hash value, the server decrypts the authentication ciphertext into a second-layer encrypted data using the ChaCha20 algorithm; maps a string of a specified position and a specified length in the second-layer encrypted data using the mapping rule to obtain a first-layer encrypted data; and decrypts the first-layer encrypted data into a user name and password, a device serial number, and a timestamp using the RSA algorithm; The server performs a validity check based on the timestamp, performs a validity check based on the device serial number, and then performs a matching check on the username and password based on the preset user management table. If the check passes: The parameter ciphertext is decrypted into secondary encrypted data using the RC6 algorithm, the 1s and 0s in the secondary encrypted data are swapped, and the binary data is inversely converted to obtain primary encrypted data, and the primary encrypted data is decrypted using the AES algorithm to obtain the virtual machine deployment parameters.

10. The virtual machine intelligent deployment system based on dynamic resource prediction according to claim 6, characterized in that: The virtual machine automatic deployment module is specifically used to: The server automatically executes the virtual machine deployment operation through RPA based on the virtual machine deployment parameters, and sets the priority of each virtual machine based on the business type of each deployed virtual machine; The resource demand prediction module is specifically used to: The server collects real-time operation data including virtual machine configuration data, virtual machine business data, virtual machine performance data, virtual machine status data and news data, pre-processes the real-time operation data and inputs it into the deployed resource demand prediction model to obtain resource demand prediction results; The deployment parameter dynamic adjustment module is specifically used to: The server analyzes the resource demand prediction result, determines whether the resource demand of each deployed virtual machine changes within a preset time range in the future, and if so, determines whether the server can meet the changed resource demand. If so, dynamically adjusts the virtual machine deployment parameters of the deployed virtual machine based on the Nash equilibrium algorithm; if not, screens the deployed virtual machines based on the priority, and dynamically adjusts the virtual machine deployment parameters of the screened virtual machines based on the Nash equilibrium algorithm; The deployment log management module is specifically used to: The server records virtual machine deployment logs in real time through Elastic Observability, creates a pair of public and private keys using the ECDSA algorithm, encrypts the virtual machine deployment logs using the private key to obtain first encrypted data, cuts the last three characters of the private key to the middle to obtain the key, encrypts the first encrypted data and the key into second encrypted data using the Serpent algorithm, shifts each character of the second encrypted data left by 8 bits to obtain an encrypted log, stores the encrypted log in a specified path, and backs up the encrypted log to the blockchain.

Images