Cloud-edge collaborative intelligent storage node dynamic deployment method and system

Abstract

The invention discloses a cloud-edge collaborative intelligent storage node dynamic deployment method and system. The method comprises the following steps: monitoring performance indexes such as edge node data traffic and storage resource state in real time; a deep learning algorithm combining time sequence analysis and an LSTM neural network is adopted to analyze traffic information, and a data access demand is predicted; edge nodes and cloud storage resource configuration are dynamically adjusted based on a prediction result, and an intelligent scheduling algorithm, a data cold and hot separation strategy and a self-adaptive fragmentation technology are introduced to allocate resources; the storage node layout is optimized in real time, and an optimal data distribution path is selected through a reinforcement learning strategy; data security and access control are realized by adopting end-to-end encryption, multi-level access control and block chain technologies. The system comprises a monitoring acquisition module, a prediction analysis module, a resource scheduling module, a path optimization module and a security control module. According to the method, the utilization efficiency of storage resources is improved, data delay is reduced, system stability and data security are enhanced, and the method is suitable for a cloud edge collaborative storage scene.

Description

Technical Field

[0001] The present invention relates to the application field of intelligent storage management in collaboration between cloud computing and edge computing, and in particular to a method and system for dynamic deployment of intelligent storage nodes in cloud-edge collaboration. Background Art

[0002] In scenarios where cloud computing and edge computing are combined, with the explosive growth of data volumes, effectively managing storage resources and enabling dynamic deployment and optimized allocation are key to ensuring system performance. Existing technologies have already achieved basic cloud-edge storage collaboration, providing initial support for big data processing.

[0003] Common data management usually relies on static configuration or dynamic adjustment mechanisms based on simple rules to manage storage nodes. However, these methods often cannot respond to changes in data traffic in real time, resulting in low storage resource utilization and increased data latency. Especially in the case of high concurrent access, the system stability and data security are insufficient and cannot meet the working requirements of data management applications. Therefore, a cloud-edge collaborative intelligent storage node dynamic deployment method and system are proposed. Summary of the Invention

[0004] The present invention provides the following technical solution: a cloud-edge collaborative intelligent storage node dynamic deployment method, comprising the following steps: S1 real-time monitoring and data collection: First, continuously monitor the data traffic and storage resource status of edge nodes, collecting performance indicators including but not limited to data traffic rate, storage space utilization, and node processing capacity; S2 traffic pattern analysis and prediction: Using machine learning and a deep learning algorithm that combines time series analysis with an LSTM neural network model to analyze the data traffic information collected in step S1, obtain historical traffic patterns and predict future data access needs; S3 dynamic resource configuration: Based on the prediction results obtained in step S2, the storage resource configuration of the edge node and the cloud is automatically adjusted, including increasing and decreasing the storage capacity of the edge node, adjusting the data caching strategy, and optimizing the data transmission path. At the same time, intelligent scheduling algorithms, data hot and cold separation strategies, and adaptive sharding technologies are introduced to dynamically allocate computing resources and storage resources according to the real-time node load. S4 real-time optimization and layout adjustment: Optimize storage node layout in real time, select the optimal data distribution path using a reinforcement learning-based strategy, and dynamically adjust data distribution between edge nodes and the cloud through algorithms; S5 Data Security and Access Control: End-to-end encryption technology and multi-level access control mechanisms are used during the data transmission process, including data transmission encryption, dynamic management of access rights, and real-time updates of security policies. Security audits and vulnerability scans are conducted regularly, and encryption standards and access control rules are updated in a timely manner. Blockchain technology is also introduced to keep tamper-proof records of data access logs.

[0005] The present invention provides a cloud-edge collaborative intelligent storage node dynamic deployment system, which adopts the above-mentioned cloud-edge collaborative intelligent storage node dynamic deployment method, including: Monitoring and collection module, prediction and analysis module, resource scheduling module, path optimization module and security control module. The monitoring and collection module is deployed at the edge node and is used to collect data flow rate, storage space utilization and node processing capacity indicators in real time; A prediction and analysis module, used to analyze historical traffic patterns and predict future demand. The prediction and analysis module is connected to the monitoring and acquisition module. The prediction and analysis module has a built-in hybrid prediction model and an architecture based on an LSTM neural network. The resource scheduling module includes an intelligent scheduling unit and a hot and cold separation unit; A resource scheduling module is used to dynamically adjust edge nodes and cloud storage resources based on prediction results. The path optimization module is connected to the resource scheduling module. The path optimization module builds a state transition model based on the Q-learning algorithm, combines the improved Dijkstra algorithm to calculate the optimal multi-hop transmission path in real time, and dynamically generates a topology-aware routing table; The security control module is used to implement end-to-end encryption and multi-level access control. The security control module integrates an encryption unit, an access control unit and an audit unit. The encryption unit adopts a double-layer encryption mechanism of the national secret SM4 and Diffie-Hellman protocol. The access control unit dynamically generates strategies based on attributes, and the audit unit realizes on-chain storage of operation logs through blockchain smart contracts.

[0006] Preferably, the real-time monitoring and data collection in step S1 further includes deploying a lightweight monitoring agent on the edge node to achieve comprehensive perception of the edge node's operating status. The monitoring agent adopts a layered architecture design, including a data acquisition layer, a data processing layer, and a data transmission layer.

[0007] Preferably, in step S2, when constructing a hybrid prediction model, a long short-term memory network is preferably used as the basic architecture, and the traffic data is decomposed into trend terms, seasonal terms and residual terms through time series decomposition technology. The training data set is dynamically updated in combination with the sliding window mechanism. The attention mechanism is introduced in the model training process to enhance the ability to capture sudden traffic characteristics, and rapid adaptation of cross-regional traffic patterns is achieved through transfer learning.

[0008] Preferably, in step S3, the storage capacity adjustment strategy adopts a hierarchical expansion mechanism. When it is detected that the node storage space utilization exceeds the preset threshold, the virtualized hierarchical expansion of the local storage medium is triggered first. If it exceeds the upper limit of the system elastic expansion, a virtual storage unit application is automatically initiated to the cloud. At the same time, a heat attenuation factor is introduced into the data caching strategy to dynamically adjust the cache level distribution according to the data access frequency.

[0009] Preferably, in step S4, the reinforcement learning strategy constructs a state transition model based on the Q-learning algorithm, sets the node load rate, link delay and storage cost as state space parameters, balances the decision weights of exploration and utilization through the ε-greedy exploration strategy, and combines the Dijkstra improved algorithm in the path selection process to calculate the optimal path of multi-hop transmission in real time, and dynamically generates a topology-aware routing table.

[0010] Preferably, in step S5, the end-to-end encryption adopts a double-layer encryption mechanism combining the national secret SM4 algorithm and the Diffie-Hellman key exchange protocol, constructs an attribute-based access control model in access permission management, and dynamically generates access policies through role inheritance and permission minimization principles. The security audit module adopts blockchain smart contract technology to realize on-chain storage of operation logs.

[0011] Preferably, the intelligent scheduling algorithm in step S3 implements a multi-dimensional resource perception mechanism, that is, it comprehensively considers the remaining storage space of the node, the current network congestion index and the task priority weight, and obtains the optimal solution for resource allocation through genetic algorithm optimization. The fitness function design in the genetic algorithm includes three dimensions: storage utilization, transmission energy consumption and service response time, and sets a dynamic penalty coefficient to constrain the generation of invalid solutions. At the same time, the hot and cold separation strategy adopts a dynamic hierarchical storage architecture, establishes a three-level storage pool according to the local characteristics of data access time, retains the hot data accessed in the last 24 hours in the high-performance SSD storage layer, migrates the data with a historical decrease in access frequency to the distributed mechanical hard disk storage layer, and transfers the data that exceeds the set cold storage period to the object storage service after compression. The data flow between the storage layers is realized through asynchronous migration threads.

[0012] Preferably, the intelligent scheduling unit of the resource scheduling module constructs a multi-dimensional resource perception matrix to track the storage space fragmentation rate, network link congestion index and task queue priority weight of the edge node in real time, and generates a resource allocation plan through a genetic algorithm optimization engine. The resource allocation plan includes a local storage tiering strategy and cloud elastic expansion instructions. The hot and cold separation unit deploys a three-level storage pool dynamic migration mechanism, which automatically divides the hot data storage area, warm data buffer area and cold data archiving area according to the timeliness characteristics of data access, and each storage interval maintains data consistency through an asynchronous replication thread.

[0013] Preferably, the audit unit of the security control module constructs a blockchain evidence chain, and its operation log generates a unique digital fingerprint after hash summary processing, and achieves log consistency confirmation among multiple verification nodes through the PBFT consensus algorithm.

[0014] In summary, compared with the prior art, the present invention provides a method and system for dynamic deployment of intelligent storage nodes in cloud-edge collaboration, which has the following beneficial effects: 1. This invention uses the monitoring and acquisition module to continuously monitor key indicators such as edge node data traffic rate and storage space utilization. Combined with the LSTM neural network-based deep learning algorithm and time series analysis in the predictive analysis module, this mechanism can accurately capture historical traffic patterns and predict future access needs. Compared to traditional static resource allocation methods, this mechanism can identify traffic peaks and troughs in advance. Through the resource scheduling module's intelligent scheduling algorithm and data hot and cold separation strategy, it dynamically adjusts the storage capacity, caching strategy, and sharding distribution of edge nodes and the cloud, improving deployment efficiency. 2. This invention utilizes a path optimization module that combines reinforcement learning with an improved Dijkstra algorithm to perceive node load rates, link latency, and network topology changes in real time. It dynamically generates a topology-aware routing table and selects the optimal data distribution path. In high-concurrency scenarios, this mechanism automatically avoids congested links and prioritizes data routing to low-load edge nodes, shortening data transmission distance and time. Furthermore, through real-time optimization and layout adjustments, it dynamically balances the load between nodes, avoiding response bottlenecks caused by single node overload, improving storage resource utilization, and reducing data latency. 3. This invention utilizes the security control module's end-to-end encryption technology, multi-level access control, and blockchain auditing mechanisms to provide comprehensive security protection covering data collection, transmission, storage, and access. Unlike traditional single-point encryption and fixed permissions management, this solution enables real-time updates of encryption standards, dynamic adjustments to access permissions, and tamper-proof storage of operation logs. This allows for proactive identification of potential attacks and updates to security policies, improving system stability and data security. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flow chart of the method of the present invention.

[0016] Figure 2 It is a system structure diagram of the present invention. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0018] See also Figure 1 The present invention provides a technical solution, a cloud-edge collaborative intelligent storage node dynamic deployment method, comprising the following steps: S1 real-time monitoring and data collection: First, the data traffic and storage resource status of edge nodes are continuously monitored, and performance indicators including but not limited to data traffic rate, storage space utilization, and node processing capacity are collected. Real-time monitoring and data collection further includes deploying lightweight monitoring agents on edge nodes to achieve comprehensive awareness of the operating status of edge nodes. The monitoring agents adopt a layered architecture design, including data collection layer, data processing layer, and data transmission layer. S2 traffic pattern analysis and prediction: Using machine learning and a deep learning algorithm that combines time series analysis with an LSTM neural network model to analyze the data traffic information collected in step S1, obtain historical traffic patterns and predict future data access needs; When building a hybrid prediction model, a long-short-term memory network is preferred as the basic architecture. Time series decomposition technology is used to break down traffic data into trend terms, seasonal terms, and residual terms. A sliding window mechanism is used to dynamically update the training dataset. An attention mechanism is introduced during model training to enhance the ability to capture sudden traffic characteristics. Transfer learning is used to achieve rapid adaptation of cross-regional traffic patterns. S3 dynamic resource configuration: Based on the prediction results obtained in step S2, the storage resource configuration of the edge node and the cloud is automatically adjusted, including increasing and decreasing the storage capacity of the edge node, adjusting the data caching strategy, and optimizing the data transmission path. At the same time, intelligent scheduling algorithms, data hot and cold separation strategies, and adaptive sharding technologies are introduced to dynamically allocate computing resources and storage resources according to the real-time node load. The storage capacity adjustment strategy adopts a hierarchical expansion mechanism. When it is detected that the node storage space utilization exceeds the preset threshold, the virtualized hierarchical expansion of the local storage medium is triggered first. If the upper limit of the system elastic expansion is exceeded, a virtual storage unit application is automatically initiated to the cloud. At the same time, a heat decay factor is introduced into the data caching strategy to dynamically adjust the cache tier distribution according to the data access frequency. The intelligent scheduling algorithm implements a multi-dimensional resource perception mechanism, which comprehensively considers the remaining storage space of the node, the current network congestion index and the task priority weight, and obtains the optimal solution for resource allocation through genetic algorithm optimization. The fitness function design in the genetic algorithm includes three dimensions: storage utilization, transmission energy consumption and service response time, and sets a dynamic penalty coefficient to constrain the generation of invalid solutions. At the same time, the hot and cold separation strategy adopts a dynamic hierarchical storage architecture, establishing a three-level storage pool based on the local characteristics of data access time. Hot data accessed in the last 24 hours is retained in the high-performance SSD storage layer, and data with a historical decrease in access frequency is migrated to the distributed mechanical hard disk storage layer. Data that exceeds the set cold storage period is compressed and transferred to the object storage service. Data flow between storage layers is achieved through asynchronous migration threads. S4 real-time optimization and layout adjustment: Optimize storage node layout in real time, select the optimal data distribution path using a reinforcement learning-based strategy, and dynamically adjust data distribution between edge nodes and the cloud through algorithms; The reinforcement learning strategy builds a state transition model based on the Q-learning algorithm, sets the node load rate, link delay and storage cost as state space parameters, and balances the decision weights of exploration and utilization through the ε-greedy exploration strategy. At the same time, the improved Dijkstra algorithm is combined in the path selection process to calculate the optimal path of multi-hop transmission in real time and dynamically generate a topology-aware routing table. The specific implementation process of the ascending method is as follows; Real-time status monitoring and data collection: The system deploys intelligent sensor modules to continuously track the operating status of edge nodes. The load rate of each node is calculated by comprehensively counting indicators such as the current storage space occupancy ratio, the number of task queue backlogs, and CPU processing efficiency. Link latency is measured by periodically sending detection data packets, measuring the round-trip time from the source node to the target node, and recording the fluctuation characteristics during periods of network congestion. Storage cost parameters include the energy consumption indicators of local storage devices and the billing unit price of cloud storage services. The system establishes a dynamic cost model based on historical data. All collected data is aggregated by the edge gateway to form a panoramic view of the node status, providing real-time input for subsequent decision-making; Reinforcement learning decision model construction: An intelligent decision-making engine is built based on the core algorithm of Q-learning. The system abstracts the node load rate, link latency and storage cost into a three-dimensional state space, and each state corresponds to a specific network operation scenario. The reward mechanism design adopts a multi-dimensional evaluation strategy: when the latency of the data transmission path is reduced or the storage resource utilization rate is improved, positive rewards are given; if congestion increases or costs exceed the limit, negative penalties are triggered. Through the ε-greedy exploration strategy, the system tries new paths (exploration) and reuses historical optimal paths (utilization) in a dynamic proportion, balancing short-term benefits and long-term strategy optimization. For example, during burst traffic peaks, low-latency paths are explored first; during stable operation, the focus is on minimizing storage costs; Path calculation and dynamic adjustment mechanism: When a data distribution request is received, the system calls the improved path calculation module. This module integrates a dynamic weight adjustment function to correct the link weight parameters based on real-time monitoring data: for example, the transmission delay of the congested link is multiplied by the dynamic penalty coefficient to reduce the probability of it being selected. Based on the corrected network topology, the system starts a multi-objective optimization algorithm to simultaneously evaluate indicators such as transmission delay, resource consumption and path reliability. Through a hierarchical screening mechanism, paths that do not meet the storage capacity constraints are first eliminated, and then the solution with the best overall performance is selected from the remaining options. After the path is selected, the system automatically triggers the data sharding strategy, dynamically divides the data block size according to the file type and access frequency, uses large-block continuous transmission for video streams with high real-time requirements, and divides database transaction data into small blocks and sends them in parallel; Dynamic generation of topology-aware routing tables: Path optimization results are synchronized to the routing table generator in real time. This module maintains dynamically updated topology mapping relationships. The routing table of each node contains the priority ranking, estimated latency, and bandwidth reservation of the currently available paths. When a network topology change is detected (such as a node failure or a new edge device), the system immediately triggers the routing table reconstruction process: first, the faulty node is isolated and its associated links are marked as unavailable; then, the optimal path in the affected area is recalculated, and the updated routing information is synchronized to adjacent nodes through a broadcast mechanism. To improve convergence speed, the routing table generator adopts an incremental update strategy, adjusting only the local paths affected by the change to avoid frequent resets of the global routing table; S5 Data Security and Access Control: End-to-end encryption technology and multi-level access control mechanisms are used during data transmission, including data transmission encryption, dynamic access rights management, and real-time updates of security policies. Regular security audits and vulnerability scans are conducted, and encryption standards and access control rules are updated in a timely manner. Blockchain technology is also introduced to ensure that data access logs are tamper-proof. End-to-end encryption uses a double-layer encryption mechanism that combines the national secret SM4 algorithm with the Diffie-Hellman key exchange protocol. An attribute-based access control model is constructed in access permission management, and access policies are dynamically generated through role inheritance and permission minimization principles. The security audit module uses blockchain smart contract technology to achieve on-chain storage of operation logs.

[0019] See also Figure 2 The present invention provides a cloud-edge collaborative intelligent storage node dynamic deployment system, which adopts the above-mentioned cloud-edge collaborative intelligent storage node dynamic deployment method, including: Monitoring and collection module, prediction and analysis module, resource scheduling module, path optimization module and security control module. The monitoring and collection module is deployed on the edge node to collect data flow rate, storage space utilization rate and node processing capacity indicators in real time; The predictive analysis module is used to analyze historical traffic patterns and predict future demand. The predictive analysis module is connected to the monitoring and acquisition module. The predictive analysis module has a built-in hybrid prediction model and an architecture based on the LSTM neural network. The resource scheduling module includes an intelligent scheduling unit and a hot and cold separation unit. The resource scheduling module is used to dynamically adjust edge nodes and cloud storage resources based on prediction results. The path optimization module is connected to the resource scheduling module. The path optimization module builds a state transition model based on the Q-learning algorithm and combines the improved Dijkstra algorithm to calculate the optimal multi-hop transmission path in real time and dynamically generate a topology-aware routing table. The security control module is used to implement end-to-end encryption and multi-level access control. The security control module integrates an encryption unit, an access control unit, and an audit unit. The encryption unit uses a dual-layer encryption mechanism based on the national secret SM4 and Diffie-Hellman protocols. The access control unit dynamically generates policies based on attributes, and the audit unit implements on-chain storage of operation logs through blockchain smart contracts. The intelligent scheduling unit of the resource scheduling module builds a multi-dimensional resource perception matrix to track the storage space fragmentation rate, network link congestion index, and task queue priority weight of edge nodes in real time. It generates a resource allocation plan through a genetic algorithm optimization engine. The resource allocation plan includes local storage tiering strategy and cloud elastic expansion instructions. The hot and cold separation unit deploys a three-level storage pool dynamic migration mechanism, automatically dividing the hot data storage area, warm data buffer area, and cold data archiving area according to the timeliness of data access. Each storage area maintains data consistency through asynchronous replication threads. The audit unit of the security control module builds a blockchain evidence chain. Its operation log is processed by hash summary to generate a unique digital fingerprint, and the PBFT consensus algorithm is used to achieve log consistency confirmation among multiple verification nodes. The implementation process of the above method is as follows; When a user initiates a data access request, the audit unit of the security control module starts the full-link recording process. First, the system captures key information of the operation behavior in real time, including metadata such as the access subject identity, operation timestamp, execution instruction type, and data flow trajectory, forming a structured log entry. To ensure the integrity of the original data, each log is attached with a dynamically generated unique digital fingerprint during the generation phase. This fingerprint is generated by performing an encryption operation on the log content. Any slight change in the content will cause the fingerprint value to change significantly. After completing fingerprint generation, the audit unit packages the log entry and the corresponding digital fingerprint into a data packet and transmits it to the log collection node in the blockchain network through an internal encrypted channel. This node, as the access layer of the blockchain network, is responsible for performing preliminary verification of the reported logs, including checking the integrity check code and timestamp validity of the data packet, and eliminating abnormal or duplicate log requests. Log data packets that pass the verification will be submitted to the consensus node pool in the blockchain network, triggering the multi-node collaborative verification process; During the consensus verification phase, multiple verification nodes within the network use a voting-based consensus mechanism to reach a consensus on the authenticity and order of the logs. Each verification node independently performs a match check between the log content and the fingerprint, while also comparing the log time series of adjacent nodes to see if there are any logical conflicts. When more than a set proportion of verification nodes confirm that the log content is legitimate, the system will generate block data containing the current log fingerprint and link it to the hash value chain of the previous block, forming an irreversible ledger record. The newly generated block is synchronized to all storage nodes via the distributed network to ensure that multiple copies of the log data are stored on physical media; To enhance the system's fault tolerance, the blockchain network employs a dynamic node management mechanism. Each verification node is equipped with a redundant backup module. When the primary node fails, the backup node automatically takes over the consensus verification task. Heartbeat detection and state synchronization mechanisms ensure the network's continued operation. For querying and auditing historical logs, the system provides a timeline-based chain search function. Authorized users can quickly locate the block location of the target log by entering search criteria. The system also verifies the continuity of the block hash value level by level to ensure that the log record has not been tampered with. To address high-frequency, concurrent log writing scenarios, the system introduces an asynchronous batch processing mechanism. The audit unit aggregates multiple logs generated within a short period of time, generating a combined log package containing batch identifiers and summary information. This batch submission reduces network transmission load. Before entering the blockchain, the combined log package undergoes a secondary aggregation check to ensure that the fingerprint values ​​of each log within the batch fully match the original record. This mechanism significantly improves system throughput while ensuring data integrity, meeting the audit requirements of large-scale node environments.

[0020] This solution uses the monitoring and collection module to continuously perceive key indicators such as edge node data traffic rate and storage space utilization. Combined with the deep learning algorithm and time series analysis based on the LSTM neural network in the predictive analysis module, it can accurately capture historical traffic patterns and predict future access needs. Compared with traditional static resource allocation methods, this mechanism can identify traffic peaks and troughs in advance. Through the intelligent scheduling algorithm and data hot and cold separation strategy of the resource scheduling module, it dynamically adjusts the storage capacity, caching strategy and shard distribution of edge nodes and the cloud, thereby improving deployment efficiency.

[0021] This solution utilizes a path optimization module that combines reinforcement learning with an improved Dijkstra algorithm. This module perceives node load, link latency, and network topology changes in real time, dynamically generates a topology-aware routing table, and selects the optimal data distribution path. In high-concurrency scenarios, this mechanism automatically avoids congested links and prioritizes data routing to low-load edge nodes, shortening data transmission distance and time. Furthermore, through real-time optimization and layout adjustments, it dynamically balances the load between nodes, avoiding response bottlenecks caused by overloading a single node, improving storage resource utilization, and reducing data latency.

[0022] This solution also provides end-to-end security protection across data collection, transmission, storage, and access through the security control module's end-to-end encryption technology, multi-level access control, and blockchain audit mechanisms. Unlike traditional single-point encryption and fixed permissions management, this solution enables real-time updates to encryption standards, dynamic adjustments to access permissions, and tamper-proof storage of operation logs. This allows for proactive identification of potential attacks and updated security policies, improving system stability and data security.

[0023] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0024] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A cloud-edge collaborative intelligent storage node dynamic deployment method, characterized in that: The following steps are involved: S1 real-time monitoring and data collection: First, continuously monitor the data traffic and storage resource status of edge nodes, collecting performance indicators including but not limited to data traffic rate, storage space utilization, and node processing capacity; S2 traffic pattern analysis and prediction: Using machine learning and a deep learning algorithm that combines time series analysis with an LSTM neural network model to analyze the data traffic information collected in step S1, obtain historical traffic patterns and predict future data access needs; S3 dynamic resource configuration: Based on the prediction results obtained in step S2, the storage resource configuration of the edge node and the cloud is automatically adjusted, including increasing and decreasing the storage capacity of the edge node, adjusting the data caching strategy, and optimizing the data transmission path. At the same time, intelligent scheduling algorithms, data hot and cold separation strategies, and adaptive sharding technologies are introduced to dynamically allocate computing resources and storage resources according to the real-time node load. S4 real-time optimization and layout adjustment: Optimize storage node layout in real time, select the optimal data distribution path using a reinforcement learning-based strategy, and dynamically adjust data distribution between edge nodes and the cloud through algorithms; S5 Data Security and Access Control: End-to-end encryption technology and multi-level access control mechanisms are used during the data transmission process, including data transmission encryption, dynamic management of access rights, and real-time updates of security policies. Security audits and vulnerability scans are conducted regularly, and encryption standards and access control rules are updated in a timely manner. Blockchain technology is also introduced to keep tamper-proof records of data access logs.

2. The cloud-edge collaborative intelligent storage node dynamic deployment method according to claim 1 is characterized by: The real-time monitoring and data collection in step S1 further includes deploying a lightweight monitoring agent on the edge node to achieve comprehensive perception of the edge node's operating status. The monitoring agent adopts a layered architecture design, including a data acquisition layer, a data processing layer, and a data transmission layer.

3. The cloud-edge collaborative intelligent storage node dynamic deployment method according to claim 1 is characterized by: In step S2, when constructing the hybrid prediction model, the long short-term memory network is preferably used as the basic architecture. The traffic data is decomposed into trend terms, seasonal terms and residual terms through time series decomposition technology. The training data set is dynamically updated in combination with the sliding window mechanism. The attention mechanism is introduced in the model training process to enhance the ability to capture sudden traffic characteristics, and rapid adaptation of cross-regional traffic patterns is achieved through transfer learning.

4. The method for dynamic deployment of intelligent storage nodes for cloud-edge collaboration according to claim 1 is characterized by: In step S3, the storage capacity adjustment strategy adopts a hierarchical expansion mechanism. When it is detected that the node storage space utilization exceeds the preset threshold, the virtualized hierarchical expansion of the local storage medium is triggered first. If it exceeds the upper limit of the system elastic expansion, a virtual storage unit application is automatically initiated to the cloud. At the same time, a heat attenuation factor is introduced into the data caching strategy to dynamically adjust the cache level distribution according to the data access frequency.

5. The cloud-edge collaborative intelligent storage node dynamic deployment method according to claim 1 is characterized by: In step S4, the reinforcement learning strategy constructs a state transition model based on the Q-learning algorithm, sets the node load rate, link delay and storage cost as state space parameters, balances the decision weights of exploration and utilization through the ε-greedy exploration strategy, and at the same time, combines the Dijkstra improved algorithm in the path selection process to calculate the optimal path of multi-hop transmission in real time, and dynamically generates a topology-aware routing table.

6. The method for dynamic deployment of intelligent storage nodes in cloud-edge collaboration according to claim 1, characterized in that: In step S5, the end-to-end encryption adopts a double-layer encryption mechanism combining the national secret SM4 algorithm and the Diffie-Hellman key exchange protocol, constructs an attribute-based access control model in access permission management, and dynamically generates access policies through role inheritance and permission minimization principles. The security audit module uses blockchain smart contract technology to realize on-chain storage of operation logs.

7. The method for dynamic deployment of intelligent storage nodes for cloud-edge collaboration according to claim 1, characterized in that: The intelligent scheduling algorithm in step S3 implements a multi-dimensional resource perception mechanism, that is, it comprehensively considers the remaining storage space of the node, the current network congestion index and the task priority weight, and obtains the optimal solution for resource allocation through genetic algorithm optimization. The fitness function design in the genetic algorithm includes three dimensions: storage utilization, transmission energy consumption and service response time, and sets a dynamic penalty coefficient to constrain the generation of invalid solutions. At the same time, the hot and cold separation strategy adopts a dynamic hierarchical storage architecture, establishes a three-level storage pool according to the local characteristics of data access time, retains the hot data accessed in the last 24 hours in the high-performance SSD storage layer, migrates the data with a historical decrease in access frequency to the distributed mechanical hard disk storage layer, and transfers the data that exceeds the set cold storage period to the object storage service after compression. The data flow between the storage layers is realized through asynchronous migration threads.

8. A cloud-edge collaborative intelligent storage node dynamic deployment system, adopting the cloud-edge collaborative intelligent storage node dynamic deployment method according to any one of claims 1 to 7, characterized in that: include: Monitoring and collection module, prediction and analysis module, resource scheduling module, path optimization module and security control module. The monitoring and collection module is deployed at the edge node and is used to collect data flow rate, storage space utilization rate and node processing capacity indicators in real time; A prediction and analysis module, used to analyze historical traffic patterns and predict future demand. The prediction and analysis module is connected to the monitoring and acquisition module. The prediction and analysis module has a built-in hybrid prediction model and an architecture based on an LSTM neural network. The resource scheduling module includes an intelligent scheduling unit and a hot and cold separation unit; A resource scheduling module is used to dynamically adjust edge nodes and cloud storage resources based on prediction results. The path optimization module is connected to the resource scheduling module. The path optimization module builds a state transition model based on the Q-learning algorithm, combines the improved Dijkstra algorithm to calculate the optimal multi-hop transmission path in real time, and dynamically generates a topology-aware routing table; The security control module is used to implement end-to-end encryption and multi-level access control. The security control module integrates an encryption unit, an access control unit and an audit unit. The encryption unit adopts a double-layer encryption mechanism of the national secret SM4 and Diffie-Hellman protocol. The access control unit dynamically generates strategies based on attributes. The audit unit realizes on-chain storage of operation logs through blockchain smart contracts.

9. The cloud-edge collaborative intelligent storage node dynamic deployment system according to claim 8, characterized in that: The intelligent scheduling unit of the resource scheduling module constructs a multi-dimensional resource perception matrix to track the storage space fragmentation rate, network link congestion index and task queue priority weight of the edge node in real time, and generates a resource allocation plan through a genetic algorithm optimization engine. The resource allocation plan includes a local storage tiering strategy and cloud elastic expansion instructions. The hot and cold separation unit deploys a three-level storage pool dynamic migration mechanism, which automatically divides the hot data storage area, warm data buffer area and cold data archiving area according to the timeliness characteristics of data access, and each storage interval maintains data consistency through an asynchronous replication thread.

10. The cloud-edge collaborative intelligent storage node dynamic deployment system according to claim 8, characterized in that: The audit unit of the security control module builds a blockchain evidence chain, and its operation log is processed by hash summary to generate a unique digital fingerprint, and log consistency confirmation is achieved among multiple verification nodes through the PBFT consensus algorithm.

Images