A strategy selection method for node backup in multi-domain networks

Abstract

The present invention relates to the field of data processing technology. The present invention provides a strategy selection method for node backup in a multi-domain network, which solves the problem that the existing backup management method cannot effectively cope with the complex resource distribution in the multi-domain network. By acquiring network system operation data and backup task requirements, a weighted undirected graph is used to represent the network structure, and a multi-domain network model is trained. A random forest algorithm is used to evaluate storage network element node attributes, and node priority ranking is determined to achieve reasonable allocation of backup tasks. During the task allocation process, node resource usage is monitored in real time, and backup strategies are dynamically adjusted to achieve maximum resource utilization and load balancing. The present invention improves backup efficiency and data reliability, reduces storage computing costs, and is suitable for scenarios such as large enterprise cloud environments and cross-domain data backup of financial institutions.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a strategy selection method for node backup in a multi-domain network. Background Art

[0002] Nodes in a multi-domain network are the basic elements that constitute the network. Nodes in a multi-domain network usually represent a network device or network service. Network devices include computers, servers, routers, etc. Network devices are connected to other devices through network interfaces and work together to complete data transmission and sharing. Nodes have multiple functions such as data exchange and routing, data processing and forwarding, network management and monitoring, so that data can reach the destination safely and efficiently along the correct path. According to different functions, locations and roles, nodes can be divided into terminal nodes, intermediate nodes, server nodes and gateway nodes.

[0003] In the existing single-domain network environment, node backup tasks are usually controlled and scheduled by a centralized backup management system. However, in a multi-domain network environment, there are many challenges. First, traditional backup strategies usually only consider the resource status and topology structure within a single domain, ignoring the global characteristics and complexity of multi-domain networks. Resource distribution and availability vary greatly between subdomains due to management policies, protocol differences, bandwidth constraints, etc. Second, different network element nodes (such as routers, switches, servers, etc.) have significant differences in computing power, storage capacity, network bandwidth, and topological location.

[0004] Existing backup management systems and methods are difficult to meet complex and changing needs in a multi-domain network environment. Existing backup solutions are limited to single-domain networks and usually rely on static configurations and fixed strategies. They lack flexibility and the ability to respond quickly to dynamic network states. Especially with the continuous expansion of network scale and the rapid growth in number, existing backup management methods cannot effectively cope with the complex resource distribution in multi-domain networks. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a strategy selection method for node backup in a multi-domain network, which solves the problem that the existing backup management method cannot effectively cope with the complex resource distribution in the multi-domain network.

[0006] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows:

[0007] The present invention provides a node backup strategy selection method in a multi-domain network, comprising:

[0008] Step S101, obtaining network system operation data, which includes node attribute data, network topology data, and node performance history data, and receiving backup task requirements, which include backup data size, backup window time, and data recovery speed;

[0009] Step S102, according to the network system operation data, determine the network between the storage network element nodes in the cloud environment, and represent the network between the storage network element nodes in the cloud environment in the form of a weighted undirected graph, the weighted undirected graph is used to show the connection relationship between the nodes, and the weight of the connection between the nodes, define node data and link attributes for each node, divide the multi-domain network to obtain autonomous domains, receive management authority and resource allocation mode, configure the management authority and resource allocation mode in the multi-domain network, train the random forest model for the nodes and links in the autonomous domain, and obtain a multi-domain network model;

[0010] Step S103, collecting attribute data of the storage network element node, the attribute data of the network element node includes CPU utilization, memory utilization, data disk utilization, network receiving bandwidth, average load within 1 minute and network domain. After collecting the attribute data of the network element node, preprocessing the attribute data of the network element node is performed to obtain the attribute data of the preprocessed network element node, and the random forest algorithm is used to perform model training. After the model training is completed, weighted processing is performed on each attribute of the network element node to obtain the attribute weight of the network element node, and the storage network element nodes are sorted from high to low according to the attribute weight of the network element node to obtain the evaluation result of the attribute of the network element node;

[0011] Step S104, determining the priority ranking of each network element node based on the evaluation result of the attributes of the network element node, receiving the allocation of backup tasks, and allocating the backup tasks in descending order according to the result of the priority ranking of the network element nodes;

[0012] Step S105, collect the data fed back by each network element node during the allocation process, substitute the data fed back by each network element node during the allocation process into the preset node resource usage detection model, and obtain the resource usage detection results of each node. The resource usage detection results of each node include normal node resource usage, tight node resource usage and node resource failure. The node with tight resource usage and the node with resource failure are removed from the allocation backup task, and the allocation backup task is reallocated to other available nodes with higher priority.

[0013] Furthermore, the node backup strategy selection method in the multi-domain network of the present invention, step S101 includes:

[0014] Node attribute data includes CPU utilization, memory utilization, and data disk utilization. Node attribute data is used to reflect the current resource status of the storage network element node;

[0015] Network topology data is used to reflect the connection relationship between storage network element nodes. Network topology data includes link bandwidth and delay;

[0016] Node performance history data includes past performance data that can provide node operation data under different loads and conditions;

[0017] The backup task requirements include: the size of the backup data, the backup window time, and the data recovery speed.

[0018] Furthermore, the node backup strategy selection method in the multi-domain network of the present invention, step S102 includes:

[0019] Based on the acquired network system operation data, the connection relationship between the storage network element nodes in the cloud environment is determined, and the nodes and the connections between them are represented in the form of a weighted undirected graph. Each node represents a storage network element, and the lines between the nodes represent the communication links between the nodes. The weights of the lines are used to reflect the bandwidth and latency network properties.

[0020] Divide the multi-domain network into multiple autonomous domains, each of which has relatively independent management authority and resource allocation methods, receive and configure management authority and resource allocation methods;

[0021] After dividing the autonomous domains and configuring management permissions, the random forest model is trained using the node and link data within the autonomous domain. The multi-domain network model will be used to evaluate the performance of different nodes and generate node priority rankings. By training the multi-domain network model, the multi-domain network model can learn the relationship between node attributes and network performance, which is used to evaluate nodes.

[0022] Furthermore, the node backup strategy selection method in the multi-domain network of the present invention, step S103 includes:

[0023] Collect and preprocess the attribute data of each storage network element node in the cloud environment, and use the preprocessed network element node attribute data to train the random forest algorithm model. During the training process, the model will learn the relationship between each attribute and node performance. After the model training is completed, the feature_importances_ attribute of the model is used to calculate the weight of each network element node attribute. Through weighted processing, the relative importance ranking of each network element node attribute is obtained;

[0024] The storage network element nodes are sorted from high to low according to the calculated network element node attribute weights. The sorting result reflects the priority of each node in the backup task. The nodes with high priority will be assigned backup tasks first.

[0025] Furthermore, the node backup strategy selection method in the multi-domain network of the present invention, step S104 includes:

[0026] Acquire the attribute evaluation result of the storage network element node, where the attribute evaluation result of the storage network element node includes the ranking of each attribute weight by the network element node;

[0027] Based on the attribute evaluation results, the priority order of each network element node is obtained, and the priority order of each network element node will be used as the basis for allocating backup tasks;

[0028] After determining the priority ranking of the network element nodes, receiving and allocating backup tasks, the allocated backup tasks include data backup requirements of various data volumes and importance levels;

[0029] Backup tasks are assigned in descending order according to the priority sorting results of the network element nodes. The node with the highest priority will be assigned tasks. If the node with the highest priority cannot accommodate the backup data, the remaining backup data will be assigned to the next node with the highest priority, and so on, until all backup data is assigned.

[0030] Furthermore, the node backup strategy selection method in the multi-domain network of the present invention, step S105 includes:

[0031] During the backup task allocation process, data feedback from each network element node is collected in real time. The data feedback from each network element node includes the node's CPU utilization, memory utilization, data disk utilization, network receiving bandwidth, average load, error information, and warning information;

[0032] Substituting the collected feedback data of each network element node into a preset resource usage detection model, obtaining the resource usage detection result of each node through the operation of the resource usage detection model, and removing the nodes with tight resource usage and faulty nodes from the current backup task allocation according to the detection result;

[0033] After removing nodes with tight resources or failures, redistribute backup tasks based on the priority of the remaining nodes, giving priority to nodes with higher priorities and good resource conditions for task allocation.

[0034] Furthermore, the node backup strategy selection method in the multi-domain network of the present invention, step S105 includes:

[0035] According to the resource usage detection result of each node obtained in step S105, it is determined that the node is in a state of tight resource usage or failure, and the node in the state of tight resource usage or failure is removed from the current backup task allocation list;

[0036] After removing nodes with insufficient resources or failures, re-evaluate the priority of the remaining nodes, obtain the node resource status and network topology information, and re-order the remaining nodes according to the latest node resource status and network topology information;

[0037] Determine if backup tasks need to be reallocated, select nodes with higher priorities and better resource conditions from the updated node priority ranking for task allocation, allocate the backup tasks that need to be reallocated to the selected nodes in turn, and continuously monitor the resource usage of the nodes and the progress of task execution.

[0038] Beneficial effects of the present invention:

[0039] The present invention represents a multi-domain network through a weighted undirected graph and trains a multi-domain network model. The present invention realizes the optimal allocation of global resources, comprehensively considers various factors such as the computing power, storage capacity, network bandwidth, and topological position of each node, avoids local optimal solutions, and improves the overall resource utilization efficiency. In the backup task allocation process, the present invention can collect data fed back by each node in real time, and dynamically adjust the backup strategy according to the resource usage detection results. The dynamic adaptation mechanism improves the backup task when the network state changes. It can still be completed efficiently and reliably, and improves the flexibility and response speed of the system.

[0040] The random forest algorithm is used to comprehensively evaluate node attributes. The present invention can accurately measure the importance of each attribute in node performance and status evaluation, improve the accuracy of evaluation results, provide a scientific basis for node sorting and backup task allocation, and achieve rational resource allocation.

[0041] The present invention avoids the situation where a single node takes on too many backup tasks by preferentially allocating backup tasks to nodes with superior performance and sufficient resources, thereby achieving load balancing, reducing the risk of single point failures, and improving system stability.

[0042] By comprehensively considering node performance, network bandwidth and topological location factors, the present invention can generate an efficient node backup strategy, significantly improve the backup speed, shorten the backup window period, and provide enterprises with a more convenient and efficient data backup solution.

[0043] In summary, the node backup strategy selection method in a multi-domain network proposed in the present invention has significant beneficial effects, can provide enterprises with a more efficient, reliable and economical data backup solution, and improve the overall performance and resource utilization of the multi-domain network. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on the drawings without paying any creative labor.

[0045] Figure 1 A schematic flow chart of a node backup strategy selection method in a multi-domain network provided by an embodiment of the present invention.

[0046] Figure 2 A multi-domain network modeling diagram provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described in conjunction with the specific embodiments of the present invention and the corresponding 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 creative work are within the scope of protection of the present invention. The technical solutions provided by the embodiments of the present invention are described in detail below in conjunction with the drawings.

[0048] In order to better understand the purpose of the present invention, the present invention is described in further detail below.

[0049] The present invention provides a node backup strategy selection method in a multi-domain network, comprising:

[0050] Step S101, obtaining network system operation data, which includes node attribute data, network topology data, and node performance history data, and receiving backup task requirements, which include backup data size, backup window time, and data recovery speed;

[0051] Obtain network system operation data. Network system operation data is the key basis for evaluating and optimizing multi-domain network node backup strategies, mainly including the following aspects:

[0052] Node attribute data directly reflects the current resource status of the storage network element node. Node attribute data is an important reference for evaluating node performance and selecting backup targets. Specifically, it includes:

[0053] CPU utilization: measures the CPU workload within a specific period of time and helps determine whether the allocation of computing resources is reasonable.

[0054] Memory usage: reflects the ratio of currently used memory to total memory, which helps to assess the memory pressure of the node.

[0055] Data disk utilization: indicates the ratio of used space to total space on a storage device, and is used to prevent overuse of storage resources.

[0056] Network topology data is used to describe the connection relationship between storage network element nodes. Network topology data is an important basis for optimizing data backup paths and load balancing. Specifically, it includes:

[0057] Link bandwidth: determines the speed and capacity of data transmission between nodes.

[0058] Link delay: affects the real-time and reliability of data transmission.

[0059] Node performance history data: Providing node operation data under different loads and conditions helps to have a more comprehensive understanding of node performance and stability, and provides a reference for future backup task allocation.

[0060] Receive backup task requirements. Backup task requirements are the direct driving force for developing node backup strategies, including:

[0061] The scale of backup data refers to the amount of data that needs to be backed up, which directly affects the complexity of the backup task and the resources required.

[0062] Backup window time: refers to the time period allowed for backup operations. The backup task must be completed within this time window to avoid affecting normal business.

[0063] Data recovery speed: When data is lost or damaged, the speed of data recovery is required, which has an important impact on the selection of backup strategies and node priority sorting.

[0064] By acquiring comprehensive network system operation data and detailed backup task requirements, the selection of backup strategies can be made more accurate and effective. Real-time acquisition of this data helps to dynamically adjust the backup strategy to adapt to changes in network status and real-time requirements of backup tasks. It provides strong data support for subsequent node evaluation, priority sorting, and backup task allocation, and is the basis for achieving efficient backup of multi-domain networks.

[0065] Step S102, according to the network system operation data, determine the network between the storage network element nodes in the cloud environment, and represent the network between the storage network element nodes in the cloud environment in the form of a weighted undirected graph, the weighted undirected graph is used to show the connection relationship between the nodes, and the weight of the connection between the nodes, define node data and link attributes for each node, divide the multi-domain network to obtain autonomous domains, receive management authority and resource allocation mode, configure the management authority and resource allocation mode in the multi-domain network, train the random forest model for the nodes and links in the autonomous domain, and obtain a multi-domain network model;

[0066] Step S102 involves network modeling, autonomous domain division, receiving and configuring management authority and resource allocation methods, and training a random forest model based on these data. The following is a detailed analysis of this step:

[0067] The weighted undirected graph representation determines the connection relationship between the storage network element nodes in the cloud environment based on the acquired network system operation data. s =(N s ,L s ) to formalize this relationship, where N s represents a set of nodes, L s Represents a collection of links. Each node represents a storage network element, the lines between nodes represent the communication links between them, and the weight of the lines reflects network properties such as bandwidth and latency.

[0068] Define node data and link attributes;

[0069] Node data, for graph G s For each node in the cluster, define its data attributes, including but not limited to CPU utilization, memory utilization, data disk utilization, etc. These attributes reflect the resource status of the node.

[0070] Link Properties,Meanwhile, defining link properties, we mainly focus on the bandwidth and,delay of the link, which are crucial in evaluating the efficiency and,reliability of data transmission.

[0071] Multi-domain network division and autonomous domain management;

[0072] Divide autonomous domains and divide the multi-domain network into multiple autonomous domains. Each autonomous domain has relatively independent management authority and resource allocation methods. The division helps to simplify network management while retaining a certain degree of flexibility and scalability.

[0073] Receive and configure management permissions: Receive management permission information for each autonomous domain and configure it in the multi-domain network to ensure that each autonomous domain can operate independently according to its management policy.

[0074] Resource allocation mode configuration, similarly, receives and configures the resource allocation mode of each autonomous domain to ensure that resources can be effectively used within the domain.

[0075] Random forest model training,model training data preparation, using the node and link data within the autonomous domain as training data, including the attribute information of the nodes and the connection relationship between the nodes.

[0076] Random forest model training: Use training data to train the random forest model. Random forest is an integrated learning algorithm that improves the accuracy and stability of prediction by building multiple decision trees. In this step, the random forest model will learn the relationship between node attributes and network performance, providing a basis for subsequent node evaluation.

[0077] Multi-domain network model generation: Through training, a multi-domain network model is generated, which can evaluate the performance of different nodes and generate node priority ranking based on the evaluation results. This provides an important basis for subsequent backup task allocation.

[0078] Accuracy of network modeling: The use of weighted undirected graphs can accurately represent the connection relationships and weights between storage network element nodes, providing a solid foundation for subsequent node evaluation and backup task allocation.

[0079] Reasonable division of autonomous domains: Reasonable division of autonomous domains helps simplify network management while retaining sufficient flexibility and scalability.

[0080] Through the training of the random forest model, the relationship between node attributes and network performance can be learned, providing a scientific basis for node evaluation and priority sorting. This helps to improve the accuracy and efficiency of backup strategy selection.

[0081] In summary, step S102 provides a basis and support for the subsequent selection of node backup strategies in multi-domain networks through network modeling, autonomous domain division, reception and configuration of management authority and resource allocation methods, and training of the random forest model.

[0082] Step S103, collecting attribute data of the storage network element node, the attribute data of the network element node includes CPU utilization, memory utilization, data disk utilization, network receiving bandwidth, average load within 1 minute and network domain. After collecting the attribute data of the network element node, preprocessing the attribute data of the network element node is performed to obtain the attribute data of the preprocessed network element node, and the random forest algorithm is used to perform model training. After the model training is completed, weighted processing is performed on each attribute of the network element node to obtain the attribute weight of the network element node, and the storage network element nodes are sorted from high to low according to the attribute weight of the network element node to obtain the evaluation result of the attribute of the network element node;

[0083] Step S103 involves the collection, preprocessing, model training and node sorting of storage network element node attribute data in the technical solution of the present invention. The following is a detailed analysis of this step:

[0084] Attribute range: Collects various attribute data of storage network element nodes, including CPU utilization, memory utilization, data disk utilization, network receiving bandwidth, average load within 1 minute, and network domain. The attributes fully reflect the resource utilization and network location information of the node.

[0085] Attribute data is usually collected in real time through monitoring systems deployed in the network to ensure the accuracy and timeliness of the data.

[0086] Attribute data preprocessing: clean the collected raw data, remove outliers and missing values, etc., to ensure data quality. Convert some attribute data, such as converting network domain attributes to binary form (1 for the same domain and 0 for different domains) to facilitate subsequent processing. Normalize the attribute data so that they have the same dimension for easy comparison and calculation.

[0087] Random forest algorithm model training;

[0088] The random forest algorithm is used for model training. Random forest is an ensemble learning algorithm that improves the accuracy and stability of predictions by building multiple decision trees. The random forest model is trained using the preprocessed network element node attribute data as training data. The model will learn the relationship between each attribute and node performance. After training, the feature_importances__ attribute of the model is used to calculate the weight of each attribute, which reflects the importance of each attribute in node performance evaluation.

[0089] Storage network element node sorting, based on the calculated attribute weights, weighted processing of each storage network element node attribute to obtain a comprehensive score for each node. Storage network element nodes are sorted from high to low based on the comprehensive score. The sorting result reflects the priority of each node in the backup task, and nodes with high priority will be assigned backup tasks first.

[0090] Collecting multiple attribute data can fully reflect the resource utilization and network location information of the node, providing a solid foundation for subsequent node evaluation and sorting. Data preprocessing can ensure data quality and improve the accuracy and stability of model training. The random forest algorithm can automatically learn the relationship between attributes and node performance, and calculate the weight of each attribute, providing a scientific basis for node sorting. Node sorting based on attribute weights can ensure the realization of goals such as maximizing resource utilization and load balancing, and improve the efficiency and reliability of backup tasks.

[0091] In summary, step S103 provides a basis for subsequent backup task allocation by collecting, preprocessing and storing attribute data of network element nodes and using the random forest algorithm for model training and node sorting.

[0092] Step S104, determining the priority ranking of each network element node based on the evaluation result of the attributes of the network element node, receiving the allocation of backup tasks, and allocating the backup tasks in descending order according to the result of the priority ranking of the network element nodes;

[0093] Step S104 is to implement node backup task allocation in the technical solution of the present invention. Step S104 determines the priority ranking of each node based on the evaluation result of the network element node attribute, and allocates the backup task accordingly. The following is a detailed analysis of this step:

[0094] According to the network element node attribute evaluation result obtained in step S103, the priority ranking of each node is determined. This ranking is based on the comprehensive performance evaluation of the node in terms of computing power, storage capacity, network bandwidth, and topological location. The ranking result reflects the priority of each node in the backup task, and the node with a high priority will have a higher priority when allocating backup tasks.

[0095] Backup tasks may be assigned manually by network administrators or scheduled by automated backup systems. Backup tasks include requirements such as the size of backup data, backup window time, and data recovery speed.

[0096] Backup tasks are assigned in descending order based on the priority of the network element nodes. This means that the node with the highest priority will be assigned the backup task first, followed by the node with the second highest priority, and so on.

[0097] If a single node can accommodate all the backup data, the data is allocated entirely to that node.

[0098] If a single node cannot accommodate all the backup data, its remainder is distributed to the next highest priority node until all the backup data has been distributed.

[0099] During the backup task allocation process, it is necessary to continuously monitor the resource usage of each node to ensure that the backup task does not cause node resource overload or performance bottleneck.

[0100] If it is found that the resource usage of a node is tight or a failure occurs, it is necessary to adjust the backup strategy in time and reallocate the task to other available and higher-priority nodes to ensure the smooth completion of the backup task.

[0101] Prioritization based on the evaluation results of network element node attributes can ensure that backup tasks are assigned to nodes with better performance, thereby improving backup efficiency and reliability.

[0102] Allocating backup tasks from high to low priority can fully utilize network resources and avoid resource waste and performance bottlenecks.

[0103] Continuous monitoring and dynamic adjustment mechanisms ensure that backup tasks can be completed efficiently and reliably when network status changes.

[0104] In summary, step S104 achieves efficient and reasonable allocation of backup tasks in a multi-domain network by determining the priority ranking of network element nodes, receiving backup tasks and allocating tasks according to the priority.

[0105] Step S105, collect the data fed back by each network element node during the allocation process, substitute the data fed back by each network element node during the allocation process into the preset node resource usage detection model, and obtain the resource usage detection results of each node. The resource usage detection results of each node include normal node resource usage, tight node resource usage and node resource failure. The node with tight resource usage and the node with resource failure are removed from the allocation backup task, and the allocation backup task is reallocated to other available nodes with higher priority.

[0106] Step S105 is the technical solution of the present invention, which involves real-time monitoring and dynamic adjustment of the usage of network element node resources during the backup task allocation process. The following is a detailed analysis of this step:

[0107] During the backup task allocation process, data feedback from each network element node is collected in real time, including but not limited to the node's CPU utilization, memory utilization, data disk utilization, network receiving bandwidth, average load, error information, and warning information.

[0108] The collection frequency is set according to actual needs to ensure that changes in node resource usage can be discovered in a timely manner. Before step S105, a node resource usage detection model needs to be preset, and the resource usage detection model can determine the node resource usage based on the input node resource usage data.

[0109] The resource usage detection model processes and analyzes the node resource usage data based on threshold judgment and trend prediction, thereby obtaining the resource usage detection results.

[0110] The resource usage detection results mainly include three situations: normal node resource usage, tight node resource usage, and node resource failure.

[0111] Normal node resource usage indicates that the current node resources are sufficient and can execute the assigned backup tasks normally.

[0112] The resource usage of the node is tight, which means that the current node resource usage is high, but has not reached the failure threshold, which may affect the execution efficiency or security of the backup task.

[0113] The node's resources fail, indicating that the current node has experienced resource overload, hardware failure, and other problems, and cannot continue to perform the backup task.

[0114] Adjust backup task allocation;

[0115] According to the resource usage detection results, nodes with tight resource usage or failures are processed. The specific operation is: remove these nodes from the current backup task allocation, and reallocate the backup tasks originally assigned to them to other available nodes with higher priority. The adjustment process is a dynamic process and requires multiple iterations based on actual conditions. After each adjustment, the resource usage of the remaining nodes needs to be re-evaluated and adjusted again as needed.

[0116] During the execution of the backup task, the resource usage of each node is continuously monitored to ensure the smooth progress of the backup task. Problems found during the monitoring process, adjustment measures taken, and final results are recorded in the log for subsequent analysis and tracing. By real-time monitoring of the resource usage of network element nodes, potential problems can be discovered and handled in a timely manner to avoid the interruption of backup tasks due to insufficient node resources or failures. Dynamically adjusting the allocation of backup tasks according to changes in node resource usage can ensure efficient and reliable execution of backup tasks and improve the resource utilization and backup efficiency of the overall network. Recording key information during the monitoring and adjustment process will help with subsequent analysis and optimization work and improve the maintainability and scalability of the system.

[0117] In summary, step S105 monitors the resource usage of network element nodes in real time and dynamically adjusts the backup task allocation according to the detection result, so that the backup task is executed efficiently and reliably in the multi-domain network.

[0118] The technical solution of the present invention solves the problem that the existing backup management method cannot effectively cope with the complex resource distribution in the multi-domain network through the following steps:

[0119] Obtain network system operation data, including node attribute data (such as CPU utilization, memory utilization, data disk utilization, etc.), network topology data (such as link bandwidth and latency), and node performance history data. At the same time, receive backup task requirements, including the size of backup data, backup window time, and data recovery speed. Based on the acquired network system operation data, use a weighted undirected graph to represent the network between storage network element nodes in the cloud environment, and clarify the connection relationship and weight between nodes. Divide the multi-domain network into multiple autonomous domains, each of which has relatively independent management authority and resource allocation methods, and use the node and link data in the autonomous domain to train the random forest model to obtain a multi-domain network model.

[0120] Collect the attribute data of the storage network element nodes, perform preprocessing, and use the random forest algorithm to train the model. Calculate the weight of each node attribute through the model's feature_importances__ attribute, and perform a comprehensive evaluation and sorting of the storage network element nodes according to the weight to obtain the priority of each node. Receive and assign backup tasks based on the evaluation results and priority sorting of node attributes. Start assigning backup tasks from the node with the highest priority. If a single node cannot accommodate all backup data, it will be assigned to the next highest priority node in turn until all backup data is assigned.

[0121] During the backup task allocation process, the data fed back by each node is collected in real time and substituted into the preset resource usage detection model to obtain the resource usage detection results of each node. For nodes with tight resource usage or failure, they are removed from the current backup task allocation, and the backup tasks are reallocated according to the priority of the remaining nodes. The resource usage and task execution progress of each node are continuously monitored to ensure the stability of the backup process.

[0122] Through the above steps, the technical solution of the present invention can effectively deal with the complex resource distribution problem in the multi-domain network. Specifically, by representing the multi-domain network with a weighted undirected graph and training the multi-domain network model, the optimal allocation of global resources is achieved. In the process of allocating backup tasks, dynamic adjustments are made based on the node feedback data collected in real time to ensure the flexibility and response speed of resource allocation. The random forest algorithm is used to comprehensively evaluate the node attributes, accurately measure the importance of each attribute in the node performance and status evaluation, and improve the accuracy and reliability of the evaluation results. It avoids the situation where a single node takes on too many backup tasks, achieves load balancing, and reduces the risk of single point failure. Through intelligent optimization algorithms and dynamic adjustment mechanisms, existing resources are rationally utilized, and unnecessary storage expansion and hardware investment costs are reduced.

[0123] Specifically, the node backup strategy selection method in the multi-domain network of the present invention, step S101 includes:

[0124] Node attribute data includes CPU utilization, memory utilization, and data disk utilization. Node attribute data is used to reflect the current resource status of the storage network element node;

[0125] Network topology data is used to reflect the connection relationship between storage network element nodes. Network topology data includes link bandwidth and delay;

[0126] Node performance history data includes past performance data that can provide node operation data under different loads and conditions;

[0127] The backup task requirements include: the size of the backup data, the backup window time, and the data recovery speed.

[0128] Node attribute data;

[0129] CPU utilization: indicates the proportion of CPU usage within a specific period of time, reflecting the computing resource usage of the node.

[0130] Memory usage: refers to the ratio of currently used memory to total memory, which measures the memory resource pressure of the node.

[0131] Data disk utilization: refers to the ratio of the used space to the total space on the storage device. It evaluates the storage resource usage of the node. The node attribute data is used to reflect the current resource status of the storage network element node and is an important basis for evaluating the node load capacity and the feasibility of the backup task.

[0132] Network topology data;

[0133] Link bandwidth: indicates the maximum data transmission rate of the communication link between nodes, which affects the data transmission efficiency of the backup task.

[0134] Latency: refers to the time required for a data packet to be transmitted in the network, which affects the real-time performance and response speed of the backup task.

[0135] Network topology data is used to reveal the connection relationship between storage network element nodes and their network properties, which helps to consider data transmission path optimization and load balancing in the backup strategy.

[0136] Node performance historical data;

[0137] Past performance data includes the operation data of nodes under different loads and conditions. The operation data of nodes under different loads and conditions includes historical CPU utilization, memory usage, etc. The node performance history data provides the performance of nodes in different scenarios, which helps to predict the behavior of nodes in future backup tasks and improve the accuracy and reliability of policy selection.

[0138] Receive backup task requirements;

[0139] The size of the backup data refers to the amount of data that needs to be backed up, which affects the storage requirements and transmission time of the backup task.

[0140] Backup window time: refers to the time period during which backup operations are allowed, which limits the execution time of backup tasks.

[0141] Data recovery speed: refers to the speed requirement for restoring data from backup when needed, which affects the reliability and practicality of the backup strategy.

[0142] Receiving the backup task requirement is step S101, which provides a clear goal and direction for the subsequent backup strategy selection.

[0143] The data collected in step S101 covers the resource status of the node, network topology, historical performance, and specific requirements of the backup task, providing comprehensive information support for policy selection. By acquiring node attributes and network topology data in real time, and referring to historical performance data, the accuracy of policy selection can be ensured, and the risks brought by blind decision-making can be avoided. Receiving backup task requirements enables policy selection to be adjusted according to different scenarios and requirements, improving the flexibility and adaptability of the method.

[0144] In summary, step S101 lays a solid foundation for the subsequent selection of node backup strategies in the multi-domain network by comprehensively and accurately collecting information such as node attributes, network topology, historical performance, and backup task requirements.

[0145] Specifically, the node backup strategy selection method in the multi-domain network of the present invention, step S102 includes:

[0146] Based on the acquired network system operation data, the connection relationship between the storage network element nodes in the cloud environment is determined, and the nodes and the connections between them are represented in the form of a weighted undirected graph. Each node represents a storage network element, and the lines between the nodes represent the communication links between the nodes. The weights of the lines are used to reflect the bandwidth and latency network properties.

[0147] Divide the multi-domain network into multiple autonomous domains, each of which has relatively independent management authority and resource allocation methods, receive and configure management authority and resource allocation methods;

[0148] After dividing the autonomous domains and configuring management permissions, the random forest model is trained using the node and link data within the autonomous domain. The multi-domain network model will be used to evaluate the performance of different nodes and generate node priority rankings. By training the multi-domain network model, the multi-domain network model can learn the relationship between node attributes and network performance, which is used to evaluate nodes.

[0149] Specifically, step S102 of the node backup strategy selection method in the multi-domain network of the present invention involves modeling of network structure, division of autonomous domains and configuration of management authority, and training of a random forest model based on such information. The following is a detailed analysis of step S102:

[0150] Determine the node connection relationship: Based on the obtained network system operation data, first determine the connection relationship between the storage network element nodes in the cloud environment, including which nodes can communicate directly and the communication link properties between them. Use a weighted undirected graph to represent these nodes and the connections between them. In the graph, each node represents a storage network element, and the lines between the nodes represent the communication links between the nodes. The weight of the line is used to reflect network properties such as bandwidth and latency, which have an important impact on the execution efficiency of the backup task and the quality of data transmission.

[0151] Divide the multi-domain network into multiple autonomous domains. Each autonomous domain has relatively independent management authority and resource allocation methods, which reflects the complexity of the structure and management of the multi-domain network. Management authority and resource allocation method configuration: Receive and configure the management authority and resource allocation method of each autonomous domain to ensure the basis for the subsequent backup strategy to be effectively implemented in each autonomous domain.

[0152] After dividing the autonomous domain and configuring management permissions, the training set is prepared using the node and link data in the autonomous domain. The data should include various attributes of the nodes (such as CPU utilization, memory usage, etc.) and the connection relationship and network properties between them. Use these training data to train the random forest model. Random forest is an algorithm based on ensemble learning. It improves the stability and accuracy of the model by building multiple decision trees and combining their prediction results.

[0153] After training is completed, the resulting multi-domain network model will be able to evaluate the performance of different nodes and generate node priority rankings. This is achieved by learning the relationship between node attributes and network performance through the model, providing a scientific basis for the allocation of subsequent backup tasks.

[0154] The use of weighted undirected graphs can intuitively represent the connection relationship and network properties between nodes, providing a clear network topology view for subsequent strategy selection. Autonomous domain division and management authority configuration are key steps in dealing with the complexity of multi-domain networks, ensuring that backup strategies can be independently and effectively executed in each autonomous domain.

[0155] Random forest model training: By training the random forest model, the complex relationship between node attributes and network performance can be learned, providing powerful intelligent support for subsequent node priority sorting and backup task allocation.

[0156] In summary, step S102 lays a solid foundation for the subsequent selection of node backup strategies in multi-domain networks through network structure modeling, autonomous domain division and management authority configuration, and random forest model training.

[0157] Specifically, the node backup strategy selection method in the multi-domain network of the present invention, step S103 includes:

[0158] Collect and preprocess the attribute data of each storage network element node in the cloud environment, and use the preprocessed network element node attribute data to train the random forest algorithm model. During the training process, the model will learn the relationship between each attribute and node performance. After the model training is completed, the feature_importances_ attribute of the model is used to calculate the weight of each network element node attribute. Through weighted processing, the relative importance ranking of each network element node attribute is obtained;

[0159] The storage network element nodes are sorted from high to low according to the calculated network element node attribute weights. The sorting result reflects the priority of each node in the backup task. The nodes with high priority will be assigned backup tasks first.

[0160] Specifically, step S103 in the strategy selection method for node backup in a multi-domain network of the present invention involves multiple key steps such as collection, preprocessing, model training, and node sorting of storage network element node attribute data. The following is a detailed analysis of step S103:

[0161] Node attribute data collection;

[0162] Attribute range: Collect various attribute data of each storage network element node in the cloud environment, including but not limited to CPU utilization, memory utilization, data disk utilization, network receiving bandwidth, average load within 1 minute, and network domain. The attribute data fully reflects the current resource status and network location of the node.

[0163] Data preprocessing: clean the collected attribute data, remove outliers, missing values, etc., to ensure the quality and integrity of the data. Transform the data as needed, such as normalizing some attribute data to the same dimension for subsequent processing. Divide the preprocessed data set into training sets and test sets, usually in a certain ratio (such as 70% training set, 30% test set), for subsequent model training and verification.

[0164] Random forest model training, select appropriate parameters to initialize the random forest model. Random forest is an algorithm based on ensemble learning, which improves the stability and accuracy of the model by building multiple decision trees and combining their prediction results. Use the training set data to train the random forest model. During the training process, the model will learn the relationship between various attributes and node performance, which is crucial for subsequent node sorting and backup task allocation. Use the test set data to verify the trained model and evaluate the model's prediction performance and generalization ability.

[0165] Attribute weight calculation and sorting;

[0166] The attribute weight calculation uses the feature_importances__ attribute of the trained random forest model to calculate the weight of each network element node attribute. The weight reflects the importance of each attribute in the node performance evaluation. According to the calculated attribute weight, the storage network element nodes are weighted and comprehensively evaluated, and sorted from high to low according to the evaluation results. The sorting result directly reflects the priority of each node in the backup task. The nodes with high priority will be assigned backup tasks first.

[0167] By calculating the attribute weights and sorting the nodes, the priority of each node in the backup task can be objectively reflected, providing guidance for the reasonable allocation of subsequent backup tasks. Step S103 achieves scientific evaluation and reasonable sorting of storage network element nodes through key steps such as node attribute data collection, preprocessing, random forest model training, and node sorting, providing a basis for the allocation of subsequent backup tasks.

[0168] Specifically, the strategy selection method for node backup in a multi-domain network of the present invention, step S104 includes:

[0169] Acquire the attribute evaluation result of the storage network element node, where the attribute evaluation result of the storage network element node includes the ranking of each attribute weight by the network element node;

[0170] Based on the attribute evaluation results, the priority order of each network element node is obtained, and the priority order of each network element node will be used as the basis for allocating backup tasks;

[0171] After determining the priority ranking of the network element nodes, receiving and allocating backup tasks, the allocated backup tasks include data backup requirements of various data volumes and importance levels;

[0172] Backup tasks are assigned in descending order according to the priority sorting results of the network element nodes. The node with the highest priority will be assigned tasks. If the node with the highest priority cannot accommodate the backup data, the remaining backup data will be assigned to the next node with the highest priority, and so on, until all backup data is assigned.

[0173] Specifically, step S104 of the node backup strategy selection method in the multi-domain network of the present invention reasonably allocates the backup task to each node based on the attribute evaluation result and priority sorting of the storage network element node. The following is a detailed analysis of step S104:

[0174] The attribute evaluation result of the storage network element node is obtained from step S103. The attribute evaluation result includes the priority ranking of the network element nodes by each attribute weight. The weight and ranking reflect multiple factors such as the performance, resource status and topological position of the node.

[0175] Based on the attribute evaluation results, a priority order is determined for each storage network element node. The priority order is an important basis for the subsequent allocation of backup tasks, which ensures that nodes with good resource conditions and superior performance can get backup tasks first.

[0176] Receive backup task requirements, including key indicators such as the size of backup data, backup window time, and data recovery speed. Indicators are crucial for developing appropriate backup strategies.

[0177] Backup tasks are assigned in descending order according to the priority ranking results of the network element nodes. This means that the node with the highest priority will be assigned tasks first. If its capacity is insufficient, the remaining backup data will be assigned to the next highest priority node, and so on, until all backup data has been assigned.

[0178] During the allocation process, dynamic adjustments are made based on the real-time resource usage and performance of the nodes. For example, if a high-priority node is found to be close to full load or performance degradation, the allocation strategy is adjusted in a timely manner to transfer part or all of the backup tasks to other available nodes with higher priority.

[0179] Accuracy of attribute evaluation results: The accuracy of attribute evaluation results directly affects the rationality of node priority sorting, and further affects the efficiency and effect of backup task allocation. Therefore, in step S103, the integrity of attribute data collection, the accuracy of preprocessing, and the adequacy of model training must be ensured.

[0180] Clarifying the needs of backup tasks is a prerequisite for formulating a reasonable allocation strategy. Only by fully understanding the requirements such as the scale of backup data, backup window time and data recovery speed can the rationality and efficiency of backup task allocation be ensured.

[0181] Since the network status and node performance may change at any time, the dynamic adjustment mechanism in step S104 is particularly important. It ensures the flexibility and responsiveness of backup task allocation, and can adjust the allocation strategy in time when node resources are tight or performance is degraded, avoiding risks such as single point overload and data loss.

[0182] In summary, step S104 achieves reasonable allocation and efficient execution of backup tasks by obtaining attribute evaluation results, determining node priority order, receiving backup task requirements, and allocating backup tasks. At the same time, the flexibility of the allocation strategy is ensured through a dynamic adjustment mechanism.

[0183] Specifically, the node backup strategy selection method in the multi-domain network of the present invention, step S105 includes:

[0184] During the backup task allocation process, data feedback from each network element node is collected in real time. The data feedback from each network element node includes the node's CPU utilization, memory utilization, data disk utilization, network receiving bandwidth, average load, error information, and warning information;

[0185] Substituting the collected feedback data of each network element node into a preset resource usage detection model, obtaining the resource usage detection result of each node through the operation of the resource usage detection model, and removing the nodes with tight resource usage and faulty nodes from the current backup task allocation according to the detection result;

[0186] After removing nodes with tight resources or failures, redistribute backup tasks based on the priority of the remaining nodes, giving priority to nodes with higher priorities and good resource conditions for task allocation.

[0187] During the backup task allocation process, data feedback from each network element node is collected in real time. The data includes the node's CPU utilization, memory utilization, data disk utilization, network receiving bandwidth, average load, error information, and warning information, which fully reflects the node's current operating status and resource usage.

[0188] The collected feedback data from each network element node is substituted into the pre-set resource usage detection model. Through complex algorithms and logical judgments, the model can accurately evaluate the resource usage status of the node and determine whether the node is in a resource-constrained or faulty state.

[0189] According to the operation results of the resource usage detection model, the resource usage detection results of each node are obtained. The detection results are divided into three situations: normal, tight resource usage and node failure.

[0190] Nodes with tight resource usage and faulty nodes are promptly removed from the current backup task allocation. This step is to prevent the risk of backup task failure or data loss due to insufficient node resources or faults.

[0191] After removing nodes with insufficient resources or failures, redistribute backup tasks according to the priority of the remaining nodes. When redistributing, nodes with higher priority and good resource conditions are given priority for task allocation to ensure smooth and efficient execution of backup tasks.

[0192] After the backup task is reallocated, the resource usage and task execution progress of each node are continuously monitored. Through real-time monitoring, potential problems can be discovered and solved in a timely manner to ensure the continuity and stability of the backup task. The real-time data collection and monitoring mechanism in step S105 ensures that the system can respond to changes in node status in a timely manner, avoiding the expansion of problems caused by delayed processing.

[0193] The accuracy of the resource usage detection model is crucial for determining the node status and adjusting the backup task allocation. Only accurate detection results can guide the subsequent task allocation adjustment. The task allocation adjustment mechanism in step S105 ensures that the system can flexibly adjust the backup task allocation strategy when facing emergencies such as node resource shortage or failure, thereby ensuring the stability of the system and the security of the data.

[0194] In summary, step S105 achieves comprehensive monitoring and flexible adjustment of the backup task allocation process through key steps such as real-time data collection, resource usage detection, task allocation adjustment and continuous monitoring.

[0195] Specifically, the node backup strategy selection method in the multi-domain network of the present invention, step S105 includes:

[0196] According to the resource usage detection result of each node obtained in step S105, it is determined that the node is in a state of tight resource usage or failure, and the node in the state of tight resource usage or failure is removed from the current backup task allocation list;

[0197] After removing nodes with insufficient resources or failures, re-evaluate the priority of the remaining nodes, obtain the node resource status and network topology information, and re-order the remaining nodes according to the latest node resource status and network topology information;

[0198] Determine if backup tasks need to be reallocated, select nodes with higher priorities and better resource conditions from the updated node priority ranking for task allocation, allocate the backup tasks that need to be reallocated to the selected nodes in turn, and continuously monitor the resource usage of the nodes and the progress of task execution.

[0199] Resource usage detection result analysis: Based on the resource usage detection results of each node obtained in step S105, nodes with tight resource usage or failure are accurately identified. The identified resource-intensive or failure nodes are removed from the current backup task allocation list to prevent these nodes from becoming bottlenecks for backup tasks or causing task failures.

[0200] After removing the problematic node, re-collect the resource status (such as CPU utilization, memory utilization, data disk utilization, etc.) and network topology information of the remaining nodes.

[0201] Based on the latest node resource status and network topology information, the priority of the remaining nodes is re-evaluated. This step ensures the accuracy and timeliness of node sorting and reflects the current actual situation of the network.

[0202] According to the updated node priority ranking, nodes with higher priority and good resource status are selected to reallocate backup tasks.

[0203] The backup tasks that need to be reallocated are assigned to the selected nodes in turn. This step ensures that the backup tasks can be efficiently and evenly distributed in the network, avoiding single point overload and performance bottlenecks.

[0204] After the backup tasks are reallocated, the resource usage and task execution progress of each node are continuously monitored. This includes real-time monitoring of key indicators such as the node's CPU utilization, memory usage, and network bandwidth.

[0205] When new resource shortages or faulty nodes are discovered, the feedback mechanism is immediately triggered to dynamically adjust the backup task allocation strategy based on the latest monitoring data. Continuous monitoring and adjustment ensure that the system can remain efficient and stable in the face of dynamic changes.

[0206] Step S105 enables the system to quickly adapt to changes in network status through real-time monitoring and dynamic adjustment, so that the backup task can proceed smoothly. By re-evaluating and sorting the remaining nodes and reallocating the backup tasks, step S105 achieves optimal resource allocation, avoiding resource waste and performance bottlenecks. The continuous monitoring and feedback adjustment mechanism ensures that the system can respond quickly to emergencies and maintains the stability and reliability of the system.

[0207] In summary, step S105 improves the efficient execution of backup tasks in a multi-domain network and the stable operation of the system through a series of sophisticated operations and monitoring mechanisms. The implementation of this step not only improves the efficiency and reliability of backup tasks, but also provides strong support for performance optimization and resource management of the entire network.

[0208] The present invention proposes a node backup task strategy selection method suitable for a multi-domain network environment, which generates the optimal node backup strategy selection method by comprehensively analyzing the computing power, storage capacity, network bandwidth, and topological location of each storage network element node in the multi-domain network, and optimizing using an intelligent optimization algorithm. The method can adjust the allocation and scheduling of backup tasks according to changes in the network state, ensuring maximum resource utilization and load balancing during the backup process, thereby improving backup efficiency, reducing storage computing costs, and significantly improving the data reliability and disaster recovery capabilities of the multi-domain network.

[0209] The specific process of the method of the present invention is as follows Figure 1 As shown. First, the resource information of the acquired storage network element nodes is analyzed, and the characteristics of each storage network element node are deeply analyzed. Then, an evaluation is performed, which is mainly divided into the evaluation of storage network element node resources and the evaluation of network topology location. The evaluation of storage network element node resources mainly includes identifying and analyzing the resource usage of each storage network element node, such as CPU utilization, memory utilization, and data disk utilization, to ensure that network resources are reasonably allocated and efficiently utilized. By evaluating the topological location, the network layout and the location of the storage network element nodes are determined to ensure that the influence of the topological location is considered when backing up the storage network element nodes, thereby optimizing the overall architecture and performance of the network. Subsequently, the optimization algorithm is applied to adjust the network, including policy adjustment and load balancing, to further improve network performance. Finally, a method for selecting the best node backup task strategy is generated based on the above content.

[0210] The network between storage network element nodes in the cloud environment can be represented by a weighted undirected graph G s =(N s ,L s ), where N s is the set of nodes, L s is a collection of links; node attributes consider computing and storage capabilities, and link attributes consider bandwidth. s It consists of multiple autonomous domains, and each domain is connected by an inter-domain link. The inter-domain link is also called the backbone link, and its bandwidth resources are much larger than the intra-domain link. The network of the hth domain is represented by G s ,h=(N s ,h,L s ,h); where N s ,h is the node set of the hth domain, L s ,h is the link set of the hth domain. Figure 2 Shown is a modeling diagram of a multi-domain network.

[0211] First, we analyze the characteristics of storage network element nodes. The attributes of storage network element nodes mainly include: CPU utilization, memory utilization, data disk utilization, network receiving bandwidth, average load within 1 minute, and network domain. As shown in Table 1, Figure 2 Taking the storage network element node in as an example, the various properties of the storage network element node are displayed. Among them, the network domain is the domain where the storage network element node is located, which can reflect the topological location information of the storage network element node; the CPU utilization is the workload of the CPU in a specific time, which can ensure that computing resources are reasonably allocated; the memory usage is the ratio of the currently used memory to the total memory, which can evaluate the memory pressure of the storage network element node; the data disk utilization is the ratio of the used space on the storage device to the total space, to prevent excessive use of storage resources; the network receiving bandwidth is the rate at which the storage network element node receives data through the network within a certain period of time, to ensure the efficiency of data transmission; the average load within 1 minute is the number of processes that the system needs to process in the last 1 minute, to determine whether the system is facing an overload risk.

[0212]

[0213] Table 1 Storage network element node attributes

[0214] The present invention uses a weighted comprehensive evaluation method to comprehensively evaluate and rank the attributes of storage network element nodes, and combines the resource status and topological location information of the storage network element nodes to prioritize the storage network element nodes, thereby providing a reasonable selection strategy for node backup tasks. In order to achieve a scientific and objective evaluation of storage network element nodes, the method introduces a random forest algorithm to calculate the weights of the attributes of the storage network element nodes, so as to accurately measure the importance of each attribute in the node performance and status evaluation, and improve the accuracy and reliability of the evaluation results.

[0215] Random forest is an algorithm based on ensemble learning. Its core idea is to improve the overall prediction performance and model stability by building multiple decision tree models and integrating their results. The method of the present invention uses a random forest algorithm to evaluate the importance of each attribute of a storage network element node to the node performance, thereby obtaining the weight corresponding to each attribute. It can be divided into the following stages: data preparation, model training, and calculation of attribute weights.

[0216] When using the random forest algorithm for evaluation, the goal is to determine the performance score of the storage network element node. First, it is necessary to collect historical data related to the storage network element node, which includes various attributes of each node (as shown in Table 1, CPU utilization, memory usage, network receiving bandwidth, etc.) and the corresponding node performance. In order to improve the quality and applicability of the data, the collected data needs to be preprocessed. For example, for the attribute "network domain", when the storage network element node and its node backup task are in the same domain, the network domain attribute value is defined as 1; conversely, if the storage network element node and its backup task are in different domains, the attribute value is defined as 0. Next, the preprocessed data set is randomly divided into a training set and a test set according to a certain ratio. The training set is used for model training, while the test set is used for model verification to ensure that the model has good generalization ability and prediction accuracy.

[0217] Select appropriate parameters to initialize the random forest model, then use the training data set to train the random forest regressor, use the test set to make predictions and calculate the accuracy of the model. The random forest will automatically evaluate the importance of each attribute of the storage network element node to the node performance through the tree construction process.

[0218] The feature_importances__ attribute of the model can be used to obtain the importance score, that is, the weight, of each attribute of the storage network element node. The score can reflect the impact of each attribute on the result when the decision tree splits.

[0219] rf.feature_importances__ is an attribute of the RandomForestClassifier or RandomForestRegressor object in scikit-learn, which is used to indicate the importance of each feature attribute to the model prediction result. Among them, the feature_importances__ attribute is an array that indicates the relative importance (i.e., weight) of each input feature in the model. The importance score is calculated by calculating the sum of the impurity reduction (such as Gini impurity or entropy) brought by each feature when all decision tree nodes are split, and averaging all trees. The specific calculation steps are as follows:

[0220] Calculation of node split contribution: When a decision tree of the random forest selects a feature (such as CPU utilization) for node splitting, this split will lead to a reduction in the impurity of the model (such as a reduction in the Gini index or an increase in information gain). The amount of reduction in impurity is proportional to the importance of the feature in this tree.

[0221] Cumulative feature importance contribution: The algorithm calculates the sum of the impurity reductions brought by each feature (CPU utilization, memory utilization, etc.) across all trees. For example, if the CPU utilization feature is frequently used as the basis for node splits in many trees, and the impurity reduction brought by each split is large, then its importance score will be high.

[0222] Normalized feature importance: Calculate the proportion of each feature's importance score in all features (that is, the sum of all feature contributions is normalized to 1). feature_importances__ ultimately yields the relative importance of these features in the entire model. If a feature (such as CPU utilization) has a particularly high importance score, it means that the feature contributes greatly to the performance of the storage network element node; conversely, the lower the score, the less important it is.

[0223] The above random forest algorithm can effectively measure the influence of each attribute on the performance evaluation of storage network element nodes and assign reasonable weight values ​​to each attribute. This provides a scientific basis and data support for comprehensive evaluation and reasonable sorting of storage network element nodes, helps optimize the allocation and management of storage node backup resources, and improves the overall system operation efficiency and reliability.

[0224] The method of the present invention uses a weighted comprehensive evaluation method to evaluate and rank storage network element nodes. The specific steps are as follows:

[0225] Attribute weight allocation,According to the above random forest algorithm, the weight value corresponding to each attribute of the storage network element node can be obtained, where the sum of the weights is 1.

[0226] Normalize the attributes of the storage network element nodes (such as CPU utilization, memory usage, network receiving bandwidth, etc.) to ensure that they have the same dimension. This method uses minimum-maximum normalization to normalize the data to a range of 0 to 1. For indicators that need to be minimized (that is, the lower the value of the attribute, the better the performance of the corresponding storage network element node, such as CPU utilization, memory usage, average load within 1 minute, etc.), use the reverse normalization formula:

[0227]

[0228] For indicators that need to be maximized (i.e., the higher the attribute value, the better the performance of the corresponding storage network element node, such as network receiving bandwidth), use the standard normalization formula:

[0229]

[0230] Where X is the original attribute value, X min and X max are the minimum and maximum values ​​of the attribute, respectively.

[0231] To calculate the score, the normalized attribute values ​​of the storage network element nodes are multiplied by their corresponding weights, and then the sum is calculated to obtain the total score of each storage network element node.

[0232]

[0233] The storage network element nodes are sorted from high to low according to the scores, and finally a node backup task strategy selection method in a multi-domain network environment is obtained.

[0234] According to the priority ranking of storage network element nodes, node backup tasks are assigned to nodes with higher priorities. First, the backup tasks are assigned from the node with the highest ranking. If a single node cannot accommodate all the backup data, the remaining backup data is assigned to the next priority node in turn until the node data is fully backed up. During the backup task process, it is necessary to continuously monitor the resource usage of each node and dynamically adjust the backup strategy when resources are tight or a node fails. For example, according to the resource usage, the priority of the storage network element node is adjusted in time, and after obtaining the new ranking, the backup task is reallocated to other available nodes.

[0235] In summary, the present invention proposes a node backup strategy selection method in a multi-domain network. By comprehensively evaluating factors such as the computing power, storage capacity, network bandwidth, and topological location of storage network element nodes, combined with an intelligent optimization algorithm, an efficient node backup strategy is generated. The backup strategy is dynamically adjusted to adapt to changes in network status. This method not only optimizes the allocation and scheduling of node backup tasks, but also provides application value for improving the overall performance of the multi-domain network.

[0236] Large-scale node backup strategy selection in cloud storage platform,In a large enterprise cloud environment consisting of multiple autonomous domains, due to the rapid growth of business, the data center is facing the pressure of surging data volume, and the original data backup solution cannot meet the current needs.,Enterprises need to adopt a new node backup selection method to improve the efficiency and reliability of data backup.

[0237] Embodiment 1, the enterprise cloud environment includes multiple autonomous domains, each of which has several storage network element nodes, and the domains are connected by high-speed backbone links. The monitoring system deployed in each domain is used to collect information such as CPU utilization, memory utilization, data disk utilization, network receiving bandwidth, and average load within 1 minute of the storage network element node in real time. The collected data is processed, the weight of each attribute is calculated using the random forest algorithm, and the storage network element nodes are comprehensively evaluated and sorted according to the weight. Based on the evaluation results, the scheduler is responsible for assigning backup tasks to nodes with high priority and adjusting the backup strategy when necessary. Through the intelligent optimization algorithm, it is ensured that the resources of the storage network element nodes are reasonably allocated and the overall resource utilization is improved. It avoids the situation where a single node undertakes too many backup tasks, realizes load balancing, and reduces the risk of single point failure. The optimized backup strategy significantly improves the backup speed and shortens the backup window period. The existing resources are reasonably used, unnecessary storage expansion is reduced, and the hardware investment cost is saved. The dynamic adjustment mechanism enables the system to respond quickly in the face of emergencies, ensuring the integrity of the data and the stability of the system.

[0238] Example 2: A financial institution has multiple data centers covering multiple autonomous domains, and the number of storage network element nodes in each domain varies. Due to the sensitivity of financial transaction data, it is necessary to ensure that this data can be quickly restored under any circumstances. The traditional manual backup method can no longer meet current business needs, especially the network delay and bandwidth limitation problems faced during cross-regional backup. The backup task needs to have high reliability and fault tolerance.

[0239] Financial institutions have multiple data centers, each of which is divided into multiple logical domains, such as transaction processing domains and risk management domains. Specialized detection software is deployed in each data center to collect real-time information such as CPU utilization, memory utilization, and network receiving bandwidth of storage nodes. The random forest algorithm is used to analyze the resource status of the nodes, and a comprehensive evaluation is performed in combination with network topology information to generate a node priority list. According to the priority list, key transaction data is backed up to the node with the best performance and synchronized to other data centers through inter-domain high-speed links for redundancy.

Claims

1. A node backup strategy selection method in a multi-domain network, characterized in that: include: Step S101, obtaining network system operation data, which includes node attribute data, network topology data, and node performance history data, and receiving backup task requirements, which include backup data size, backup window time, and data recovery speed; Step S102, according to the network system operation data, determine the network between the storage network element nodes in the cloud environment, and represent the network between the storage network element nodes in the cloud environment in the form of a weighted undirected graph, the weighted undirected graph is used to show the connection relationship between the nodes, and the weight of the connection between the nodes, define node data and link attributes for each node, divide the multi-domain network to obtain autonomous domains, receive management authority and resource allocation mode, configure the management authority and resource allocation mode in the multi-domain network, train the random forest model for the nodes and links in the autonomous domain, and obtain a multi-domain network model; Step S103, collecting attribute data of the storage network element node, the attribute data of the network element node includes CPU utilization, memory utilization, data disk utilization, network receiving bandwidth, average load within 1 minute and network domain. After collecting the attribute data of the network element node, preprocessing the attribute data of the network element node is performed to obtain the attribute data of the preprocessed network element node, and the random forest algorithm is used to perform model training. After the model training is completed, weighted processing is performed on each attribute of the network element node to obtain the attribute weight of the network element node, and the storage network element nodes are sorted from high to low according to the attribute weight of the network element node to obtain the evaluation result of the attribute of the network element node; Step S104, determining the priority ranking of each network element node based on the evaluation result of the attributes of the network element node, receiving the allocation of backup tasks, and allocating the backup tasks in descending order according to the result of the priority ranking of the network element nodes; Step S105, collect the data fed back by each network element node during the allocation process, substitute the data fed back by each network element node during the allocation process into the preset node resource usage detection model, and obtain the resource usage detection results of each node. The resource usage detection results of each node include normal node resource usage, tight node resource usage and node resource failure. The node with tight resource usage and the node with resource failure are removed from the allocation backup task, and the allocation backup task is reallocated to other available nodes with higher priority.

2. The node backup strategy selection method in a multi-domain network according to claim 1, characterized in that: The step S101 includes: Node attribute data includes CPU utilization, memory utilization, and data disk utilization. Node attribute data is used to reflect the current resource status of the storage network element node; Network topology data is used to reflect the connection relationship between storage network element nodes. Network topology data includes link bandwidth and delay; Node performance history data includes past performance data that can provide node operation data under different loads and conditions; The backup task requirements include: the size of the backup data, the backup window time, and the data recovery speed.

3. The node backup strategy selection method in a multi-domain network according to claim 1, characterized in that: The step S102 includes: Based on the acquired network system operation data, the connection relationship between the storage network element nodes in the cloud environment is determined, and the nodes and the connections between them are represented in the form of a weighted undirected graph. Each node represents a storage network element, and the lines between the nodes represent the communication links between the nodes. The weights of the lines are used to reflect the bandwidth and latency network properties. Divide the multi-domain network into multiple autonomous domains, each of which has relatively independent management authority and resource allocation methods, receive and configure management authority and resource allocation methods; After dividing the autonomous domains and configuring management permissions, the random forest model is trained using the node and link data within the autonomous domain. The multi-domain network model will be used to evaluate the performance of different nodes and generate node priority rankings. By training the multi-domain network model, the multi-domain network model can learn the relationship between node attributes and network performance, which is used to evaluate nodes.

4. The node backup strategy selection method in a multi-domain network according to claim 1, characterized in that: The step S103 includes: Collect and preprocess the attribute data of each storage network element node in the cloud environment, and use the preprocessed network element node attribute data to train the random forest algorithm model. During the training process, the model will learn the relationship between each attribute and node performance. After the model training is completed, the feature_importances_ attribute of the model is used to calculate the weight of each network element node attribute. Through weighted processing, the relative importance ranking of each network element node attribute is obtained; The storage network element nodes are sorted from high to low according to the calculated network element node attribute weights. The sorting result reflects the priority of each node in the backup task. The nodes with high priority will be assigned backup tasks first.

5. The node backup strategy selection method in a multi-domain network according to claim 1, characterized in that: The step S104 includes: Acquire the attribute evaluation result of the storage network element node, where the attribute evaluation result of the storage network element node includes the ranking of each attribute weight by the network element node; Based on the attribute evaluation results, the priority order of each network element node is obtained, and the priority order of each network element node will be used as the basis for allocating backup tasks; After determining the priority ranking of the network element nodes, receiving and allocating backup tasks, the allocated backup tasks include data backup requirements of various data volumes and importance levels; Backup tasks are assigned in descending order according to the priority sorting results of the network element nodes. The node with the highest priority will be assigned tasks. If the node with the highest priority cannot accommodate the backup data, the remaining backup data will be assigned to the next node with the highest priority, and so on, until all backup data is assigned.

6. The node backup strategy selection method in a multi-domain network according to claim 1, characterized in that: The step S105 includes: During the backup task allocation process, data feedback from each network element node is collected in real time. The data feedback from each network element node includes the node's CPU utilization, memory utilization, data disk utilization, network receiving bandwidth, average load, error information, and warning information; Substituting the collected feedback data of each network element node into a preset resource usage detection model, obtaining the resource usage detection result of each node through the operation of the resource usage detection model, and removing the nodes with tight resource usage and faulty nodes from the current backup task allocation according to the detection result; After removing nodes with tight resources or failures, redistribute backup tasks based on the priority of the remaining nodes, giving priority to nodes with higher priorities and good resource conditions for task allocation.

7. The node backup strategy selection method in a multi-domain network according to claim 6, characterized in that: The step S105 includes: According to the resource usage detection result of each node obtained in step S105, it is determined that the node is in a state of tight resource usage or failure, and the node in the state of tight resource usage or failure is removed from the current backup task allocation list; After removing nodes with insufficient resources or failures, re-evaluate the priority of the remaining nodes, obtain the node resource status and network topology information, and re-order the remaining nodes according to the latest node resource status and network topology information; Determine if backup tasks need to be reallocated, select nodes with higher priorities and better resource conditions from the updated node priority ranking for task allocation, allocate the backup tasks that need to be reallocated to the selected nodes in turn, and continuously monitor the resource usage of the nodes and the progress of task execution.

Images