Method and device for allocating physical network resources

Abstract

The embodiment of the invention provides a method and a device for allocating physical network resources. The method comprises the following steps: acquiring a virtual undirected weighted graph of a virtual network; acquiring a physical undirected weighted graph of the physical network; and inputting the virtual undirected weighted graph and the physical undirected weighted graph into a target neural network model, and deploying virtual nodes and virtual links of a virtual network to a physical network through the target neural network model. Through the method and the device, the problem of low physical network resource allocation efficiency in related technologies is solved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of communications, and more specifically, to a method and apparatus for allocating physical network resources. Background Technology

[0002] With the rapid development of satellite communication technology, satellite networks have demonstrated unique advantages in achieving global coverage and high-availability communication.

[0003] The "cloud-edge-device collaboration" mechanism combines cloud computing, edge computing, and terminal computing into a single physical network resource. It allows for collaborative cooperation among physical resources at different levels to improve the overall system performance and efficiency. However, in a cloud-edge-device collaborative satellite network environment, the rational allocation of physical network resources still faces a series of challenges. Traditional methods of physical network resource allocation have certain limitations. For example, in traditional terrestrial base station-based networks, due to geographical constraints and uneven spectrum distribution, flexible resource configuration and optimization are difficult to achieve, resulting in the physical network's performance not being fully realized.

[0004] Existing technologies primarily employ linear programming to address the problem of uneven allocation of physical network resources. Linear programming uses appropriate parameters to summarize the problem into a mathematical programming model, aiming to achieve optimal allocation of physical network resources. However, this method may suffer from high computational complexity and low efficiency in complex satellite network environments.

[0005] There is currently no effective solution to the above problems. Summary of the Invention

[0006] This invention provides a method and apparatus for allocating physical network resources, thereby at least addressing the problem of low efficiency in physical network resource allocation in related technologies.

[0007] According to an embodiment of the present invention, a method for allocating physical network resources is provided, comprising: obtaining a virtual undirected weighted graph of a virtual network, wherein the virtual undirected weighted graph is a set representing virtual nodes and virtual links contained in the virtual network, and resources required for deploying the virtual network; obtaining a physical undirected weighted graph of a physical network, wherein the physical undirected weighted graph is a set representing physical nodes and physical links contained in the physical network, and resources possessed by the physical network; inputting the virtual undirected weighted graph and the physical undirected weighted graph into a target neural network model, and using the target neural network model to allocate the virtual nodes and virtual links of the virtual network. The nodes and the virtual links are deployed to the physical network; wherein, the target neural network model is trained using a sample virtual undirected weighted graph and a sample physical undirected weighted graph, the sample virtual undirected weighted graph representing the set of physical nodes and physical links contained in the physical network, and the set of resources possessed by the physical network, and the sample physical undirected weighted graph representing the set of sample physical nodes and sample physical links contained in the sample physical network, and the set of resources possessed by the sample physical network; the target neural network model satisfies a preset convergence condition between the total time and the benefit-cost ratio of deploying the sample virtual network to the sample physical network.

[0008] In an exemplary embodiment, deploying the virtual nodes and virtual links of the virtual network to the physical network using the target neural network model includes: inputting the virtual undirected weighted graph and the physical undirected weighted graph into a first neural network model; determining physical nodes matching each virtual node in the virtual node set using the first neural network model to obtain a first physical node subset, wherein the target neural network model includes the first neural network model, the virtual undirected weighted graph includes the virtual node set in the virtual network, and the physical undirected weighted graph includes the physical node set in the physical network; determining a physical weighted subgraph in the physical undirected weighted graph, wherein the physical weighted subgraph includes the first physical node subset and a first physical link subset, the first physical link subset including links between each physical node in the first physical node subset; and deploying the virtual nodes and virtual links of the virtual network to the physical network using the physical weighted subgraph and the virtual undirected weighted graph.

[0009] In an exemplary embodiment, deploying the virtual nodes and virtual links of the virtual network to the physical network using the physical weighted subgraph and the virtual undirected weighted graph includes: inputting the virtual undirected weighted graph and the physical weighted subgraph into a second neural network model; determining a target link in the first physical link subset using the second neural network model to obtain a target physical link subset, wherein the target neural network model includes the second neural network model; determining the network consisting of the first physical node subset and the target physical link subset in the physical network as a target physical subnetwork; and deploying the virtual nodes and virtual links of the virtual network to the target physical subnetwork.

[0010] In an exemplary embodiment, the virtual undirected weighted graph further includes: resources required to deploy the virtual nodes; the physical undirected weighted graph further includes: available resources of the physical nodes; and a first physical node subset is obtained by determining, through the first neural network model, physical nodes matching each virtual node in the set of physical nodes, to obtain a first physical node subset, comprising: the first neural network model performing the following operations on each virtual node in the set of virtual nodes, wherein the virtual node performing the following operations is taken as the current virtual node; the convolutional layer of the first neural network model extracts features from the current virtual node, its neighboring nodes, the virtual links between the current virtual node and its neighboring nodes, and the resources required to deploy the current virtual node to obtain a first feature matrix; the convolutional layer of the first neural network model extracts features from each physical node in the set of physical nodes, the physical links between the physical nodes, and the available resources of each physical node to obtain a second feature matrix; and the first and second feature matrices are used to determine the current physical node matching the current virtual node in the set of physical nodes, wherein the first physical node subset includes the current physical node.

[0011] In an exemplary embodiment, determining a current physical node matching the current virtual node in the physical node set using the first feature matrix and the second feature matrix includes: inputting the first feature matrix and the second feature matrix into a probability layer of the first neural network model, obtaining the mapping probability of each physical node in the physical node set through the probability layer; inputting the first feature matrix and the second feature matrix into a filtering layer of the first neural network model, determining multiple physical nodes satisfying preset constraints in the physical node set through the filtering layer; and determining the node with the highest mapping probability among the multiple physical nodes satisfying the preset constraints as the current physical node matching the current virtual node.

[0012] In an exemplary embodiment, determining multiple physical nodes that satisfy preset constraints in the set of physical nodes through the filtering layer includes: determining physical nodes in the set of physical nodes whose available resources are greater than or equal to the resources required to deploy the current virtual node through the filtering layer.

[0013] In an exemplary embodiment, the virtual undirected weighted graph further includes: resources required to deploy the virtual link; the physical undirected weighted graph further includes: available resources of the physical link; inputting the virtual undirected weighted graph and the physical weighted subgraph into a second neural network model; determining a target link in the first physical link subset through the second neural network model to obtain a target physical link subset includes: the second neural network model performing the following operations on each virtual link in the virtual link set, wherein the virtual link performing the following operations is designated as the current virtual link; the convolutional layer of the second neural network model extracts features from the current virtual link, the virtual nodes connected to the current virtual link, and the resources required to deploy the current virtual link to obtain a third feature matrix; the convolutional layer of the second neural network model extracts features from each physical link in the first physical link subset, the physical nodes connected to the physical link, and the available resources of the physical link to obtain a fourth feature matrix; determining a current physical link matching the current virtual link in the first physical link subset through the third feature matrix and the fourth feature matrix, wherein the target physical link subset includes the current physical link.

[0014] In an exemplary embodiment, determining the current physical link matching the current virtual link in the first physical link subset using the third feature matrix and the fourth feature matrix includes: inputting the third feature matrix and the fourth feature matrix into the probability layer of the second neural network model, obtaining the mapping probability of each physical link in the first physical link subset through the probability layer; inputting the third feature matrix and the fourth feature matrix into the filtering layer of the second neural network model, determining multiple physical links satisfying the second constraint condition in the first physical link subset through the filtering layer; and determining the physical link with the highest mapping probability among the multiple physical links satisfying the second constraint condition as the current physical link matching the current virtual link.

[0015] In an exemplary embodiment, determining a plurality of physical links satisfying a second constraint condition in the first physical link subset through the filtering layer includes: determining physical links in the first physical link subset whose available resources are greater than or equal to the resources required to deploy the current virtual link through the filtering layer of the second neural network model.

[0016] In an exemplary embodiment, before inputting the virtual undirected weighted graph and the physical undirected weighted graph into the target neural network model, the method further includes: obtaining the sample virtual undirected weighted graph of the sample virtual network, wherein the sample virtual undirected weighted graph further includes: resources required to deploy the sample virtual nodes and resources required to deploy the sample virtual links; obtaining the sample physical undirected weighted graph of the sample physical links, wherein the sample physical undirected weighted graph further includes: available resources of the sample physical nodes and available resources of the sample physical links; training an initial neural network model using the sample virtual undirected weighted graph and the sample physical undirected weighted graph, and ending the training to obtain the target neural network model when the convergence condition is met between the total time and the benefit-cost ratio of deploying the sample virtual network to the sample physical network.

[0017] In an exemplary embodiment, the training to obtain the target neural network model ends when the convergence condition between the total time of deploying the sample virtual network to the sample physical network and the benefit-cost ratio is met includes: determining the total time of deploying the sample virtual nodes to the sample physical nodes and the sample virtual links to the sample physical links; determining the benefit-cost ratio between deploying the sample virtual nodes to the sample physical nodes and the sample virtual links to the sample physical links; determining a target optimization value based on the total time and the benefit-cost ratio; and determining that the convergence condition is met when the target optimization value reaches a preset value, thus ending the training to obtain the target neural network model.

[0018] In an exemplary embodiment, determining the total duration for deploying the sample virtual node to the sample physical node and the sample virtual link to the sample physical link includes: determining the duration for deploying the sample virtual node to the sample physical node as a first duration; determining the duration for deploying the sample virtual link to the sample physical link as a second duration; and determining the sum of the first duration and the second duration as the total duration.

[0019] In an exemplary embodiment, determining the cost-benefit ratio of deploying the sample virtual node to the sample physical node and deploying the sample virtual link to the sample physical link includes: determining the amount of central processing unit resources occupied by deploying the sample virtual node to the sample physical node as a first resource amount; determining the amount of bandwidth resources occupied by deploying the sample virtual link to the sample physical link as a second resource amount; and obtaining the cost-benefit ratio using the first resource amount and the second resource amount.

[0020] According to another embodiment of the present invention, an apparatus for allocating physical network resources is provided, comprising: a first acquisition module, configured to acquire a virtual undirected weighted graph of a virtual network, wherein the virtual undirected weighted graph is a set representing virtual nodes and virtual links contained in the virtual network, and resources required for deploying the virtual network; a second acquisition module, configured to acquire a physical undirected weighted graph of a physical network, wherein the physical undirected weighted graph is a set representing physical nodes and physical links contained in the physical network, and resources possessed by the physical network; and a deployment module, configured to input the virtual undirected weighted graph and the physical undirected weighted graph into a target neural network model, and through the target neural network model deploy the virtual nodes and virtual links of the virtual network to the physical network;

[0021] The target neural network model is trained using a sample virtual undirected weighted graph and a sample physical undirected weighted graph. The sample virtual undirected weighted graph represents the sample virtual nodes and sample virtual links contained in the sample virtual network, as well as the set of resources required to deploy the sample virtual network. The sample physical undirected weighted graph represents the sample physical nodes and sample physical links contained in the sample physical network, as well as the set of resources possessed by the sample physical network.

[0022] The target neural network model achieves a convergence condition between the total time and the cost-benefit ratio of deploying the sample virtual network to the sample physical network, as preset.

[0023] According to yet another embodiment of the present invention, a computer-readable storage medium is also provided, wherein a computer program is stored therein, wherein the computer program, when executed by a processor, implements the steps of the method described in any of the preceding claims.

[0024] According to yet another embodiment of the present invention, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0025] According to yet another embodiment of the present invention, a computer program product is also provided, comprising a computer program that, when executed by a processor, implements the steps of the method described in any of the preceding claims.

[0026] This invention constructs a virtual undirected weighted graph for a virtual network and a physical undirected weighted graph for a physical network. A target neural network model is then used to deploy virtual nodes and virtual links from the virtual network to the physical network. Since the target neural network model satisfies preset conditions including the total time and cost-benefit ratio for deploying sample virtual nodes and links, it achieves the technical effect of reducing complexity and improving efficiency. Therefore, it can solve the problem of low efficiency in physical network resource allocation. Attached Figure Description

[0027] Figure 1 This is a hardware structure block diagram of a mobile terminal according to an embodiment of the present invention, which describes a method for allocating physical network resources.

[0028] Figure 2 This is a flowchart of a method for allocating physical network resources according to an embodiment of the present invention;

[0029] Figure 3 This is a schematic diagram illustrating the deployment of a virtual network according to an embodiment of the present invention;

[0030] Figure 4 This is a structural block diagram of an apparatus for allocating physical network resources according to an embodiment of the present invention. Detailed Implementation

[0031] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings and examples.

[0032] It should be noted that the terms "first," "second," etc., in the specification, claims, and drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0033] The methods and embodiments provided in this application can be executed on a mobile terminal, computer terminal, or similar computing device. Taking running on a mobile terminal as an example, Figure 1 This is a hardware structure block diagram of a mobile terminal according to an embodiment of the present invention, which describes a method for allocating physical network resources. Figure 1 As shown, a mobile terminal may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 (which may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and a memory 104 for storing data are also shown. The mobile terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the mobile terminal described above. For example, the mobile terminal may also include components that are more... Figure 1 The more or fewer components shown, or having the same Figure 1The different configurations shown.

[0034] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the method for allocating physical network resources in this embodiment of the invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, thereby implementing the above-described method. The memory 104 may include high-speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0035] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the mobile terminal's communication provider. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module used for wireless communication with the Internet.

[0036] This embodiment provides a method for allocating physical network resources on the aforementioned mobile terminal. Figure 2 This is a flowchart of a method for allocating physical network resources according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps:

[0037] Step S202: Obtain a virtual undirected weighted graph, wherein the virtual undirected weighted graph is used to represent the set of virtual nodes and virtual links contained in the virtual network, as well as the resources required to deploy the virtual network;

[0038] Specifically, the virtual undirected weighted graph includes: a set of virtual nodes in the virtual network and a set of virtual links in the virtual network. The set of virtual nodes includes multiple virtual nodes, and the set of virtual links includes multiple virtual links. The virtual links are links between the virtual nodes.

[0039] Specifically, graph wheels can be used to construct an undirected weighted graph of a virtual network:

[0040] G VR ={N VR ,L VR,RC N ,RB L}

[0041] Where N VR L represents the set of virtual nodes in a virtual network. The set of virtual nodes includes multiple virtual nodes in the virtual network. VR This represents the set of virtual links in a virtual network request. The set of virtual links includes multiple virtual links within the virtual network; a virtual link is a link between virtual nodes. It can be represented by l. vr (i,j) represents a virtual node To virtual node Virtual links between them. RC N RB represents the resources required to deploy virtual nodes. L This indicates the resources required to deploy a virtual link.

[0042] Step S204: Obtain a physical undirected weighted graph, wherein the physical undirected weighted graph is used to represent the set of physical nodes and physical links contained in the physical network, as well as the set of resources possessed by the physical network.

[0043] Specifically, the physical undirected weighted graph includes: a set of physical nodes in the physical network and a set of physical links in the physical network. The set of physical nodes includes multiple physical nodes, and the set of physical links includes multiple physical links. The physical links are links between the physical nodes.

[0044] Constructing an undirected weighted graph of the physical network:

[0045] G S ={N SC N SE N ST ,L S C N B L}

[0046] Where, N SC N SE N ST These are physical nodes in a physical network, which is a cloud-edge-device collaborative physical network. N SC It is a cloud server node, N SE It is an edge server node, L S It is the set of physical links in a physical network. The set of physical links includes physical links in the physical network, and physical links are links between physical nodes. C N B is the available resource of a physical node. L It refers to the available resources of the physical link.

[0047] Step S206: Input the virtual undirected weighted graph and the physical undirected weighted graph into the target neural network model, and deploy the virtual nodes and virtual links of the virtual network to the physical network through the target neural network model;

[0048] The target neural network model is trained using a sample virtual undirected weighted graph and a sample physical undirected weighted graph. The sample virtual undirected weighted graph represents the set of sample virtual nodes and sample virtual links contained in the sample virtual network, as well as the set of resources required to deploy the sample virtual network. Specifically, the sample virtual undirected weighted graph includes a set of sample virtual nodes and a set of sample virtual links in the sample virtual network. The set of sample virtual nodes includes multiple sample virtual nodes, and the set of sample virtual links includes multiple sample virtual links. The sample virtual links are links between the sample virtual nodes. The sample physical undirected weighted graph represents the set of sample virtual nodes and sample virtual links contained in the sample virtual network, as well as the set of resources required to deploy the sample virtual network. Specifically, the sample physical undirected weighted graph includes a set of sample physical nodes and a set of sample physical links in the sample physical network. The set of sample physical nodes includes multiple sample physical nodes, and the set of sample physical links includes multiple sample physical links. The sample physical links are links between the sample physical nodes.

[0049] The target neural network model achieves a convergence condition between the total time and the cost-benefit ratio of deploying the sample virtual network to the sample physical network, as preset.

[0050] Optionally, the entity performing the above steps may be a background processor or other devices with similar processing capabilities, or a machine that integrates at least an image acquisition device and a data processing device. The image acquisition device may include an image acquisition module such as a camera, and the data processing device may include a terminal such as a computer or a mobile phone, but is not limited thereto.

[0051] The aforementioned target neural network model includes a first neural network model (also known as a virtual node mapping model) and a second neural network model (also known as a link mapping model).

[0052] The aforementioned first neural network model is used to map virtual nodes in a virtual network to physical nodes in a physical network. Specifically, the virtual undirected weighted graph and the physical undirected weighted graph are input into the first neural network model, and the first neural network model determines the physical nodes that match each virtual node in the set of physical nodes, thus obtaining the first subset of physical nodes.

[0053] For example, the first neural network model performs the following operations on each virtual node in the set of virtual nodes, wherein the virtual node performing the following operations is designated as the current virtual node:

[0054] The convolutional layer of the first neural network model extracts features from the current virtual node, the neighboring nodes of the current virtual node, the virtual links between the current virtual node and the neighboring nodes, and the resources required to deploy the current virtual node to obtain the first feature matrix.

[0055] The convolutional layer of the first neural network model extracts features from each physical node in the set of physical nodes, the physical links between the physical nodes, and the available resources of each physical node to obtain a second feature matrix.

[0056] The current physical node matching the current virtual node is determined from the set of physical nodes using a first feature matrix and a second feature matrix, wherein the first subset of physical nodes includes the current physical node, including:

[0057] The first feature matrix and the second feature matrix are input into the probability layer of the first neural network model, and the mapping probability of each physical node in the physical node set is obtained through the probability layer.

[0058] The first feature matrix and the second feature matrix are input into the filtering layer of the first neural network model, and the filtering layer determines multiple physical nodes that meet the preset constraints from the set of physical nodes.

[0059] The node with the highest mapping probability among multiple physical nodes that meet the preset constraints is determined as the current physical node that matches the current virtual node.

[0060] The process of determining multiple physical nodes that meet preset constraints in the set of physical nodes through a filtering layer includes: determining physical nodes in the set of physical nodes whose available resources are greater than or equal to the resources required to deploy the current virtual node through the filtering layer.

[0061] In the undirected weighted graph, a physical weighted subgraph is determined. The physical weighted subgraph includes a first physical node subset and a first physical link subset. The first physical link subset includes the links between the physical nodes in the first physical node subset.

[0062] The aforementioned second neural network model is used to determine the target physical link subset in the first physical link subset of the physical undirected weighted subgraph, determine the network composed of the first physical node subset and the target physical link subset as the target physical subnetwork, and deploy the virtual nodes and virtual links of the virtual network to the target physical subnetwork.

[0063] Specifically, the virtual undirected weighted graph and the physical weighted subgraph are input into the second neural network model. The target link is determined through the second neural network model in the first physical link subset, and the target physical link subset is obtained. The target neural network model includes the second neural network model. The network composed of the first physical node subset and the target physical link subset in the physical network is determined as the target physical subnetwork, and the virtual nodes and virtual links of the virtual network are deployed to the target physical subnetwork.

[0064] For example, the second neural network model performs the following operations on each virtual link in the set of virtual links, where the virtual link performing the following operations is designated as the current virtual link:

[0065] The convolutional layer of the second neural network model extracts features from the current virtual link, the virtual nodes connected to the current virtual link, and the resources required to deploy the current virtual link, thus obtaining the third feature matrix;

[0066] The convolutional layer of the second neural network model extracts features from each physical link in the first physical link subset, the physical nodes connected by the physical links, and the available resources of the physical links to obtain the fourth feature matrix.

[0067] The target physical link subset, which matches the current virtual link, is determined using the third and fourth feature matrices. This subset includes the current physical link and comprises:

[0068] The third and fourth feature matrices are input into the probability layer of the second neural network model, and the mapping probability of each physical link in the first physical link subset is obtained through the probability layer.

[0069] The third and fourth feature matrices are input into the filtering layer of the second neural network model, and the filtering layer determines multiple physical links that satisfy the second constraint condition in the first physical link subset.

[0070] The physical link with the highest mapping probability among the plurality of physical links that satisfy the second constraint is determined as the current physical link that matches the current virtual link.

[0071] Specifically, a filtering layer determines multiple physical links that satisfy the second constraint condition in the first physical link subset: the filtering layer of the second neural network model determines physical links in the first physical link subset whose available resources are greater than or equal to the resources required to deploy the current virtual link.

[0072] As an optional implementation, the target neural network model described above is obtained through machine learning and can be trained in the following way:

[0073] Obtain the sample virtual undirected weighted graph of the sample virtual network, wherein the sample virtual undirected weighted graph further includes: the resources required to deploy the sample virtual nodes, and the resources required to deploy the sample virtual links;

[0074] Obtain a sample physical undirected weighted graph of the sample physical links, wherein the sample physical undirected weighted graph further includes: the available resources of the sample physical nodes, and the available resources of the sample physical links;

[0075] The initial neural network model is trained using the sample virtual undirected weighted graph and the sample physical undirected weighted graph. The training ends when the total time for deploying the sample virtual network to the sample physical network and the benefit-cost ratio reach the convergence condition, thus obtaining the target neural network model.

[0076] Wherein, when the total time for deploying the sample virtual network to the sample physical network reaches the convergence condition between the benefit-cost ratio, training ends to obtain the target neural network model, including:

[0077] Determine the total time for deploying the sample virtual node to the sample physical node and the sample virtual link to the sample physical link, including:

[0078] The duration for which the sample virtual node is deployed to the sample physical node is determined as the first duration; the duration for which the sample virtual link is deployed to the sample physical link is determined as the second duration; and the sum of the first duration and the second duration is determined as the total duration.

[0079] Specifically, the total duration can be obtained using the following formula:

[0080]

[0081] in, Indicates the deployment of sample virtual nodes Required time Indicates the deployment of sample virtual links l vr (k,l) represents the required time. It is a binary variable, specifically defined as follows: Successfully deployed to hour Not deployed to hour l vr (k,l) was successfully deployed to l. s When (i,j), l vr (k,l) is not deployed to l sWhen (i,j),

[0082] Determining the cost-benefit ratio of deploying the sample virtual node to the sample physical node and deploying the sample virtual link to the sample physical link includes: determining the amount of central processing unit resources occupied by deploying the sample virtual node to the sample physical node as a first resource amount;

[0083] The amount of bandwidth resources occupied by the sample virtual link deployed to the sample physical link is determined as the second resource amount; the revenue-cost ratio is obtained by using the first resource amount and the second resource amount;

[0084] Specifically, the benefit-cost ratio formula is as follows:

[0085]

[0086] in, It is to make virtual nodes The amount of central processing unit resources used by the deployment to physical nodes. It is to make virtual link l vr (k,l) represents the amount of bandwidth resources used to deploy to the physical link.

[0087] It is a binary variable, specifically defined as follows: Successfully deployed to hour Not deployed to hour l vr (k,l) was successfully deployed to l. s When (i,j), l vr (k,l) is not deployed to l s When (i,j),

[0088] The target optimization value is determined based on the total duration and the benefit-cost ratio;

[0089] Specifically, the target optimization value S can be obtained using the following formula:

[0090]

[0091] Here, α and β are preset values ​​that can be set according to the situation, for example, α = 0.5, β = 0.5.

[0092] If the target optimization value reaches the preset value and the preset conditions are met, the training ends and the target neural network model is obtained.

[0093] The physical network in the above embodiments can be a satellite network. A cloud-edge-device collaborative satellite network resource allocation method is proposed to rationally and effectively adjust resources based on spatiotemporal differences, maximizing resource utilization. In this mechanism, the cloud provides powerful computing and storage capabilities for processing large-scale data and complex tasks, while edge and terminal devices provide low-latency, high-reliability computing and services. Specifically addressing the heterogeneity and time-varying nature of cloud-edge-device collaborative satellite networks, this method can improve resource utilization while minimizing latency.

[0094] Figure 3 This is a schematic diagram illustrating the deployment of a virtual network according to an embodiment of the present invention, comprising two virtual networks and one physical network. Virtual network 1 contains 5 virtual nodes and 6 virtual links, with the virtual nodes represented by hexagons. Virtual network 2 contains 4 virtual nodes and 4 virtual links, with the virtual nodes represented by circles. The physical network contains 12 virtual nodes and 15 virtual links, with the virtual nodes represented by squares.

[0095] The following is in conjunction with the appendix Figure 3 The embodiments of the present invention will be further explained with specific examples:

[0096] This example uses the Networkx tool to build a cloud-edge-device collaborative satellite underlying network topology. The satellite underlying network in this example consists of 20 satellite nodes, 60 cloud server nodes, 120 edge server nodes, and 800 links. Satellite node computing resources are randomly generated within the range of 30-100 units; cloud server node computing resources are randomly generated within the range of 100-200 units; edge server node computing resources are randomly generated within the range of 20-80 units; and link bandwidth capacity is randomly generated within the range of 10-100 units. The virtual network in this example is divided into a test set and a training set. Each virtual network request contains 3-15 nodes, with a 50% probability of a link existing between nodes. The computing resource requirements of each node are set to a range of 3-60 units, randomly generated within this range. The bandwidth resource requirements of each link are randomly generated within the range of 3-30 units. The number of virtual network requests follows a Poisson distribution. In this embodiment, 1500 virtual network requests are used as a training set to train the virtual network mapping model based on graph neural networks and deep reinforcement learning proposed in this invention, and 1500 virtual network requests are used as a test set to evaluate the performance of the virtual network mapping model proposed in this invention.

[0097] The specific algorithm for virtual network mapping in this example models the relationship between the physical network, virtual network, and computing resources as a graph structure. Nodes can represent tasks, computing nodes, etc., and edges can represent connections or communication needs between resources. The graph structure is input to the constructed policy network, which receives the state vector as input. Through multi-layer graph convolution, it iteratively aggregates the information of nodes and their neighbors, gradually updating the node representation to better capture the local and global information of the graph. The output is a probability distribution representing the probability of selecting each node in a given state. The parameters of the policy network are randomly generated, and the reward function includes minimizing latency and maximizing resource utilization. The policy network parameters are updated using gradient descent until a predetermined number of iterations is reached. Finally, the virtual network embedding rate, long-term revenue-cost ratio, resource utilization, and mapping latency are used as performance indicators. Experiments show that this method outperforms other heuristic or deep learning-based mapping methods in terms of virtual network embedding rate, long-term revenue-cost ratio, resource utilization, and mapping latency.

[0098] A virtual node mapping model (first neural network model) is constructed using graph neural networks (GNNs) and deep reinforcement learning techniques. This first network model consists of an input layer, convolutional layers, a probability layer, a filtering layer, and an output layer. The satellite node states are input to the input layer, which then passes the undirected weighted graph data to the convolutional layers. The structure of the GNN is embedded within the convolutional layers, forming the graph convolutional layer. This layer extracts and learns useful features from the input, which help solve specific tasks. For the graph data of the satellite network topology, the GNN's graph convolutional layer can progressively extract information from the graph at each layer by considering the relationships between nodes and their neighbors, obtaining the available resource vectors of the nodes. This facilitates the probability layer in making mapping probabilities for each physical node that better match the current network state. The probability layer is mainly responsible for calculating the mapping probability of the current virtual node among each physical node based on the node's available resource vector. The filtering layer removes nodes that do not meet the requirements according to various constraints, and the output layer outputs the physical node with the highest probability of being mapped to the current virtual node.

[0099] A virtual link mapping model (second neural network model) is constructed using graph neural networks and deep reinforcement learning techniques. The policy network of the same model as the node mapping model is constructed using the same method. The policy network consists of an input layer, convolutional layers, probability layers, filtering layers, and an output layer. The convolutional layers use graph convolutional networks to extract link features. The link mapping model can be trained using a pre-trained node mapping model to obtain the total time and cost-benefit ratio of the overall mapping process.

[0100] This application is applicable to highly dynamic satellite networks. For real-world satellite network applications, when users access satellite networks, it is necessary to consider the available computing resources, available bandwidth resources, processing latency, and other characteristics of satellite nodes. A graph convolutional-based policy network is used to extract the characteristic states of the dynamic satellite network topology. The graph convolutional layers can progressively extract information from the graph at each layer by considering the relationships between nodes and their neighbors to obtain the available resource vector of the satellite network, thereby improving the ability to perceive dynamic satellite networks.

[0101] It enables coordinated optimization of latency and resource utilization. When training the virtual network mapping model, we minimize the latency-to-cost ratio as the reward function, comprehensively considering latency, node computing resources, and link bandwidth resources. This results in low latency and high resource utilization in the mapping, thereby improving the utilization of the underlying resources of the satellite network and enhancing the user's service experience.

[0102] This application has the following technical effects:

[0103] By establishing a suitable virtual network request model, we can better understand the characteristics of virtual network requests, providing a foundation for subsequent mapping. Modeling the satellite physical network is equally crucial, requiring consideration of its topology, communication performance, resource constraints, and its mapping relationship with virtual network requests. Establishing an accurate satellite physical network model can better support the mapping and management of virtual networks.

[0104] This application strikes a balance between latency and energy consumption. By comprehensively considering multiple objectives, it can better meet the diverse needs of satellite networks and improve overall performance.

[0105] The model possesses adaptive learning capabilities, allowing it to self-adjust and optimize based on changes in virtual network requests and current network conditions. Achieving this adaptive learning capability requires the model to have strong generalization ability and the capacity to cope with unknown situations; therefore, realizing this adaptive learning capability is a key aspect of this proposal.

[0106] This proposal employs techniques such as graph neural networks and deep reinforcement learning, which can efficiently process large-scale data and perform complex calculations. By applying these techniques, the computational power and processing efficiency of this proposal can be significantly improved, thereby better supporting the mapping and management of virtual networks.

[0107] It can respond to virtual network requests in real time and generate corresponding mapping strategies, thus ensuring real-time performance. This real-time performance guarantee is crucial for satellite networks because their dynamic and heterogeneous nature requires timely response and processing.

[0108] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0109] This embodiment also provides an apparatus for implementing the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the apparatus described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0110] Figure 4 This is a structural block diagram of an apparatus for allocating physical network resources according to an embodiment of the present invention, such as... Figure 4 As shown, the device includes:

[0111] The first acquisition module 402 is used to acquire a virtual undirected weighted graph of the virtual network, wherein the virtual undirected weighted graph is used to represent the set of virtual nodes and virtual links contained in the virtual network, as well as the resources required to deploy the virtual network;

[0112] The second acquisition module 404 is used to acquire a physical undirected weighted graph of the physical network, wherein the physical undirected weighted graph is a set of physical nodes and physical links contained in the physical network, as well as the resources owned by the physical network.

[0113] Deployment module 406 is used to input the virtual undirected weighted graph and the physical undirected weighted graph into the target neural network model, and through the target neural network model, deploy the virtual nodes and virtual links of the virtual network to the physical network.

[0114] The target neural network model is trained using a sample virtual undirected weighted graph and a sample physical undirected weighted graph. The sample virtual undirected weighted graph represents the sample virtual nodes and sample virtual links contained in the sample virtual network, as well as the set of resources required to deploy the sample virtual network. The sample physical undirected weighted graph represents the sample physical nodes and sample physical links contained in the sample physical network, as well as the set of resources possessed by the sample physical network.

[0115] The target neural network model achieves a convergence condition between the total time and the cost-benefit ratio of deploying the sample virtual network to the sample physical network, as preset.

[0116] In an exemplary embodiment, the apparatus is further configured to input the virtual undirected weighted graph and the physical undirected weighted graph into a first neural network model, determine physical nodes matching each virtual node in the set of physical nodes through the first neural network model, thereby obtaining a first physical node subset, wherein the target neural network model includes the first neural network model, the virtual undirected weighted graph includes the set of virtual nodes in the virtual network, and the physical undirected weighted graph includes the set of physical nodes in the physical network; determine a physical weighted subgraph in the physical undirected weighted graph, wherein the physical weighted subgraph includes the first physical node subset and a first physical link subset, the first physical link subset including links between each physical node in the first physical node subset; and deploy the virtual nodes and virtual links of the virtual network to the physical network through the physical weighted subgraph and the virtual undirected weighted graph.

[0117] In an exemplary embodiment, the above-described apparatus is further configured to input the virtual undirected weighted graph and the physical weighted subgraph into a second neural network model, determine a target link in the first physical link subset through the second neural network model, and obtain a target physical link subset, wherein the target neural network model includes the second neural network model; determine the network composed of the first physical node subset and the target physical link subset in the physical network as a target physical subnetwork, and deploy the virtual nodes and the virtual links of the virtual network to the target physical subnetwork.

[0118] In an exemplary embodiment, the virtual undirected weighted graph further includes: resources required to deploy the virtual nodes, and the physical undirected weighted graph further includes: available resources of the physical nodes. The apparatus is further configured to perform the following operations on each of the virtual nodes in the set of virtual nodes using the first neural network model, wherein the virtual node performing the following operations is designated as the current virtual node: extracting features from the current virtual node, its neighboring nodes, the virtual links between the current virtual node and its neighboring nodes, and the resources required to deploy the current virtual node using the convolutional layer of the first neural network model to obtain a first feature matrix; extracting features from each of the physical nodes in the set of physical nodes, the physical links between the physical nodes, and the available resources of each physical node using the convolutional layer of the first neural network model to obtain a second feature matrix; and determining a current physical node in the set of physical nodes that matches the current virtual node using the first feature matrix and the second feature matrix, wherein the first subset of physical nodes includes the current physical node.

[0119] In an exemplary embodiment, the above-described apparatus is further configured to input the first feature matrix and the second feature matrix into the probability layer of the first neural network model, and obtain the mapping probability of each physical node in the physical node set through the probability layer; input the first feature matrix and the second feature matrix into the filtering layer of the first neural network model, and determine multiple physical nodes in the physical node set that satisfy preset constraints through the filtering layer; and determine the node with the highest mapping probability among the multiple physical nodes that satisfy the preset constraints as the current physical node that matches the current virtual node.

[0120] In an exemplary embodiment, the above-described apparatus is further configured to determine, through the filtering layer, physical nodes in the set of physical nodes whose available resources are greater than or equal to the resources required to deploy the current virtual node.

[0121] In an exemplary embodiment, the virtual undirected weighted graph further includes: resources required to deploy the virtual link, and the physical undirected weighted graph further includes: available resources of the physical link. The apparatus is further configured to perform the following operations on each virtual link in the virtual link set using the second neural network model, wherein the virtual link performing the following operations is designated as the current virtual link: extracting features from the current virtual link, the virtual nodes connected to the current virtual link, and the resources required to deploy the current virtual link using the convolutional layer of the second neural network model to obtain a third feature matrix; extracting features from each physical link in the first physical link subset, the physical nodes connected to the physical link, and the available resources of the physical link using the convolutional layer of the second neural network model to obtain a fourth feature matrix; and determining a current physical link matching the current virtual link in the first physical link subset using the third feature matrix and the fourth feature matrix, wherein the target physical link subset includes the current physical link.

[0122] In an exemplary embodiment, the above-described apparatus is further configured to input the third feature matrix and the fourth feature matrix into the probability layer of the second neural network model, and obtain the mapping probability of each physical link in the first physical link subset through the probability layer; input the third feature matrix and the fourth feature matrix into the filtering layer of the second neural network model, and determine multiple physical links satisfying the second constraint condition in the first physical link subset through the filtering layer; and determine the physical link with the highest mapping probability among the multiple physical links satisfying the second constraint condition as the current physical link that matches the current virtual link.

[0123] In an exemplary embodiment, the above-described apparatus is further configured to determine, within the first subset of physical links, physical links whose available resources are greater than or equal to the resources required to deploy the current virtual link, through the filtering layer of the second neural network model.

[0124] In an exemplary embodiment, the apparatus is further configured to, before inputting the virtual undirected weighted graph and the physical undirected weighted graph into the target neural network model, obtain the sample virtual undirected weighted graph of the sample virtual network, wherein the sample virtual undirected weighted graph further includes: resources required to deploy the sample virtual nodes and resources required to deploy the sample virtual links; obtain the sample physical undirected weighted graph of the sample physical links, wherein the sample physical undirected weighted graph further includes: available resources of the sample physical nodes and available resources of the sample physical links; train the initial neural network model using the sample virtual undirected weighted graph and the sample physical undirected weighted graph, and terminate the training to obtain the target neural network model when the convergence condition is met between the total time and the benefit-cost ratio of deploying the sample virtual network to the sample physical network.

[0125] In an exemplary embodiment, the above-described apparatus is further configured to: determine the total duration of deploying the sample virtual node to the sample physical node and deploying the sample virtual link to the sample physical link; determine the benefit-cost ratio of deploying the sample virtual node to the sample physical node and deploying the sample virtual link to the sample physical link; determine a target optimization value based on the total duration and the benefit-cost ratio; and, if the target optimization value reaches a preset value, determine that the convergence condition is met and training is terminated to obtain the target neural network model.

[0126] In an exemplary embodiment, the above-described apparatus is further configured to determine the duration for which the sample virtual node is deployed to the sample physical node as a first duration; determine the duration for which the sample virtual link is deployed to the sample physical link as a second duration; and determine the sum of the first duration and the second duration as the total duration.

[0127] In an exemplary embodiment, the above-described apparatus is further configured to determine the amount of central processing unit resources occupied by the sample virtual node deployed to the sample physical node as a first resource amount; determine the amount of bandwidth resources occupied by the sample virtual link deployed to the sample physical link as a second resource amount; and obtain the revenue-cost ratio through the first resource amount and the second resource amount.

[0128] It should be noted that the above modules can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above modules are located in the same processor; or, the above modules are located in different processors in any combination.

[0129] Embodiments of the present invention also provide a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of the method described in any of the preceding claims.

[0130] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0131] Embodiments of the present invention also provide an electronic device including a memory and a processor, the memory storing a computer program and the processor being configured to run the computer program to perform the steps in any of the above method embodiments.

[0132] In one exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.

[0133] Specific examples in this embodiment can be found in the examples described in the above embodiments and exemplary implementations, and will not be repeated here.

[0134] Embodiments of the present invention also provide a computer program product, including a computer program that, when executed by a processor, implements the steps of the methods described in various embodiments of the present application.

[0135] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. They can be implemented using computer-executable program code, and thus can be stored in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those described herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0136] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for allocating physical network resources, characterized in that, include: Obtain a virtual undirected weighted graph, wherein the virtual undirected weighted graph is used to represent the set of virtual nodes and virtual links contained in the virtual network, as well as the resources required to deploy the virtual network; Obtain a physical undirected weighted graph, wherein the physical undirected weighted graph is used to represent the set of physical nodes and physical links contained in the physical network, as well as the resources owned by the physical network; The virtual undirected weighted graph and the physical undirected weighted graph are input into the target neural network model, and the virtual nodes and virtual links of the virtual network are deployed to the physical network through the target neural network model; The target neural network model is trained using a sample virtual undirected weighted graph and a sample physical undirected weighted graph. The sample virtual undirected weighted graph represents the sample virtual nodes and sample virtual links contained in the sample virtual network, as well as the set of resources required to deploy the sample virtual network. The sample physical undirected weighted graph represents the sample physical nodes and sample physical links contained in the sample physical network, as well as the set of resources possessed by the sample physical network. The target neural network model achieves a convergence condition between the total time and the cost-benefit ratio of deploying the sample virtual network to the sample physical network, as preset.

2. The method according to claim 1, characterized in that, Deploying the virtual nodes and virtual links of the virtual network to the physical network using the target neural network model includes: The virtual undirected weighted graph and the physical undirected weighted graph are input into a first neural network model. The first neural network model determines the physical nodes that match each virtual node in the set of physical nodes in the set of virtual nodes, thus obtaining a first physical node subset. The target neural network model includes the first neural network model, the virtual undirected weighted graph includes the set of virtual nodes in the virtual network, and the physical undirected weighted graph includes the set of physical nodes in the physical network. In the physical undirected weighted graph, a physical weighted subgraph is determined, wherein the physical weighted subgraph includes the first physical node subset and the first physical link subset, and the first physical link subset includes the links between each physical node in the first physical node subset; The virtual nodes and virtual links of the virtual network are deployed to the physical network through the physical weighted subgraph and the virtual undirected weighted graph.

3. The method according to claim 2, characterized in that, Deploying the virtual nodes and virtual links of the virtual network to the physical network through the physical weighted subgraph and the virtual undirected weighted graph includes: The virtual undirected weighted graph and the physical weighted subgraph are input into the second neural network model. The target link is determined in the first physical link subset by the second neural network model to obtain the target physical link subset. The target neural network model includes the second neural network model. The network consisting of the first subset of physical nodes and the target subset of physical links in the physical network is determined as the target physical sub-network, and the virtual nodes and virtual links of the virtual network are deployed to the target physical sub-network.

4. The method according to claim 2, characterized in that, The virtual undirected weighted graph further includes: the resources required to deploy the virtual nodes; the physical undirected weighted graph further includes: the available resources of the physical nodes; and the first physical node subset is obtained by determining, through the first neural network model, physical nodes that match each of the virtual nodes in the set of physical nodes, thus obtaining a first physical node subset, including: The first neural network model performs the following operations on each of the virtual nodes in the set of virtual nodes, wherein the virtual node performing the following operations is designated as the current virtual node: The convolutional layer of the first neural network model extracts features from the current virtual node, the neighboring nodes of the current virtual node, the virtual links between the current virtual node and the neighboring nodes, and the resources required to deploy the current virtual node to obtain a first feature matrix; The convolutional layer of the first neural network model extracts features from each physical node in the set of physical nodes, the physical links between the physical nodes, and the available resources of each physical node to obtain a second feature matrix; The first feature matrix and the second feature matrix are used to determine the current physical node that matches the current virtual node in the set of physical nodes, wherein the first subset of physical nodes includes the current physical node.

5. The method according to claim 4, characterized in that, Determining the current physical node matching the current virtual node from the set of physical nodes using the first feature matrix and the second feature matrix includes: The first feature matrix and the second feature matrix are input into the probability layer of the first neural network model, and the mapping probability of each physical node in the physical node set is obtained through the probability layer. The first feature matrix and the second feature matrix are input into the filtering layer of the first neural network model, and the filtering layer determines multiple physical nodes in the physical node set that meet the preset constraints. The node with the highest mapping probability among multiple physical nodes that satisfy the preset constraints is determined as the current physical node that matches the current virtual node.

6. The method according to claim 5, characterized in that, The filtering layer determines multiple physical nodes in the set of physical nodes that satisfy preset constraints, including: The filtering layer identifies physical nodes in the set of physical nodes whose available resources are greater than or equal to the resources required to deploy the current virtual node.

7. The method according to claim 3, characterized in that, The virtual undirected weighted graph further includes: resources required to deploy the virtual links; the physical undirected weighted graph further includes: available resources of the physical links; the virtual undirected weighted graph and the physical weighted subgraph are input into a second neural network model; the second neural network model determines the target links in the first physical link subset, resulting in a target physical link subset, including: The second neural network model performs the following operations on each virtual link in the virtual link set, wherein the virtual link performing the following operations is designated as the current virtual link: The convolutional layer of the second neural network model extracts features from the current virtual link, the virtual nodes connected to the current virtual link, and the resources required to deploy the current virtual link, to obtain a third feature matrix; The convolutional layer of the second neural network model extracts features from each physical link in the first physical link subset, the physical nodes connected by the physical links, and the available resources of the physical links to obtain a fourth feature matrix; The target physical link subset includes the current physical link, which is determined in the first physical link subset by using the third feature matrix and the fourth feature matrix.

8. The method according to claim 7, characterized in that, Determining the current physical link matching the current virtual link in the first physical link subset using the third feature matrix and the fourth feature matrix includes: The third feature matrix and the fourth feature matrix are input into the probability layer of the second neural network model, and the mapping probability of each physical link in the first physical link subset is obtained through the probability layer. The third feature matrix and the fourth feature matrix are input into the filtering layer of the second neural network model, and the filtering layer determines multiple physical links that satisfy the second constraint condition in the first physical link subset. The physical link with the highest mapping probability among the plurality of physical links that satisfy the second constraint is determined as the current physical link that matches the current virtual link.

9. The method according to claim 8, characterized in that, The filtering layer identifies multiple physical links satisfying the second constraint condition within the first subset of physical links, including: The filtering layer of the second neural network model determines physical links in the first physical link subset whose available resources are greater than or equal to the resources required to deploy the current virtual link.

10. The method according to claim 1, characterized in that, Before inputting the virtual undirected weighted graph and the physical undirected weighted graph into the target neural network model, the method further includes: Obtain the sample virtual undirected weighted graph of the sample virtual network, wherein the sample virtual undirected weighted graph further includes: the resources required to deploy the sample virtual nodes, and the resources required to deploy the sample virtual links; Obtain a sample physical undirected weighted graph of the sample physical links, wherein the sample physical undirected weighted graph further includes: the available resources of the sample physical nodes, and the available resources of the sample physical links; The initial neural network model is trained using the sample virtual undirected weighted graph and the sample physical undirected weighted graph. The training ends when the total time and cost-benefit ratio of deploying the sample virtual network to the sample physical network meet the convergence condition, thus obtaining the target neural network model.

11. The method according to claim 10, characterized in that, When the convergence condition is met between the total time and the cost-benefit ratio of deploying the sample virtual network to the sample physical network, training ends to obtain the target neural network model, including: Determine the total duration for deploying the sample virtual node to the sample physical node and the sample virtual link to the sample physical link; Determine the cost-benefit ratio of deploying the sample virtual node to the sample physical node and deploying the sample virtual link to the sample physical link; The target optimization value is determined based on the total duration and the benefit-cost ratio; If the target optimization value reaches the preset value, the convergence condition is satisfied, and the training ends to obtain the target neural network model.

12. The method according to claim 11, characterized in that, Determine the total time for deploying the sample virtual node to the sample physical node and the sample virtual link to the sample physical link, including: The duration for which the sample virtual node is deployed to the sample physical node is defined as the first duration; The duration for which the sample virtual link is deployed to the sample physical link is determined as the second duration; The sum of the first duration and the second duration is determined as the total duration.

13. The method according to claim 11, characterized in that, Determining the cost-benefit ratio of deploying the sample virtual node to the sample physical node versus deploying the sample virtual link to the sample physical link includes: The amount of central processing unit resources occupied by the sample virtual node deployed to the sample physical node is defined as the first resource amount. The amount of bandwidth resources occupied by the sample virtual link deployed to the sample physical link is determined as the second resource amount; The revenue-cost ratio is obtained by using the first resource quantity and the second resource quantity.

14. An apparatus for allocating physical network resources, characterized in that, include: The first acquisition module is used to acquire a virtual undirected weighted graph of the virtual network, wherein the virtual undirected weighted graph is used to represent the set of virtual nodes and virtual links contained in the virtual network, as well as the resources required to deploy the virtual network; The second acquisition module is used to acquire the physical undirected weighted graph of the physical network, wherein the physical undirected weighted graph is used to represent the physical nodes and physical links contained in the physical network, as well as the set of resources owned by the physical network. The deployment module is used to input the virtual undirected weighted graph and the physical undirected weighted graph into the target neural network model, and through the target neural network, the model deploys the virtual nodes and virtual links of the virtual network to the physical network. The target neural network model is trained using a sample virtual undirected weighted graph and a sample physical undirected weighted graph. The sample virtual undirected weighted graph represents the sample virtual nodes and sample virtual links contained in the sample virtual network, as well as the set of resources required to deploy the sample virtual network. The sample physical undirected weighted graph represents the sample physical nodes and sample physical links contained in the sample physical network, as well as the set of resources possessed by the sample physical network. The target neural network model achieves a convergence condition between the total time and the cost-benefit ratio of deploying the sample virtual network to the sample physical network, as preset.

15. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein the computer program, when executed by a processor, implements the steps of the method described in any one of claims 1 to 13.

16. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to perform the method as described in any one of claims 1 to 13.

17. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method described in any one of claims 1 to 13.

Images