CN111106960B - Mapping method and mapping device of virtual network and readable storage medium - Google Patents
Mapping method and mapping device of virtual network and readable storage medium Download PDFInfo
- Publication number
- CN111106960B CN111106960B CN201911337626.0A CN201911337626A CN111106960B CN 111106960 B CN111106960 B CN 111106960B CN 201911337626 A CN201911337626 A CN 201911337626A CN 111106960 B CN111106960 B CN 111106960B
- Authority
- CN
- China
- Prior art keywords
- network
- sample
- node
- physical
- virtual
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/08—Configuration management of networks or network elements
- H04L41/0893—Assignment of logical groups to network elements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The application provides a mapping method, a mapping device and a readable storage medium of a virtual network, wherein the mapping method of the virtual network comprises the following steps: generating a node characteristic vector of each physical network node based on node information of each physical network node in a physical network to be mapped; constructing a network characteristic matrix of the physical network to be mapped based on the node characteristic vector of each physical network node; and inputting the network characteristic matrix into a pre-trained network mapping model, and acquiring mapping position information of each virtual network node in the virtual network after the physical network nodes are mapped to the virtual network, so that the mapping result is more accurate, the time consumed in the node mapping process is effectively reduced, and the node mapping efficiency and the resource utilization rate of the physical network are favorably improved.
Description
Technical Field
The present application relates to the field of communications technologies, and in particular, to a mapping method, a mapping apparatus, and a readable storage medium for a virtual network.
Background
With the continuous development of network technology and networking services, more and more people realize resource sharing through networks, and the number and the scale of the networks increase exponentially. Network virtualization can share a plurality of virtual networks on the basis of one physical network, so that the utilization rate of physical network resources is improved. Different virtual network mapping problems may arise in virtual networks for different optimization goals and application scenarios.
At present, in the process of mapping an actual physical network to a virtual network, direct solutions or heuristic solutions are mostly adopted for mapping, but due to high computation time and computation complexity, the direct solutions are gradually unable to be applied to an increasingly large-scale network, and heuristic solutions are mostly based on experience, nodes and the like are statically configured, so that the embedding effect is poor, the accuracy is low, and a deviation exists between the mapping result and the actual demand, therefore, how to efficiently, accurately and quickly virtualize the network is a problem to be solved urgently at present.
Disclosure of Invention
In view of this, an object of the present application is to provide a mapping method, a mapping apparatus, and a readable storage medium for a virtual network, which can map a physical network node into the virtual network by dynamically referring to a connection between nodes in the same virtual request through a trained network mapping model in combination with node information of each physical network node in the physical network, so that a mapping result is more accurate, time consumed in a node mapping process is effectively reduced, and node mapping efficiency and resource utilization rate of the physical network are improved.
The embodiment of the application provides a mapping method of a virtual network, which comprises the following steps:
generating a node characteristic vector of each physical network node based on node information of each physical network node in a physical network to be mapped;
constructing a network characteristic matrix of the physical network to be mapped based on the node characteristic vector of each physical network node;
and inputting the network characteristic matrix into a pre-trained network mapping model, and acquiring mapping position information of each virtual network node in the virtual network after the physical network node is mapped to the virtual network.
Further, the inputting the network feature matrix into a pre-trained network mapping model, and obtaining the location information of each virtual network node in the virtual network after mapping the physical network node into the virtual network, includes:
inputting the network characteristic matrix into a coding network layer in the network mapping model, and determining a network characteristic vector of the physical network;
and inputting the network characteristic vector into a decoding network layer in a network mapping model, and determining the position information of each virtual network node in the virtual network after the physical network node is mapped to the virtual network.
Further, the inputting the network feature vector into a decoding network layer in a network mapping model to determine location information of each virtual network node in the virtual network after the physical network node is mapped to the virtual network includes:
taking the network characteristic vector as a decoding output vector at the current moment output by a decoding network layer in the network mapping model, and acquiring a current-moment intermediate network parameter output by the decoding network layer corresponding to the current moment;
sequentially inputting the decoding output vector at the current moment and the intermediate network parameter at the current moment output by the decoding network layer into the decoding network layer again, and determining the decoding output vector at the next moment and the intermediate network parameter at the next moment corresponding to the next moment at the current moment;
and determining the position information of each virtual network node in the virtual network when the physical network is mapped to the virtual network based on the determined decoding output vector at each next moment.
Further, for each physical network node, the node information includes at least one of the following information:
computing resource surplus information of the physical network node; a number of neighboring network nodes connected to the physical network node; the product of the total amount of bandwidth of each link bandwidth connected to the physical network node and the remaining amount of computing resources of the physical network node.
Further, the mapping method trains the network mapping model by:
the method comprises the steps of obtaining a plurality of sample physical networks, sample node information of each sample physical network node in each sample physical network, an actual network feature vector of each sample physical network, a sample virtual network corresponding to each sample physical network and actual position information of each sample virtual network node in each sample virtual network in the virtual network;
for each sample physical network, generating a sample node feature vector of each sample physical network node based on the sample node information of each sample physical network node;
constructing a sample network characteristic matrix of each sample physical network based on the sample node characteristic vector of each sample physical network node;
training the constructed recurrent neural network based on the sample network feature matrix of each sample physical network, the actual network feature vector of each sample physical network and the actual position information of each sample virtual network node in each sample virtual network to obtain the network mapping model.
Further, the training of the constructed recurrent neural network based on the sample network feature matrix of each sample physical network, the sample network feature vector of each sample physical network, and the actual location information of each sample virtual network node in each sample virtual network to obtain the network mapping model includes:
for each sample physical network, inputting a sample network feature matrix of the sample physical network into a coding network layer in the recurrent neural network, and determining a sample network feature vector of the sample physical network;
training network parameters in a coding network layer in the recurrent neural network based on the sample network feature vector and the actual network feature vector of each sample physical network;
inputting the actual network feature vector of the sample physical network into a decoding network layer in the recurrent neural network, and determining the sample position information of each sample virtual network node in the sample virtual network after the sample physical network node is mapped to the corresponding sample virtual network;
training a decoding network layer in the recurrent neural network based on the sample position information and the actual position information of each sample virtual network node in each sample virtual network;
and determining the network mapping model based on the trained coding network layer and the trained decoding network layer.
Further, the training a decoding network layer in the recurrent neural network based on the sample location information and the actual location information of each sample virtual network node in each sample virtual network includes:
determining a loss function, a reward value and a training accuracy threshold of a decoding network layer in the recurrent neural network based on sample position information and actual position information of each virtual network node in each sample virtual network;
determining a gradient value of a decoding network layer in the recurrent neural network according to the loss function, the reward value and the training accuracy threshold;
and adjusting the network parameters of a decoding network layer in the recurrent neural network based on the gradient value, and obtaining a trained decoding network layer.
An embodiment of the present application further provides a mapping apparatus for a virtual network, where the mapping apparatus includes:
the generation module is used for generating a node feature vector of each physical network node based on the node information of each physical network node in the physical network to be mapped;
the building module is used for building a network feature matrix of the physical network to be mapped based on the node feature vector of each physical network node generated by the generating module;
and the acquisition module is used for inputting the network characteristic matrix constructed by the construction module into a pre-trained network mapping model and acquiring mapping position information of each virtual network node in the virtual network after the physical network node is mapped to the virtual network.
Further, the obtaining module includes:
the first determining unit is used for inputting the network characteristic matrix constructed by the construction module into a coding network layer in the network mapping model and determining a network characteristic vector of the physical network;
and the second determining unit is used for inputting the network feature vector determined by the first determining unit into a decoding network layer in a network mapping model, and determining the position information of each virtual network node in the virtual network after the physical network node is mapped to the virtual network.
Further, the second determining unit is specifically configured to:
taking the network characteristic vector as a decoding output vector at the current moment output by a decoding network layer in the network mapping model, and acquiring a current-moment intermediate network parameter output by the decoding network layer corresponding to the current moment;
sequentially inputting the decoding output vector at the current moment and the intermediate network parameter at the current moment output by the decoding network layer into the decoding network layer again, and determining the decoding output vector at the next moment and the intermediate network parameter at the next moment corresponding to the next moment at the current moment;
and determining the position information of each virtual network node in the virtual network when the physical network is mapped to the virtual network based on the determined decoding output vector at each next moment.
Further, for each physical network node, the node information includes at least one of the following information:
computing resource surplus information of the physical network node; a number of neighboring network nodes connected to the physical network node; the product of the total amount of bandwidth of each link bandwidth connected to the physical network node and the remaining amount of computing resources of the physical network node.
Further, the mapping apparatus further includes a training module, and the training module includes:
the system comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring a plurality of sample physical networks, sample node information of each sample physical network node in each sample physical network and an actual network feature vector of each sample physical network, and a sample virtual network corresponding to each sample physical network and actual position information of each sample virtual network node in each sample virtual network in the virtual network;
a generating unit, configured to generate, for each sample physical network, a sample node feature vector of each sample physical network node based on the sample node information of each sample physical network node acquired by the acquiring unit;
the construction unit is used for constructing a sample network characteristic matrix of each sample physical network based on the sample node characteristic vector of each sample physical network node generated by the generation unit;
and the training unit is used for training the constructed recurrent neural network based on the sample network feature matrix of each sample physical network, the actual network feature vector of each sample physical network and the actual position information of each sample virtual network node in each sample virtual network, which are constructed by the construction unit, so as to obtain the network mapping model.
Further, the training unit is specifically configured to:
for each sample physical network, inputting a sample network feature matrix of the sample physical network into a coding network layer in the recurrent neural network, and determining a sample network feature vector of the sample physical network;
training network parameters in a coding network layer in the recurrent neural network based on the sample network feature vector and the actual network feature vector of each sample physical network;
inputting the actual network feature vector of the sample physical network into a decoding network layer in the recurrent neural network, and determining the sample position information of each virtual network node in the sample virtual network after the sample physical network node is mapped to the corresponding sample virtual network;
training a decoding network layer in the recurrent neural network based on the sample position information and the actual position information of each sample virtual network node in each sample virtual network;
and determining the network mapping model based on the trained coding network layer and the trained decoding network layer.
Further, when the training unit trains the decoding network layer in the recurrent neural network based on the sample location information and the actual location information of each sample virtual network node in each sample virtual network, the training unit is specifically configured to:
determining a loss function, a reward value and a training accuracy threshold of a decoding network layer in the recurrent neural network based on sample position information and actual position information of each virtual network node in each sample virtual network;
determining a gradient value of a decoding network layer in the recurrent neural network according to the loss function, the reward value and the training accuracy threshold;
and adjusting the network parameters of a decoding network layer in the recurrent neural network based on the gradient value, and obtaining a trained decoding network layer.
An embodiment of the present application further provides an electronic device, including: a processor, a memory and a bus, the memory storing machine-readable instructions executable by the processor, the processor and the memory communicating via the bus when the electronic device is running, the machine-readable instructions when executed by the processor performing the steps of the mapping method of virtual networks as described above.
An embodiment of the present application further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of the mapping method for a virtual network as described above are performed.
The mapping method, the mapping device and the readable storage medium of the virtual network provided by the embodiment of the application generate the node feature vector of each physical network node based on the node information of each physical network node in the physical network to be mapped; constructing a network characteristic matrix of the physical network to be mapped based on the node characteristic vector of each physical network node; and inputting the network characteristic matrix into a pre-trained network mapping model, and acquiring mapping position information of each virtual network node in the virtual network after the physical network node is mapped to the virtual network.
Therefore, the physical network nodes can be mapped to the virtual network through the network mapping model by combining the node information of each physical network node in the physical network, and the relation among all nodes in the same virtual request is considered, so that the mapping result is more accurate, and the resource utilization rate of the physical network is improved.
In order to make the aforementioned objects, features and advantages of the present application more comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are required to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained from the drawings without inventive effort.
FIG. 1 is a diagram of a system architecture in one possible application scenario;
fig. 2 is a flowchart of a mapping method for a virtual network according to an embodiment of the present disclosure;
fig. 3 is a flowchart of a mapping method for a virtual network according to another embodiment of the present application;
FIG. 4 is a schematic diagram of a network mapping model structure;
fig. 5 is a schematic structural diagram of a mapping apparatus of a virtual network according to an embodiment of the present disclosure;
FIG. 6 is a schematic diagram of the structure of the acquisition module shown in FIG. 5;
FIG. 7 is a schematic diagram of the structure of the training module shown in FIG. 5;
fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all the embodiments. The components of the embodiments of the present application, generally described and illustrated in the figures herein, can be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present application, presented in the accompanying drawings, is not intended to limit the scope of the claimed application, but is merely representative of selected embodiments of the application. Every other embodiment that can be obtained by a person skilled in the art without making creative efforts based on the embodiments of the present application falls within the protection scope of the present application.
First, an application scenario to which the present application is applicable will be described. The method and the device can be applied to the technical field of communication. The physical network nodes can be mapped to the virtual network through the network mapping model by combining the node information of each physical network node in the physical network, and the relation among all the nodes in the same virtual request is considered, so that the mapping result is more accurate, and the resource utilization rate of the physical network is improved. As shown in fig. 1, the system includes an information storage device in which node information of nodes in a physical network is stored, and a mapping device that obtains node information of nodes in the physical network from the information storage device and obtains mapping location information of each virtual network node in a virtual network after the nodes in the physical network are mapped to the virtual network.
Research shows that in the existing virtual network mapping technology, in the node mapping stage, the mapping process of each node of the same virtual request is usually separated independently, so that the association relationship between the nodes in the same virtual request is ignored.
Based on this, the embodiment of the present application provides a mapping method for a virtual network, which may map physical network nodes into the virtual network by dynamically referring to the relations among nodes in the same virtual request through a trained network mapping model in combination with node information of each physical network node in the physical network, so that the mapping result is more accurate, the time consumed in the node mapping process is effectively reduced, and the efficiency of node mapping and the resource utilization rate of the physical network are improved.
Referring to fig. 2, fig. 2 is a flowchart illustrating a mapping method for a virtual network according to an embodiment of the present disclosure. As shown in fig. 2, a mapping method for a virtual network provided in an embodiment of the present application includes:
In the step, node information of each physical network node in the physical network to be mapped is acquired, and a node feature vector of each physical network node is generated based on the acquired node information of each physical network node in the physical network.
Here, the node information includes, for each physical network node, at least one of the following information: computing resource surplus information of the physical network node; a number of neighboring network nodes connected to the physical network node; the product of the total amount of bandwidth of each link bandwidth connected to the physical network node and the remaining amount of computing resources of the physical network node.
In the step, a network feature matrix corresponding to the physical network to be mapped is constructed based on the generated node feature vector of each physical network node.
In the step, a network characteristic matrix corresponding to the constructed physical network to be mapped is input into a pre-trained network mapping model, and mapping position information of virtual network nodes corresponding to the nodes in the physical network to be mapped in the virtual network after the physical network to be mapped is mapped to the virtual network is obtained.
In this way, the physical network node can be mapped into the virtual network according to the identified mapping position information of the virtual network node.
The mapping method of the virtual network provided by the embodiment of the application generates the node characteristic vector of each physical network node based on the node information of each physical network node in the physical network to be mapped; constructing a network characteristic matrix of the physical network to be mapped based on the node characteristic vector of each physical network node; and inputting the network characteristic matrix into a pre-trained network mapping model, and acquiring mapping position information of each virtual network node in the virtual network after the physical network is mapped to the virtual network.
Therefore, the node information of each physical network node in the physical network can be combined, the trained network mapping model dynamically refers to the relation among all nodes in the same virtual request, and the physical network nodes are mapped to the virtual network, so that the mapping result is more accurate, the time consumed in the node mapping process is effectively reduced, and the node mapping efficiency and the resource utilization rate of the physical network are improved.
Referring to fig. 3, fig. 3 is a flowchart of a mapping method for a virtual network according to another embodiment of the present application. As shown in fig. 3, a mapping method for a virtual network provided in an embodiment of the present application includes:
And step 303, inputting the network feature matrix into a coding network layer in a network mapping model, and determining the network feature vector of the physical network.
In the step, the constructed network characteristic matrix corresponding to the physical network to be mapped is input into a coding network layer in a pre-trained network mapping model, the network characteristic matrix corresponding to the physical network to be mapped is coded, and the network characteristic vector corresponding to the physical network is determined.
Inputting a network characteristic vector corresponding to the physical network obtained by coding into a decoding network layer in a pre-trained network mapping model, and decoding the network characteristic vector to determine the position information of each virtual node in the virtual network after the nodes in the physical network are mapped to the virtual network.
The descriptions of step 301 to step 302 may refer to the descriptions of step 201 to step 202, and the same technical effects can be achieved, which are not described in detail.
Further, step 304 further includes taking the network feature vector as a current-time decoding output vector output by a decoding network layer in the network mapping model, and obtaining a current-time intermediate network parameter output by the decoding network layer corresponding to the current time; sequentially inputting the decoding output vector at the current moment and the intermediate network parameter at the current moment output by the decoding network layer into the decoding network layer again, and determining the decoding output vector at the next moment and the intermediate network parameter at the next moment corresponding to the next moment at the current moment; and determining the position information of each virtual network node in the virtual network when the physical network is mapped to the virtual network based on the determined decoding output vector at each next moment.
In this step, as shown in fig. 4, which is a schematic structural diagram of a network mapping model, a network feature vector corresponding to a physical network obtained by encoding is used as a current time decoding output vector (e.g., N1 in fig. 4) output by a decoding network layer in the network mapping model, and a current time intermediate network parameter (e.g., W1 in fig. 4) output by the decoding network layer corresponding to the current time is obtained, the current time decoding output vector output by the decoding network layer and the current time intermediate network parameter output by the decoding network layer corresponding to the current time are sequentially input into the decoding network layer in the network mapping model again, a next time decoding output vector (e.g., N2 in fig. 4) and a next time intermediate network parameter (e.g., W2 in fig. 4) corresponding to the next time of the current time are obtained, step 304 is repeatedly executed until decoding is completed, and based on each next time decoding output vector determined, determining location information for each virtual network node in a virtual network when mapping the physical network to the virtual network.
In a decoding network layer of the network mapping model, decoding is started from the fact that a decoding signal is confirmed to be acquired, and when the decoding ending signal is acquired, decoding is ended; or starting decoding from the confirmation of obtaining the decoding signal, and finishing decoding after all the nodes to be mapped in the physical network are mapped.
The N1 and the N2 can be the same shared network parameter or different network parameters, and if the N1 and the N2 are the same shared network parameter, the calculation amount of the virtual network mapping model training process is reduced; if N1 and N2 are different network parameters, the accuracy of the virtual network mapping model can be improved.
Further, the mapping method trains the network mapping model by:
the method comprises the steps of obtaining a plurality of sample physical networks, sample node information of each sample physical network node in each sample physical network, an actual network feature vector of each sample physical network, a sample virtual network corresponding to each sample physical network, and actual position information of each sample virtual network node in each sample virtual network in the virtual network. For each sample physical network, a sample node feature vector for each sample physical network node is generated based on the sample node information for each sample physical network node. And constructing a sample network feature matrix of each sample physical network based on the sample node feature vector of each sample physical network node. Training the constructed recurrent neural network based on the sample network feature matrix of each sample physical network, the actual network feature vector of each sample physical network and the actual position information of each sample virtual network node in each sample virtual network to obtain the network mapping model.
The method comprises the steps of obtaining a plurality of sample physical networks, and after sample node information of each sample physical network node in each sample physical network, an actual network feature vector of each sample physical network, and a sample virtual network corresponding to each sample physical network are mapped to the virtual network, and the node in each sample physical network is mapped to the virtual network, the actual position information of each physical network node in each sample virtual network in the virtual network; for each sample physical network, obtaining sample node information of each sample physical network node in advance, and generating a sample node feature vector of each sample physical network node based on the sample node information of each sample physical network node. Constructing a corresponding sample network characteristic matrix of each sample physical network based on the generated sample node characteristic vector of each sample physical network node; and then training the constructed recurrent neural network by using the sample network characteristic matrix corresponding to each constructed sample physical network, the pre-acquired actual network characteristic vector of each sample physical network and the pre-acquired actual position information of each sample virtual network node in each sample virtual network, thereby obtaining a network mapping model.
Further, the training of the constructed recurrent neural network based on the sample network feature matrix of each sample physical network, the sample network feature vector of each sample physical network, and the actual location information of each sample virtual network node in each sample physical network to obtain the network mapping model includes: and for each sample physical network, inputting the sample network feature matrix of the sample physical network into an encoding network layer in the recurrent neural network, and determining a sample network feature vector of the sample physical network. Training network parameters in an encoding network layer in the recurrent neural network based on the sample network feature vectors and actual network feature vectors for each sample physical network. And inputting the actual network feature vector of the sample physical network into a decoding network layer in the recurrent neural network, and determining the sample position information of each virtual network node in the sample virtual network after the sample physical network is mapped to the corresponding sample virtual network. Training a decoding network layer in the recurrent neural network based on the sample location information and the actual location information of each virtual network node in each sample virtual network. And determining the network mapping model based on the trained coding network layer and the trained decoding network layer.
In the step, aiming at each sample physical network in a plurality of sample physical networks, inputting a constructed sample network characteristic matrix of each sample physical network into a constructed coding network layer in a recurrent neural network, and determining a sample network characteristic vector of the sample physical network through coding; the method comprises the steps of training network parameters in a coding network layer in a constructed recurrent neural network based on a plurality of constructed sample network characteristic vectors and a pre-acquired actual network characteristic vector of each sample physical network by using the same construction method aiming at the plurality of sample network characteristic vectors constructed by the plurality of sample physical networks, and continuously updating the network parameters in the coding network layer.
And for each sample physical network in a plurality of sample physical networks, inputting a pre-acquired actual network feature vector of the sample physical network into a decoding network layer in the recurrent neural network, and determining sample position information of each virtual network node in the sample virtual network after the node in the sample physical network is mapped to the corresponding sample virtual network. For a plurality of sample physical networks, training a decoding network layer in a recurrent neural network through sample position information of each sample virtual network node in each sample virtual network and pre-acquired actual position information of each sample virtual network node in each sample virtual network in the virtual network, and updating network parameters in the decoding network layer in the recurrent neural network. And determining a network mapping model based on the trained coding network layer and the trained decoding network layer.
Further, the training a decoding network layer in the recurrent neural network based on the sample location information and the actual location information of each sample virtual network node in each sample virtual network includes: determining a loss function, a reward value, and a training accuracy threshold of a decoding network layer in the recurrent neural network based on sample location information and actual location information of each virtual network node in each sample virtual network. And determining a gradient value of a decoding network layer in the recurrent neural network according to the loss function, the reward value and the training accuracy threshold value. And adjusting the network parameters of a decoding network layer in the recurrent neural network based on the gradient value, and obtaining a trained decoding network layer.
In the step, a loss function, a reward value and a training accuracy threshold value of a decoding layer of a cyclic neural network are determined according to sample position information of each sample virtual network node in each sample virtual network and pre-acquired actual position information of each sample virtual network node in each sample virtual network in the virtual network; and calculating to obtain a gradient value of a decoding network layer in the recurrent neural network according to the loss function, the reward value and the training accuracy threshold value, and further adjusting network parameters of the decoding network layer in the recurrent neural network according to the calculated gradient value to obtain a trained decoding network layer.
Wherein the gradient value of the decoding network layer in the recurrent neural network is calculated by the following formula:
g＝α·r·gf；
wherein g is a gradient value of a decoding network layer in the recurrent neural network, alpha is a preset training accuracy threshold value, and r is a reward value of the recurrent neural network.
Wherein, gfCalculated by the following formula:
wherein, L is a loss function, and W is a network parameter in the recurrent neural network.
Wherein r is the reward value of the recurrent neural network, and the reward value of the recurrent neural network is determined by calculating the link mapping result which is successfully mapped to the node in the virtual network through the shortest path algorithm.
Wherein alpha is a preset training accuracy threshold value, and if the preset alpha value is larger, the accuracy of the network mapping model is higher; conversely, the smaller the preset α value, the lower the accuracy of the model, but the faster the training speed of the model.
The mapping method for virtual network mapping provided by the embodiment of the application generates a node feature vector of each physical network node based on the node information of each physical network node in a physical network to be mapped; constructing a network characteristic matrix of the physical network to be mapped based on the node characteristic vector of each physical network node; inputting the network characteristic matrix into a coding network layer in a network mapping model, and determining a network characteristic vector of the physical network; and inputting the network characteristic vector into a decoding network layer in a network mapping model, and determining the position information of each virtual network node in the virtual network after the physical network is mapped to the virtual network.
Therefore, the node characteristic vector from each physical network node is generated based on the acquired node information in the physical network to be mapped, the network characteristic matrix of the physical network to be mapped is constructed through the node characteristic vector from each physical network node, the network characteristic vector of the physical network is determined through the coding network layer in the network mapping model, and then the physical network nodes are mapped to the virtual network through the decoding network layer in the network mapping model, so that the mapping result is more accurate, the time consumed in the node mapping process is effectively reduced, and the node mapping efficiency and the resource utilization rate of the physical network are improved.
Referring to fig. 5 to 7, fig. 5 is a schematic structural diagram of a mapping apparatus of a virtual network according to an embodiment of the present disclosure, and fig. 6 is a structural diagram of an obtaining module shown in fig. 5; fig. 7 is a block diagram of the training module shown in fig. 5. As shown in fig. 5, the mapping apparatus 500 of the virtual network includes:
further, as shown in fig. 5, the mapping apparatus 500 includes:
a generating module 510, configured to generate a node feature vector of each physical network node based on node information of each physical network node in the physical network to be mapped;
a constructing module 520, configured to construct a network feature matrix of the physical network to be mapped based on the node feature vector of each physical network node generated by the generating module 510;
an obtaining module 530, configured to input the network feature matrix constructed by the constructing module 520 into a pre-trained network mapping model, and obtain mapping position information of each virtual network node in the virtual network after the nodes in the physical network are mapped to the virtual network.
Further, as shown in fig. 6, the obtaining module 530 includes:
a first determining unit 531, configured to input the network feature matrix constructed by the constructing module 520 into a coding network layer in the network mapping model, and determine a network feature vector of the physical network;
a second determining unit 532, configured to input the network feature vector determined by the first determining unit 531 into a decoding network layer in a network mapping model, and determine location information of each virtual network node in the virtual network after the node in the physical network is mapped to the virtual network.
Further, as shown in fig. 5, the mapping apparatus further includes a training module 540, and as shown in fig. 7, the training module 540 includes:
an obtaining unit 541, configured to obtain a plurality of sample physical networks, sample node information of each sample physical network node in each sample physical network, and an actual network feature vector of each sample physical network, and a sample virtual network corresponding to each sample physical network and actual location information of each sample virtual network node in each sample virtual network in the virtual network;
a generating unit 542, configured to generate, for each sample physical network, a sample node feature vector of each sample physical network node based on the sample node information of each sample physical network node acquired by the acquiring unit 541;
a constructing unit 543, configured to construct a sample network feature matrix of each sample physical network based on the sample node feature vector of each sample physical network node generated by the generating unit 542;
a training unit 544, configured to train the constructed recurrent neural network based on the sample network feature matrix of each sample physical network, the actual network feature vector of each sample physical network, and the actual location information of each sample virtual network node in each sample virtual network, which are constructed by the construction unit 543, so as to obtain the network mapping model.
The mapping device of the virtual network provided by the embodiment of the application generates the node characteristic vector of each physical network node based on the node information of each physical network node in the physical network to be mapped; constructing a network characteristic matrix of the physical network to be mapped based on the node characteristic vector of each physical network node; and inputting the network characteristic matrix into a pre-trained network mapping model, and acquiring mapping position information of each virtual network node in the virtual network after the nodes in the physical network are mapped to the virtual network.
Therefore, the physical network nodes can be mapped to the virtual network through the network mapping model by combining the node information of each physical network node in the physical network, and the relation among all nodes in the same virtual request is considered, so that the mapping result is more accurate, and the resource utilization rate of the physical network is improved.
Referring to fig. 8, fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the present disclosure. As shown in fig. 8, the electronic device 800 includes a processor 810, a memory 820, and a bus 830.
The memory 820 stores machine-readable instructions executable by the processor 810, when the electronic device 800 runs, the processor 810 communicates with the memory 820 through the bus 830, and when the machine-readable instructions are executed by the processor 810, the steps of the mapping method for the virtual network in the method embodiments shown in fig. 2 and fig. 3 may be executed.
An embodiment of the present application further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the step of the mapping method for a virtual network in the method embodiments shown in fig. 2 and fig. 3 may be executed.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other ways. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units is only one logical division, and there may be other divisions when actually implemented, and for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection of devices or units through some communication interfaces, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a non-volatile computer-readable storage medium executable by a processor. Based on such understanding, the technical solution of the present application or portions thereof that substantially contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
Finally, it should be noted that: the above-mentioned embodiments are only specific embodiments of the present application, and are used for illustrating the technical solutions of the present application, but not limiting the same, and the scope of the present application is not limited thereto, and although the present application is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: any person skilled in the art can modify or easily conceive the technical solutions described in the foregoing embodiments or equivalent substitutes for some technical features within the technical scope disclosed in the present application; such modifications, changes or substitutions do not depart from the spirit and scope of the exemplary embodiments of the present application, and are intended to be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (7)
1. A mapping method of a virtual network, the mapping method comprising:
generating a node characteristic vector of each physical network node based on node information of each physical network node in a physical network to be mapped;
constructing a network characteristic matrix of the physical network to be mapped based on the node characteristic vector of each physical network node;
inputting the network characteristic matrix into a pre-trained network mapping model, and acquiring mapping position information of each virtual network node in a virtual network after the physical network node is mapped to the virtual network;
for each physical network node, the node information comprises at least one of the following information:
computing resource surplus information of the physical network node; a number of neighboring network nodes connected to the physical network node; the product of the total bandwidth of each link bandwidth connected to the physical network node and the remaining amount of computing resources of the physical network node;
the mapping method trains the network mapping model by:
the method comprises the steps of obtaining a plurality of sample physical networks, sample node information of each sample physical network node in each sample physical network, an actual network feature vector of each sample physical network, a sample virtual network corresponding to each sample physical network and actual position information of each sample virtual network node in each sample virtual network in the virtual network;
for each sample physical network, generating a sample node feature vector of each sample physical network node based on the sample node information of each sample physical network node;
constructing a sample network characteristic matrix of each sample physical network based on the sample node characteristic vector of each sample physical network node;
training a constructed recurrent neural network based on a sample network feature matrix of each sample physical network, an actual network feature vector of each sample physical network and actual position information of each sample virtual network node in each sample virtual network to obtain the network mapping model;
the training of the constructed recurrent neural network based on the sample network feature matrix of each sample physical network, the actual network feature vector of each sample physical network and the actual position information of each sample virtual network node in each sample virtual network to obtain the network mapping model comprises:
for each sample physical network, inputting a sample network feature matrix of the sample physical network into a coding network layer in the recurrent neural network, and determining a sample network feature vector of the sample physical network;
training network parameters in a coding network layer in the recurrent neural network based on the sample network feature vector and the actual network feature vector of each sample physical network;
inputting the actual network feature vector of the sample physical network into a decoding network layer in the recurrent neural network, and determining the sample position information of each sample virtual network node in the sample virtual network after the sample physical network node is mapped to the corresponding sample virtual network;
training a decoding network layer in the recurrent neural network based on the sample position information and the actual position information of each sample virtual network node in each sample virtual network;
and determining the network mapping model based on the trained coding network layer and the trained decoding network layer.
2. The mapping method according to claim 1, wherein the inputting the network feature matrix into a pre-trained network mapping model to obtain the location information of each virtual network node in the virtual network after the physical network node is mapped into the virtual network comprises:
inputting the network characteristic matrix into a coding network layer in the network mapping model, and determining a network characteristic vector of the physical network;
inputting the network feature vector into a decoding network layer in the network mapping model, and determining the position information of each virtual network node in the virtual network after the physical network node is mapped to the virtual network.
3. The mapping method according to claim 2, wherein the inputting the network feature vector into a decoding network layer in a network mapping model to determine the location information of each virtual network node in the virtual network after the physical network node is mapped to the virtual network comprises:
taking the network characteristic vector as a decoding output vector at the current moment output by a decoding network layer in the network mapping model, and acquiring a current-moment intermediate network parameter output by the decoding network layer corresponding to the current moment;
sequentially inputting the decoding output vector at the current moment and the intermediate network parameter at the current moment output by the decoding network layer into the decoding network layer again, and determining the decoding output vector at the next moment and the intermediate network parameter at the next moment corresponding to the next moment at the current moment;
and determining the position information of each virtual network node in the virtual network when the physical network is mapped to the virtual network based on the determined decoding output vector at each next moment.
4. The mapping method of claim 1, wherein training a decoding network layer in the recurrent neural network based on sample location information and actual location information for each sample virtual network node in each sample virtual network comprises:
determining a loss function, a reward value and a training accuracy threshold of a decoding network layer in the recurrent neural network based on sample position information and actual position information of each virtual network node in each sample virtual network;
determining a gradient value of a decoding network layer in the recurrent neural network according to the loss function, the reward value and the training accuracy threshold;
and adjusting the network parameters of a decoding network layer in the recurrent neural network based on the gradient value, and obtaining a trained decoding network layer.
5. A mapping apparatus of a virtual network, the mapping apparatus comprising:
the generation module is used for generating a node feature vector of each physical network node based on the node information of each physical network node in the physical network to be mapped;
the building module is used for building a network feature matrix of the physical network to be mapped based on the node feature vector of each physical network node generated by the generating module;
the acquisition module is used for inputting the network characteristic matrix constructed by the construction module into a pre-trained network mapping model and acquiring mapping position information of each virtual network node in a virtual network after the physical network node is mapped to the virtual network;
for each physical network node, the node information comprises at least one of the following information:
computing resource surplus information of the physical network node; a number of neighboring network nodes connected to the physical network node; the product of the total bandwidth of each link bandwidth connected to the physical network node and the remaining amount of computing resources of the physical network node;
the mapping apparatus further includes a training module, the training module including:
the system comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring a plurality of sample physical networks, sample node information of each sample physical network node in each sample physical network and an actual network feature vector of each sample physical network, and a sample virtual network corresponding to each sample physical network and actual position information of each sample virtual network node in each sample virtual network in the virtual network;
a generating unit, configured to generate, for each sample physical network, a sample node feature vector of each sample physical network node based on the sample node information of each sample physical network node acquired by the acquiring unit;
the construction unit is used for constructing a sample network characteristic matrix of each sample physical network based on the sample node characteristic vector of each sample physical network node generated by the generation unit;
the training unit is used for training the constructed recurrent neural network based on the sample network feature matrix of each sample physical network, the actual network feature vector of each sample physical network and the actual position information of each sample virtual network node in each sample virtual network, which are constructed by the construction unit, so as to obtain the network mapping model;
the training unit is specifically configured to:
for each sample physical network, inputting a sample network feature matrix of the sample physical network into a coding network layer in the recurrent neural network, and determining a sample network feature vector of the sample physical network;
training network parameters in a coding network layer in the recurrent neural network based on the sample network feature vector and the actual network feature vector of each sample physical network;
inputting the actual network feature vector of the sample physical network into a decoding network layer in the recurrent neural network, and determining the sample position information of each virtual network node in the sample virtual network after the sample physical network node is mapped to the corresponding sample virtual network;
training a decoding network layer in the recurrent neural network based on the sample position information and the actual position information of each sample virtual network node in each sample virtual network;
and determining the network mapping model based on the trained coding network layer and the trained decoding network layer.
6. An electronic device, comprising: processor, memory and bus, the memory storing machine-readable instructions executable by the processor, the processor and the memory communicating over the bus when the electronic device is running, the machine-readable instructions when executed by the processor performing the steps of the mapping method of virtual networks according to any one of claims 1 to 4.
7. A computer-readable storage medium, characterized in that a computer program is stored thereon, which computer program, when being executed by a processor, performs the steps of the mapping method for virtual networks according to any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911337626.0A CN111106960B (en) | 2019-12-23 | 2019-12-23 | Mapping method and mapping device of virtual network and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911337626.0A CN111106960B (en) | 2019-12-23 | 2019-12-23 | Mapping method and mapping device of virtual network and readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111106960A CN111106960A (en) | 2020-05-05 |
| CN111106960B true CN111106960B (en) | 2021-03-05 |
Family
ID=70423655
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911337626.0A Active CN111106960B (en) | 2019-12-23 | 2019-12-23 | Mapping method and mapping device of virtual network and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111106960B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112436992B (en) * | 2020-11-10 | 2022-01-25 | 北京邮电大学 | Virtual network mapping method and device based on graph convolution network |
| CN113726692B (en) * | 2021-08-30 | 2023-04-28 | 电子科技大学 | Virtual network mapping method and device based on generation of countermeasure network |
| CN115412401B (en) * | 2022-08-26 | 2024-04-19 | 京东科技信息技术有限公司 | Method and device for training virtual network embedding model and virtual network embedding |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108989122A (en) * | 2018-08-07 | 2018-12-11 | 北京邮电大学 | Virtual network requests mapping method, device and realization device |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109754085A (en) * | 2019-01-09 | 2019-05-14 | 中国人民解放军国防科技大学 | Deep reinforcement learning-based large-scale network collapse method, storage device and storage medium |
| CN109995435B (en) * | 2019-01-31 | 2022-09-23 | 全球能源互联网研究院有限公司 | Virtual network mapping method and system based on multi-fiber core optical network |
| CN110365568A (en) * | 2019-06-18 | 2019-10-22 | 西安交通大学 | A kind of mapping method of virtual network based on deeply study |
-
2019
- 2019-12-23 CN CN201911337626.0A patent/CN111106960B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108989122A (en) * | 2018-08-07 | 2018-12-11 | 北京邮电大学 | Virtual network requests mapping method, device and realization device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111106960A (en) | 2020-05-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111106960B (en) | Mapping method and mapping device of virtual network and readable storage medium | |
| CN111709533B (en) | Distributed training method and device of machine learning model and computer equipment | |
| US11018979B2 (en) | System and method for network slicing for service-oriented networks | |
| US11861474B2 (en) | Dynamic placement of computation sub-graphs | |
| CN108681482B (en) | Task scheduling method and device based on graph data | |
| EP4350572A1 (en) | Method, apparatus and system for generating neural network model, devices, medium and program product | |
| CN109408590B (en) | Method, device and equipment for expanding distributed database and storage medium | |
| CN110058936B (en) | Method, apparatus and computer program product for determining an amount of resources of a dedicated processing resource | |
| CN113095474A (en) | Resource usage prediction for deep learning models | |
| CN111143039B (en) | Scheduling method and device of virtual machine and computer storage medium | |
| CN111865644B (en) | Recommendation method and device of computing resources, electronic equipment and storage medium | |
| CN113572697A (en) | Load balancing method based on graph convolution neural network and deep reinforcement learning | |
| CN108280513B (en) | Model generation method and device | |
| CN113627536A (en) | Model training method, video classification method, device, equipment and storage medium | |
| US20200210884A1 (en) | Removing unnecessary history from reinforcement learning state | |
| CN116962438A (en) | Gradient data synchronization method, system, electronic equipment and readable storage medium | |
| CN109905880B (en) | Network partitioning method, system, electronic device and storage medium | |
| CN115292044A (en) | Data processing method and device, electronic equipment and storage medium | |
| CN115456194A (en) | Model training control method, device and system based on asynchronous federal learning | |
| CN107025099B (en) | Asynchronous graph calculation implementation method and system based on double-queue model | |
| CN116755866B (en) | Resource scheduling method and device, electronic equipment and readable storage medium | |
| CN116187464B (en) | Blind quantum computing processing method and device and electronic equipment | |
| CN113836438B (en) | Method, electronic device, and storage medium for post recommendation | |
| CN111262716A (en) | Registration probability estimation method and device and probability estimation model construction method and device | |
| CN116011693A (en) | Path determination method, path determination device, computer equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |