CN115842744A - Node deployment method, device, equipment and storage medium - Google Patents
Node deployment method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN115842744A CN115842744A CN202310140237.9A CN202310140237A CN115842744A CN 115842744 A CN115842744 A CN 115842744A CN 202310140237 A CN202310140237 A CN 202310140237A CN 115842744 A CN115842744 A CN 115842744A
- Authority
- CN
- China
- Prior art keywords
- node
- service function
- physical
- determining
- nodes
- 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.)
- Granted
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Abstract
The embodiment of the disclosure discloses a node deployment method, a node deployment device, a node deployment apparatus and a storage medium, wherein the method comprises the following steps: acquiring a service function node set and a physical node set to be deployed; determining a first weighted graph corresponding to a service function node set based on the execution duration corresponding to each service function node and the transmission efficiency among the service function nodes; determining a second weighted graph corresponding to the physical node set based on the residual resource amount corresponding to each physical node and the residual bandwidth between the physical nodes; determining a mapping relation between each service function node and each physical node based on the first weighted graph and the second weighted graph; and the mapping relation is used for deploying each service function node to the corresponding physical node. The embodiment of the disclosure can improve the deployment efficiency and the deployment precision of the service function node band.
Description
Technical Field
The present disclosure relates to, but not limited to, the field of computer technologies, and in particular, to a node deployment method, apparatus, device, and storage medium.
Background
Network Function Virtualization (NFV) technology realizes that a Network Function is converted from dedicated hardware into a software middlebox, and Network services or application programs are deployed in the form of a Virtual Network Function (VNF), thereby realizing flexible and scalable deployment and management. The Service Function Chain (SFC) is composed of a series of ordered virtual network functions, and the network Service quality depends on the deployment effect of the Service function chain. Therefore, the network service provider needs to place dynamically arriving SFC requests in real time in the physical network under a variety of resource constraints. In the related art, although a manual experience method may be adopted to deploy a VNF to a corresponding physical node in a physical network, a large amount of manpower is required, and the deployment efficiency is low.
Disclosure of Invention
In view of this, the embodiments of the present disclosure at least provide a node deployment method, apparatus, device and storage medium.
The technical scheme of the embodiment of the disclosure is realized as follows:
in one aspect, an embodiment of the present disclosure provides a node deployment method, including: acquiring a service function node set and a physical node set to be deployed; the service function node set carries execution time corresponding to each service function node and transmission efficiency among the service function nodes; the physical node set carries the residual resource amount corresponding to each physical node and the residual bandwidth between the physical nodes; determining a first weighted graph corresponding to the service function node set based on the execution duration corresponding to each service function node and the transmission efficiency among the service function nodes; determining a second weighted graph corresponding to the physical node set based on the residual resource amount corresponding to each physical node and the residual bandwidth between the physical nodes; determining a mapping relation between each service function node and each physical node based on the first weighted graph and the second weighted graph; and the mapping relation is used for deploying each service function node to a corresponding physical node.
In another aspect, an embodiment of the present disclosure provides a node deployment apparatus, including: the system comprises an acquisition module, a configuration module and a configuration module, wherein the acquisition module is used for acquiring a service function node set and a physical node set to be deployed; the service function node set carries execution time corresponding to each service function node and transmission efficiency among the service function nodes; the physical node set carries the residual resource amount corresponding to each physical node and the residual bandwidth between the physical nodes; a first determining module, configured to determine, based on an execution duration corresponding to each service function node and transmission efficiency between the service function nodes, a first weighted graph corresponding to the service function node set; a second determining module, configured to determine, based on a remaining resource amount corresponding to each physical node and a remaining bandwidth between the physical nodes, a second weighted graph corresponding to the set of physical nodes; a third determining module, configured to determine, based on the first weighted graph and the second weighted graph, a mapping relationship between each service function node and each physical node; and the mapping relation is used for deploying each service function node to a corresponding physical node.
In still another aspect, the present disclosure provides a computer device, which includes a memory and a processor, where the memory stores a computer program that is executable on the processor, and the processor executes the computer program to implement some or all of the steps in the method.
In yet another aspect, the disclosed embodiments provide a computer-readable storage medium, on which a computer program is stored, which, when executed by a processor, implements some or all of the steps of the above-described method.
In yet another aspect, the disclosed embodiments provide a computer program comprising computer readable code which, when run in a computer device, a processor in the computer device executes some or all of the steps for implementing the above method.
In yet another aspect, the disclosed embodiments provide a computer program product comprising a non-transitory computer readable storage medium storing a computer program, which when read and executed by a computer, implements some or all of the steps of the above method.
In the embodiment of the disclosure, firstly, a service function node set and a physical node set to be deployed are obtained; the service function node set carries execution duration corresponding to each service function node and transmission efficiency among the service function nodes; the physical node set carries the remaining resource amount corresponding to each physical node and the remaining bandwidth between each physical node. Then, based on the execution duration corresponding to each service function node and the transmission efficiency among the service function nodes, determining a first weighted graph corresponding to a service function node set; and determining a second weighted graph corresponding to the physical node set based on the residual resource amount corresponding to each physical node and the residual bandwidth between the physical nodes. Finally, determining the mapping relation between each service function node and each physical node based on the first weighted graph and the second weighted graph; and the mapping relation is used for deploying each service function node to the corresponding physical node. Therefore, the local and global dependency relationship between the service function nodes and the physical nodes can be determined simultaneously through the weighted graph formed by the service function nodes and the weighted graph formed by the physical nodes, the mapping relationship between each service function node and each physical node can be determined rapidly and accurately, the deployment efficiency and the deployment precision can be improved, and the large-scale and cross-domain SFC deployment and the like can be realized.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the technical aspects of the disclosure.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the disclosure.
Fig. 1 is a schematic implementation flow diagram of a first node deployment method provided in the embodiment of the present disclosure;
fig. 2 is a schematic flow chart of an implementation of a second node deployment method according to an embodiment of the present disclosure;
fig. 3 is a schematic flow chart of an implementation of a third node deployment method according to the embodiment of the present disclosure;
fig. 4 is a schematic flow chart illustrating an implementation of a fourth node deployment method according to an embodiment of the present disclosure;
fig. 5 is a schematic structural diagram of a service function chain according to an embodiment of the present disclosure;
FIG. 6 is a schematic diagram illustrating a service function chain fusion provided by an embodiment of the present disclosure;
fig. 7 is a schematic structural diagram of a physical link according to an embodiment of the present disclosure;
FIG. 8 is a diagram illustrating a mapping relationship according to an embodiment of the present disclosure;
fig. 9 is a schematic structural diagram of a node deployment apparatus according to an embodiment of the present disclosure;
fig. 10 is a hardware entity diagram of a computer device according to an embodiment of the present disclosure.
Detailed Description
To make the objects, technical solutions and advantages of the present disclosure clearer, the technical solutions of the present disclosure are further elaborated with reference to the drawings and the following embodiments, which should not be construed as limiting the present disclosure, and all other embodiments obtained by a person of ordinary skill in the art without making creative efforts fall within the protection scope of the present disclosure.
In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict. Reference to the terms "first/second/third" merely distinguishes similar objects and does not denote a particular ordering with respect to the objects, it being understood that "first/second/third" may, where permissible, be interchanged in a particular order or sequence so that embodiments of the disclosure described herein can be practiced in other than the order shown or described herein.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. The terminology used herein is for the purpose of describing the disclosure only and is not intended to be limiting of the disclosure.
Embodiments of the present disclosure provide a node deployment method, which may be performed by a processor of a computer device. The computer device may be a server, a cloud platform, a notebook computer, a tablet computer, a desktop computer, a mobile device (e.g., a mobile phone, a portable video player, a personal digital assistant, a dedicated messaging device), and other devices with a node deployment capability. Fig. 1 is a schematic flow chart of an implementation process of a node deployment method provided in an embodiment of the present disclosure, as shown in fig. 1, the method includes the following steps S101 to S104:
step S101, acquiring a service function node set and a physical node set to be deployed.
Here, the set of service function nodes may include at least two service function nodes (also referred to as virtual network function nodes), and the service function nodes may be understood as Virtual Network Functions (VNFs) in a Service Function Chain (SFC). For example: receiving at least two SFC requests uploaded by a user, wherein the SFC requests carry service chain path information, and the service chain path information represents the number of service function nodes of a service chain and the connection relation between the service function nodes; determining at least two service function chains based on the at least two service chain path information; for example: the first service function chain is from a first service function node to a second service function node to a third service function node; the second service function chain comprises a first service function node, a fourth service function node, a fifth service function node and the like; and determining the first service function node, the second service function node, the third service function node, the fourth service function node and the fifth service function node as a service function node set.
The physical node set may include at least two physical nodes, and a physical node may be understood as a service device (e.g., a server, a base station, etc.) in a physical link on an underlying physical network, where the service device has resources such as computation, storage, and network, and is used to provide resources to a service function node in a service function chain, so that the service function node completes corresponding service processing. For example: all service devices in a preset area can be obtained in advance, and the service devices with network connection are determined as a physical node set, wherein the current remaining resources in all the service devices are larger than a preset resource threshold value.
The service function node set carries the execution time length corresponding to each service function node and the transmission efficiency among the service function nodes. Different service function nodes may have different functions, resulting in different service function nodes having different execution durations; the execution duration may refer to a duration for processing a packet or traffic data passing through the service function node, for example, the first service function node has a firewall function and the execution duration is 0.3 second; the second service function node has a load balancing function and the execution time is 0.1 second; the third service function node has the function of an intrusion prevention system, the execution time is 0.2 second and the like. The execution duration of the service function node may be set in a user-defined manner according to factors such as the function of the service function node, and is not limited here. Due to the difference of bandwidth and network protocol between service function nodes in a service function chain, the transmission efficiency between the service function nodes can be different; the transmission efficiency may refer to a transmission size of a data amount per unit time, for example, the transmission efficiency between the first service function node and the second service function node is 5 megabytes per second, the transmission efficiency between the second service function node and the third service function node is 3 megabytes per second, and the like. The transmission efficiency between the service function nodes can be set by self according to factors such as a service chain for performing service processing by the service function chain, and the like, which is not limited herein.
The physical node set carries the remaining resource amount corresponding to each physical node and the remaining bandwidth between the physical nodes. Different physical nodes can have different types of resources (such as computing resources, storage resources, and the like), the resource parameters of the physical nodes can include a total resource amount, a used resource amount, a remaining resource amount, and the like, and the sound resource amount of the physical node can be obtained by subtracting the total resource amount and the used resource amount. The remaining bandwidth between physical nodes may also be different due to factors such as the distance between physical nodes in a physical link, the protocol, etc. For example: the total bandwidth and the used bandwidth between the physical nodes can be determined, and the subtraction processing is performed on the total bandwidth and the used bandwidth, so that the residual bandwidth between the first physical node and the second physical node is 10 million per second, the residual bandwidth between the second physical node and the third physical node is 8 million per second, and the like. In some embodiments, one physical node may carry one service function node, or may carry multiple service function nodes, and the functions of the multiple service function nodes may be the same or different, which is not limited herein.
Step S102, determining a first weighted graph corresponding to the service function node set based on the execution duration corresponding to each service function node and the transmission efficiency between the service function nodes.
Here, the first weighted graph may represent a topology structure diagram of attributes such as an association relationship between service function nodes, a weight between service function nodes, and a weight of each service function node, where each service function node has a first weight, and an edge between service function nodes has a second weight. For example: the execution duration corresponding to each service function node may be determined as a first weight, and the transmission efficiency between each service function node may be determined as a second weight. For example: the execution duration of the first service function node is 1 second, the execution duration of the second service function node is 2 seconds, and the transmission efficiency between the first service function node and the second service function node is 5 mega per second, so that it can be determined that the first weight corresponding to the first service function node is 1, the first weight corresponding to the second service function node is 2, and the second weight corresponding to the edge between the first service function node and the second service function node is 5.
Step S103, determining a second weighted graph corresponding to the physical node set based on the remaining resource amount corresponding to each physical node and the remaining bandwidth between the physical nodes.
Here, the second weighted graph may represent a topology structure diagram of attributes such as an association relationship between physical nodes, a weight between physical nodes, and a weight of each physical node, where each physical node has a third weight, and an edge between physical nodes has a fourth weight. For example: the remaining resource amount corresponding to each physical node may be determined as a third weight, and the remaining bandwidth between the physical nodes may be determined as a fourth weight, and so on. For example: if the remaining resource amount of the first physical node is 3 million, the remaining resource amount of the second physical node is 8 million, and the remaining bandwidth between the first physical node and the second physical node is 7 million per second, it may be determined that the third weight corresponding to the first physical node is 3, the third weight corresponding to the second physical node is 8, and the fourth weight corresponding to the edge between the first physical node and the second physical node is 7.
And step S104, determining the mapping relation between each service function node and each physical node based on the first weighted graph and the second weighted graph.
Here, the mapping relationship is used to deploy each service function node to the corresponding physical node, for example, the mapping relationship represents that the first service function node is deployed to the first physical node, the second service function node is deployed to the third physical node, the third service function node is deployed to the second physical node, the fourth service function node is deployed to the third physical node, and so on. Step S104 includes: the trained neural network model can be used for processing the first weighted graph and the second weighted graph to obtain the confidence coefficient of each service function node deployed to each physical node; the trained neural network model may refer to a preset machine learning model, such as a neural network model for performing confidence prediction; based on all confidence levels, a mapping relationship is determined. For example: when the confidence coefficient of the first service function node to the first physical node is 0.9, the confidence coefficient of the first service function node to the second physical node is 0.5, and the confidence coefficient of the first service function node to the third physical node is 0.4, determining that the first service function node can be deployed to the first physical node; for the mapping relationship of other service function nodes, the corresponding physical nodes can be sequentially determined in the same manner.
In the embodiment of the disclosure, firstly, a service function node set and a physical node set to be deployed are obtained; the service function node set carries execution duration corresponding to each service function node and transmission efficiency among the service function nodes; the physical node set carries the remaining resource amount corresponding to each physical node and the remaining bandwidth between each physical node. Then, determining a first weighted graph corresponding to the service function node set based on the execution duration corresponding to each service function node and the transmission efficiency among the service function nodes; and determining a second weighted graph corresponding to the physical node set based on the residual resource amount corresponding to each physical node and the residual bandwidth between the physical nodes. Finally, determining the mapping relation between each service function node and each physical node based on the first weighted graph and the second weighted graph; and the mapping relation is used for deploying each service function node to the corresponding physical node. Therefore, the local and global dependency relationship between the service function nodes and the physical nodes can be determined simultaneously through the weighted graph formed by the service function nodes and the weighted graph formed by the physical nodes, the mapping relationship between each service function node and each physical node can be determined rapidly and accurately, the deployment efficiency and the deployment precision can be improved, and the large-scale and cross-domain SFC deployment and the like can be realized.
The embodiment of the present disclosure provides a node deployment method, where the service function node set guarantees at least two service function chains. As shown in fig. 2, the method includes steps S201 to S206 as follows:
step S201, a service function node set and a physical node set to be deployed are obtained.
Step S205, determining a second weighted graph corresponding to the physical node set based on the remaining resource amount corresponding to each physical node and the remaining bandwidth between the physical nodes.
Step S206, determining a mapping relationship between each service function node and each physical node based on the first weighted graph and the second weighted graph.
Step S201 corresponds to step S101, and reference may be made to the specific implementation of step S101 in implementation; steps S205 to S206 correspond to steps S103 to S104, respectively, and specific embodiments of steps S103 to S104 can be referred to in the implementation.
Step S202, determining the execution function of the service function node in each service function chain.
Here, the execution function may refer to a type of processing traffic performed by the service function node, such as a storage function, a calculation function, and the like, a filtering function, and the like. The service type processed by the service function chain and the information such as the execution function of each service function node can be added in advance in the SFC request, and after the SFC request is received, the SFC request is analyzed to obtain the execution function of each service function node. For example: the first service function chain comprises a first service function node, a second service function node and a third service function node, and the corresponding execution functions are a first execution function, a second execution function and a third execution function respectively; the second service function chain comprises a fourth service function node, a fifth service function node and a sixth service function node, and the corresponding execution functions are a fourth execution function, a second execution function, a fifth execution function and the like.
Step S203, fusing all the service function chains based on the execution function and the execution duration corresponding to each service function node, to obtain the fused execution duration of each service function node.
Here, a plurality of service function chains may be fused to obtain a fused service function chain, which is helpful for reducing the number of service function chains and improving the relevance between service function nodes. For example: the second service function node in the first service function chain is a second execution function, the fifth service function node in the second service function chain is also a second execution function, the execution time of the second service function node is 1 second, and the execution time of the fifth service function node is 2 seconds, so that the first service function chain and the second service function chain can be fused, and the execution time of the fused second service function node (or fifth service function node) is 3 seconds.
Step S204, determining a first weighted graph corresponding to the service function node set based on the execution duration of each service function node after fusion and the transmission efficiency between the service function nodes.
For example: the execution function of the second service function node in the first service function chain is the same as that of the fifth service function node in the second service function chain, the first weight of the second service function node is 1, the first weight of the fifth service function node is 2, then the first weight of the fused second service function node (or fifth service function node) in the first weighted graph is 3, and the second weight between the service function nodes is unchanged.
In the embodiment of the present disclosure, by fusing service function nodes having the same execution function, nodes in the first weighted graph can be reduced, the relevance between nodes in the first weighted graph is improved, and then the mapping relationship can be obtained more accurately in the subsequent step.
The embodiment of the present disclosure provides a node deployment method, as shown in fig. 3, the method includes the following steps S301 to S307:
step S301, a service function node set and a physical node set to be deployed are obtained.
Step S302, determining a first weighted graph corresponding to the service function node set based on the execution duration corresponding to each service function node and the transmission efficiency between the service function nodes.
Step S303, determining a second weighted graph corresponding to the physical node set based on the remaining resource amount corresponding to each physical node and the remaining bandwidth between the physical nodes.
The steps S301 to S303 correspond to the steps S101 to S103, respectively, and the specific implementation of the steps S101 to S103 can be referred to.
Step S304, determining a first adjacency matrix corresponding to the first weighted graph, and determining a second adjacency matrix corresponding to the second weighted graph.
Here, the adjacency matrix may refer to a two-dimensional array for storing data of inter-node relationships (edges or arcs), the first adjacency matrix being used to represent the inter-node relationships of the first weighted graph, and the second adjacency matrix being used to represent the inter-node relationships of the second weighted graph. The values of the elements in the first adjacency matrix corresponding to the first service function node and the second service function node may be determined by a first weight of the first service function node, a first weight of the second service function node, and a second weight between the first service function node and the second service function node, for example, the first weight of the first service function node, the first weight of the second service function node, and the second weight between the first service function node and the second service function node are added to obtain the values of the elements in the first adjacency matrix corresponding to the first service function node and the second service function node. The element values of the elements in the second adjacent matrix corresponding to the first physical node and the second physical node may be determined by the third weight of the first physical node, the third weight of the second physical node, and the fourth weight between the first physical node and the second physical node, for example, the third weight of the first physical node, the third weight of the second physical node, and the fourth weight between the first physical node and the second physical node are added to obtain the element values of the elements in the second adjacent matrix corresponding to the first physical node and the second physical node.
Step S305, determining a vector representation of each service function node based on the first adjacency matrix.
Here, the vector representation of the service function node may be used to represent the features of the service function node, and the trained first vector representation model may be used to process the first adjacency matrix to obtain the vector representation of each service function node; the trained first vector characterization model may refer to a pre-configured machine learning model, such as a neural network model for performing vector characterization learning. For example: the first adjacency matrix may be input to a trained first vector characterization model (e.g., an Embedding network), resulting in a vector characterization for each service function node.
Step S306, determining a vector characterization of each physical node based on the second adjacency matrix.
Here, the vector representation of the physical node may be used to represent the features of the physical node, and the second adjacency matrix may be processed by using a trained second vector representation model to obtain a vector representation of each physical node; the trained second vector characterization model may refer to a pre-configured machine learning model, such as a neural network model for performing vector characterization learning. For example: the second adjacency matrix may be input to a trained second vector characterization model (e.g., an Embedding network) to obtain a vector characterization for each physical node.
Step S307, determining a mapping relationship between each service function node and each physical node based on the vector characterization of each service function node and the vector characterization of each physical node.
Here, the trained mapping model may be used to process the vector representation of the service function node and the vector representation of the physical node to obtain a mapping relationship between each service function node and each physical node; the trained mapping model may refer to a pre-set machine learning model, such as a neural network model for performing node mapping. For example: the vector representation of each service function node and the vector representation of each physical node may be input to the trained mapping model, so as to obtain that the first service function node is deployed to the second physical node, the second service function node is deployed to the third physical node, the third service function node is deployed to the first physical node, and the like.
In the embodiment of the disclosure, the first adjacency matrix and the second adjacency matrix can be accurately determined through the first weighted graph and the second weighted graph, and then the mapping relationship between each service function node and each physical node is accurately determined through the first adjacency matrix and the second adjacency matrix. Compared with the related art, the mapping relation is directly determined based on the vector of each service function node in the service function chain, so that the incidence relation between service function nodes and between physical nodes can be increased, and the deployment efficiency and accuracy are improved.
In some embodiments, the step S304 may include the following steps S3041 to S3042:
step S3041, determining the transmission delay between the service function nodes as the element value of the non-diagonal element in the first adjacent matrix, and determining the execution duration corresponding to each service function node as the element value of the diagonal element in the first adjacent matrix.
Here, the transmission delay is determined based on the transmission efficiency and a preset amount of data to be transmitted, for example, the preset amount of data to be transmitted is divided by the transmission efficiency to obtain the transmission delay. If the number of the service function nodes in the first weighted graph is m (m is a positive integer), the dimension of the first adjacent matrix is m x m, and the elements in the first row represent the association relationship between the first service function node and each of the first service function node, the second service function node, the third service function node and the like. The first service function node and the element corresponding to the first service function node are diagonal elements; the elements corresponding to the first service function node, the second service function node, the third service function node and the like are non-diagonal elements; the element value of the diagonal element corresponding to the first service function node and the first service function node may be an execution duration (i.e., a first weight) corresponding to the first service function node; the element value of the non-diagonal element corresponding to the first service function node and the second service function node may be a transmission delay (i.e., a second weight) between the first service function node and the second service function node.
Step S3042, determining the remaining bandwidth between the physical nodes as the element value of the non-diagonal element in the second adjacent matrix, and determining the remaining resource amount corresponding to each physical node as the element value of the diagonal element in the second adjacent matrix.
Here, if the number of service function nodes in the second weighted graph is n (n is a positive integer), the dimension of the second adjacency matrix is n × n, and the elements in the first row represent the association relationship between the first physical node and the first physical node, the second physical node, the third physical node, and the like. The elements corresponding to the first physical node and the first physical node are diagonal elements; the elements corresponding to the first physical node, the second physical node, the third physical node and the like are non-diagonal elements; the element value of the diagonal element corresponding to the first physical node and the first physical node may be the remaining resource amount (i.e., the third weight) corresponding to the first physical node; the element value of the non-diagonal element corresponding to the first physical node and the second physical node may be a remaining bandwidth (i.e., a fourth weight) between the first physical node and the second physical node.
Compared with the related art, the element values of the diagonal elements in the adjacency matrix are directly set to zero; in the embodiment of the disclosure, the execution duration corresponding to each service function node is determined as the element value of the diagonal element in the first adjacent matrix, and the remaining resource amount corresponding to each physical node is determined as the element value of the diagonal element in the second adjacent matrix; therefore, the overall relevance of the service function nodes and the overall relevance of the physical nodes can be increased, and the efficiency and the accuracy of deployment are improved.
In some embodiments, the step S305 may include the following steps S3051 to S3056:
step S3051, performing normalization processing on the element value of each element in the first adjacency matrix to obtain a normalized first adjacency matrix.
Here, the element value of each element in the first adjacency matrix may be linearly normalized, for example, a sum of elements of all elements in the first adjacency matrix is determined, and the element value of each element is divided by the sum of elements to obtain a normalized element value of each element.
In some embodiments, the normalization process may be performed using the following equation:
in the formula (1), the first and second groups,
indicates the fifth->
Admission, based on the number of service function nodes>
Representing a service function node->
And &>
Based on the transfer delay before normalization, and->
Representing a service function node->
And &>
The normalized transmission delay between the first and second nodes,
representing a service function node->
The set of in-degree nodes. The second neighbor matrix corresponding to the physical node set may also be normalized by the same method, which is not limited herein.
And S3052, extracting the characteristics of the normalized first adjacency matrix by using a preset long-short term memory network to obtain the initial characterization of each service function node.
Here, a Long Short-Term Memory Network (LSTM) is a kind of temporal Recurrent Neural Network (RNN) with Long-Term Memory capability, and its Network structure contains one or more unit components with forgetting and memorizing functions, and can be used in data and scenes with time-series characteristics. The initial characterization may be a characterization obtained using a long-short term memory network, and the dimension of the service function node characterized by the initial characterization may be lower than the dimension of the service function node characterized by the vector characterization. For example: and inputting the normalized first adjacency matrix into the trained long-term and short-term memory network to obtain the initial representation of each service function node.
And S3053, determining an in-degree node set and an out-degree node set corresponding to each service function node.
Here, the in-degree node of the current service function node may refer to a service function node whose quantity traffic flows to the current service function node, and the out-degree node of the current service function node may refer to a service function node that receives the quantity traffic of the current service function node. An in-degree node set and an out-degree node set corresponding to each service function node can be determined according to the flow direction of data traffic in a preset service function chain. For example: the in-degree node set of the third service function node comprises a first service function node and a second service function node, and the out-degree node set of the third service function node comprises a fourth service function node, a fifth service function node and the like.
And S3054, performing aggregation processing on the initial representations of the entry nodes in the entry node set to obtain the entry vector representations of the corresponding service function nodes.
Here, the representation of the admission vector of the service function node may refer to an overall representation corresponding to all admission nodes of the service function node, and the initial representations of the admission nodes in the admission node set of the current service function node may be aggregated by using a preset first convolution network or a first aggregation function (e.g., a GraphSAGE function), so as to obtain the representation of the admission vector of the current service function node.
And S3055, performing aggregation processing on the initial representations of the out-degree nodes in the out-degree node set to obtain out-degree vector representations of the corresponding service function nodes.
Here, the expression of the degree vector of the service function node may refer to an overall expression corresponding to all the degree nodes of the service function node, and the initial expression of the degree node in the set of degree nodes of the current service function node may be aggregated by using a preset second convolution network or a second aggregation function, so as to obtain the expression of the degree vector of the current service function node.
In some embodiments, the following formula may be used to determine the in-degree vector characterization:
in the formula (2), the first and second groups,
indicates iteration pick>
The next fifth->
The in-degree vector representation of each service function node,
represents a set of entry nodes, and>
representing service functionsNode->
And &>
The transmission time delay before the normalization is carried out,
indicates iteration pick>
A next entry vector characterization, based on the value of the entry vector>
Indicates the fifth->
An initial characterization of each of the service function nodes,
and &>
Represents a preset confidence level (hyper-parameter) of the service function node, and->
Is a parameter matrix, is->
Represents a bias vector, <' > based on the status of the bias>
Represents a Relu activation function, < >>
，/>
The iteration times of the preset convolution aggregation operation can be set according to actual conditions. For the expression of the output degree vector of the service function node, the expression can also be based on the output degree node set
The determination is performed in the same manner, and is not limited herein.
Step S3056, determining the vector representation of the corresponding service function node based on the in-degree vector representation and the out-degree vector representation of each service function node.
Here, the preset third convolution network or third aggregation function may be used to perform aggregation processing on the in-degree vector representation and the out-degree vector representation of the current service function node, so as to obtain the vector representation of the current service function node.
In some embodiments, the vector characterization of the serving function node may be determined using the following formula:
in the formula (3), the first and second groups of the compound,
represents an entry node characterization, and>
represents a degree node characterization, and>
for a preset model parameter, is>
Is a Sigmiod function.
In the embodiment of the disclosure, the vector representation of the service function node can be determined quickly and accurately by respectively determining the in-degree representation and the out-degree representation of the service function node and then based on the in-degree representation and the out-degree representation of the service function node.
In some embodiments, the step S306 may include the following steps S3061 to S3064:
step S3061, perform normalization processing on the element value of each element in the second adjacency matrix to obtain a normalized second adjacency matrix.
Here, the element value of each element in the second adjacency matrix may be linearly normalized, for example, a sum of elements of all elements in the second adjacency matrix is determined, and the element value of each element is divided by the sum of elements to obtain a normalized element value of each element.
Step S3062, feature extraction is performed on the normalized second adjacency matrix by using a preset long-short term memory network, so as to obtain an initial representation of each physical node.
Here, for example: and inputting the normalized second adjacency matrix into the trained long-term and short-term memory network to obtain the initial representation of each physical node. The long-short term memory network for extracting features of the normalized first adjacency matrix may be the same as or different from the long-short term memory network for extracting features of the normalized second adjacency matrix, and is not limited herein.
Step S3063, determine the neighboring node set corresponding to each physical node.
Here, the neighboring node corresponding to the current physical node may refer to a physical node that establishes a network connection with the current physical node, and includes an in-degree node and an out-degree node of the current physical node. The method includes the steps that a network connection adjacent node set corresponding to a physical node is determined from a physical node set based on a preset network connection relation of each physical node.
Step S3064, performing iterative aggregation processing on the initial token of each physical node and the initial tokens of the neighboring nodes in the corresponding neighboring node set to obtain the vector token of the corresponding physical node.
Here, the preset fourth convolution network or a fourth aggregation function may be used to perform aggregation processing on the initial representations of all neighboring nodes of the current physical node, so as to obtain a vector representation of the current physical node.
In some embodiments, the vector characterization of the current physical node may be determined using the following formula:
in the formula (4), the first and second groups,
indicates iteration pick>
The next fifth->
Vector characterization for a physical node, based on the comparison of the values of the parameters>
Represents a set of adjacent nodes, in combination>
Representing a service function node->
And adjacent node->
Based on the remaining bandwidth before normalization, is selected>
Indicates iteration pick>
The next vector characterization, based on the comparison>
Represents a fifth or fifth party>
Initial characterization of a service function node, based on the comparison of the value of the parameter>
And &>
Represents a preset confidence level (hyper-parameter) of the service function node, and->
Is a parameter matrix, is->
A vector of the offset is represented, and,
represents a Relu activation function, < >>
，/>
The iteration times of the preset convolution aggregation operation can be set according to actual conditions.
In the embodiment of the disclosure, the vector representation of the physical node can be determined quickly and accurately by determining the connected nodes of the physical node and then based on the initial representation of the connected nodes.
In some embodiments, the step S307 may include the following steps S3071 to S3074:
step S3071, extracting the characteristics of the vector characterization of the service function node to obtain the high-order characterization of each service function node.
Here, the higher-order token may be a token obtained by performing feature extraction on the vector token again, and the dimension of the service function node represented by the higher-order token may be higher than the dimension of the service function node represented by the vector token. The trained first feature extraction network can be used for carrying out feature extraction on the vector representation of the service function nodes to obtain high-order representation of each service function node; the trained first feature extraction network may refer to a preset machine learning model, such as a neural network model for performing feature extraction. For example: and inputting the vector representations of all the service function nodes into the trained first feature extraction network, wherein the trained first feature extraction network can obtain the high-order representation of each service function node based on the vector representation of at least one service function node.
And S3072, performing feature extraction on the vector characterization of the physical nodes to obtain a high-order characterization of each physical node.
The dimension of the physical node represented by the high-order representation can be higher than the dimension of the physical node represented by the vector representation, and the vector representation of the physical node can be subjected to feature extraction by utilizing a trained second feature extraction network to obtain the high-order representation of each physical node; the trained first feature extraction network may refer to a preset machine learning model, such as a neural network model for performing feature extraction. For example: and inputting the vector representations of all the physical nodes into the trained second feature extraction network, wherein the trained second feature extraction network can obtain the high-order representation of each physical node based on the vector representation of at least one physical node.
Step S3073, determining a mapping probability of each service function node deploying a corresponding physical node based on the high-order characterization of each service function node and the high-order characterization of each physical node.
Here, the mapping probability may refer to a possible degree of deployment of the current service function node to the current physical node, and the high-order representation of each service function node and the high-order representation of each physical node may be processed by using a trained probability prediction model to obtain the high-order representation of each physical node; the trained probabilistic predictive model may refer to a preset machine learning model, such as a neural network model for performing probabilistic prediction. For example: inputting the high-order representation of the first service function node and the high-order representation of the first physical node into the trained probability prediction model to obtain that the preset probability of the first service function node being deployed to the first physical node is 0.9; and inputting the high-order representation of the first service function node and the high-order representation of the second physical node into the trained probability prediction model to obtain that the preset probability of the first service function node being deployed to the second physical node is 0.6 and the like.
Step S3074, determining a mapping relationship between each service function node and each physical node based on all the mapping probabilities.
For example: determining that the preset probability of the first service function node being deployed to the first physical node is 0.9, the preset probability of the first service function node being deployed to the second physical node is 0.6, and the preset probability of the first service function node being deployed to the third physical node is 0.5; the method comprises the steps that since the preset probability that a first service function node is deployed to a first physical node is the highest among the preset probabilities between the first service function node and all physical nodes, it is determined that the first service function node is deployed to the first physical node; then determining that the preset probability that the second service function node is deployed to the third physical node is the highest, and determining that the second service function node is deployed to the third physical node; and determining that the preset probability of the third service function node to be deployed to the second physical node is the highest, and determining that the third service function node is deployed to the second physical node.
In the embodiment of the disclosure, the high-order representation of the service function node is obtained by performing feature extraction on the vector representation of the service function node again, and the high-order representation of the physical node is obtained by performing feature extraction on the vector representation of the physical node; therefore, the representation of the high-order representation of the service function node on the whole result of all the service function nodes can be improved, and the representation of the high-order representation of the physical node on the whole result of all the physical nodes can be improved, so that the mapping probability of the physical nodes corresponding to the service function node deployment can be quickly and accurately obtained, and the mapping relation between each service function node and each physical node can be quickly and accurately determined.
In some embodiments, the step S3071 may include the following steps S311 to S313:
step S311, determining a first meta-path between each service function node and a corresponding adjacent node.
Here, the first meta path may refer to a possible path between two service function nodes, for example, the first meta path from the first service function node to the second service function node may include: the first service function node can be in network connection with the second service function node through the first intermediate service function node; the first service function node can be in network connection with the second service function node through a second intermediate service function node; the first service function node can be in network connection with the second service function node through the first intermediate service function node and the third intermediate service function node. All first meta-paths between each service function node and the corresponding neighboring nodes may be determined using a recursive algorithm or a heterogeneous network.
Step S312, performing linear transformation on the vector representation of the service function node corresponding to each first meta path to obtain the semantic representation of the corresponding service function node.
Here, the semantic representation of the service function node may refer to a representation of the service function node with respect to the first meta-path. For example: the first meta path of the first service function node includes the first service function node, the third service function node, and the second service function node, and then the vector representation of the first service function node, the vector representation of the third service function node, and the vector representation of the second service function node may be weighted to obtain the semantic representation of the first service function node.
Step 313, normalizing the semantic representation of each service function node to obtain a high-order representation of the corresponding service function node.
Here, the semantic representations of all first unary paths of the current service function node may be added to obtain a token sum; dividing the semantic representation of each first unitary path with the representation sum to obtain normalized semantic representations; multiplying the normalized semantic representation with the semantic representation before normalization to obtain a representation product; and adding all the feature products to obtain the high-order feature of the current service function node.
In some embodiments, first, a meta-path and a set of neighbor nodes may be constructed for each serving function node or physical node: finding a meta-path of length T for each node, the node being
Is used for the meta path of>
Indicates that, then, the node->
Neighbor node set based on meta-path to be @>
. Then, the slave meta-path neighbor node set +>
And (3) middle sampling: setting the sampling number as S, if the vertex neighbor number is less than S, adopting a sampling method with replacement until S vertexes are sampled; if the number of the vertex neighbors is larger than S, sampling without putting back is adopted, namely a resampling method or a negative sampling method with putting back is adopted to reach S. Second, the meta-path node characterizes the aggregate->
: and obtaining the representation of the node on each element path by adopting an average pooling aggregation method. It is contemplated that each node in the network contains many types of semantic information, i.e., each meta-path may correspond to a semantic information. And finally, fusing semantic information: an attention mechanism is designed to learn the importance of different meta-paths and fuse them. The input of the method is that the node generates P group node embedding (semantic representation corresponding to the meta-path) under each meta-path to represent->
Make->
The learned weight for each meta path can be calculated using the following formula:
in the formula (5), the first and second groups of the chemical reaction materials are selected from the group consisting of,
represents a neural network for learning the weight, and>
indicating an output, i.e. a high-order representation>
A vector representation representing the input, i.e. the service function node.
In some embodiments, in order to learn the importance of each meta-path, a linear transformation is used to transform node embedding under a specific semantic, a layer structure multi-layer Perceptron (MLP) may be utilized, and the following formula is used to determine the semantic representation of the service function node:
in the formula (6), the first and second groups,
semantic representation of service function nodes or meta-paths>
Represents a semantic representation corresponding to the set of service function nodes, based on the value of the semantic representation>
Represents a predetermined parameter matrix, based on the evaluation of the status of the evaluation unit>
And &>
Represents a predetermined parameter, is present>
And representing the semantic representation corresponding to the meta path.
In some embodiments, the importance of each unary path may be determined using the following formula:
in the formula (7), the first and second groups,
represents the degree of significance of the meta path>
The semantic representation to which the meta-path corresponds,
indicating the number of meta-paths.
In some embodiments, the semantic representations may be normalized using the following formula:
in the formula (8), the first and second groups,
high-level characterization, representing a service function node>
Represents the degree of significance of the meta path>
And representing semantic representations corresponding to the meta-paths. For the high-order characterization of the physical node, the calculation may also be performed in the same manner, which is not limited herein.
In the embodiment of the disclosure, by determining the semantic representation of the service function node for each first meta path, the high-order representation of the service function node can be determined quickly and accurately.
In some embodiments, the step S3072 may include the following steps S321 to S323:
step S321, determining a second meta-path between each of the physical nodes and the corresponding neighboring node.
Here, the second meta path may refer to a possible path between two physical nodes, for example, the second meta path between the first physical node to the second physical node may include: the first physical node can be in network connection with the second physical node through the first intermediate physical node; the first physical node can be in network connection with the second physical node through the second intermediate physical node; the first physical node may be in network connection with the second physical node through the first intermediate physical node, the third intermediate physical node, and so on. All second meta-paths between each physical node and the corresponding neighboring node may be determined using a recursive algorithm.
Step S322, performing linear transformation on the vector representation of the physical node corresponding to each second element path to obtain the semantic representation of the corresponding physical node.
Here, the semantic representation of the physical node may refer to a representation of the physical node for the second meta-path. For example: the second meta-path of the first physical node includes the first physical node, the third physical node, and the second physical node, and then the vector representation of the first physical node, the vector representation of the third physical node, and the vector representation of the second physical node may be weighted to obtain the semantic representation of the first physical node.
Step S323, normalization processing is carried out on the semantic representation of each physical node, and high-order representation of the corresponding physical node is obtained.
Here, the semantic representations of all second meta paths of the current physical node may be added to obtain a feature sum; dividing the semantic representation of each second element path with the representation sum to obtain normalized semantic representations; multiplying the normalized semantic representation with the semantic representation before normalization to obtain a representation product; and adding all the feature products to obtain the high-order feature of the current physical node.
In the embodiment of the disclosure, by determining the semantic representation of the physical node for each second element path, the high-order representation of the physical node can be determined quickly and accurately.
In some embodiments, the mapping probability comprises a first probability and a second probability; the step S3073 may include the following steps S331 to S335:
step S331, determining a current service function node from the service function node set, and determining at least two candidate physical nodes matching the previous service function node from the physical node set.
Here, the candidate physical node may refer to a physical node that meets a traffic processing requirement of the current service function node, and for example, if the traffic processing requirement of the current service function node is 10 million per second, the remaining bandwidth of the first physical node is 12 million per second, and the remaining bandwidth of the second physical node is 7 million per second, the first physical node may be determined as a candidate physical node of the current service function node.
Step S332, performing logistic regression on the high-order characterization of the current service function node and the high-order characterization of each candidate physical node to obtain a first probability that the current service function node is deployed to each candidate physical node.
Here, the first probability may be understood as a possible degree of deployment of the current service function node to the candidate physical node. The Logistic regression is also called Logistic regression analysis, is a generalized linear regression analysis model, and belongs to supervised learning in machine learning, for example, logistic regression processing is respectively performed on the high-order characterization of the current service function node and the high-order characterization of the first candidate physical node or the second candidate physical node by using a Sigmoid function, and the first probability that the current service function node is deployed to the first candidate physical node is obtained to be 0.9, the first probability that the current service function node is deployed to the second candidate physical node is obtained to be 0.6, and the like.
In some embodiments, the first probability may be determined using the following equation:
in the formula (9), the first and second groups,
represents a first probability that>
Representing service function nodes>
Is higher order characterized by->
Indicating a physical node pick>
Is higher order characterized by->
Representing a set of service function nodes.
Step S333, determining a current candidate node from at least two candidate physical nodes based on the first probability.
Here, the current candidate node may refer to a physical node that the current service function node determines to deploy. For example: the first probability that the current service function node is deployed to the first candidate physical node is 0.9, and the first probability that the current service function node is deployed to the second candidate physical node is 0.6, and the first candidate physical node may be determined as the current candidate node.
Step S334 is to obtain a next service function node and a next candidate node corresponding to the next service function node.
Here, the next service function node may refer to an out-degree node of the current service function node, for example, the second service function node is connected to the current service function node, and may receive data traffic sent by the current service function node, and may determine the second service function node as the next service function node. The current service function node and the next service function node can form a service function link set
The path length between the current service function node and the next service function node ≥ is>
Greater than 0. The physical node which meets the requirement of the next service function node for service processing can be determined from the physical node set and will be fullAnd determining the physical node meeting the service processing requirement of the next service function node as a next candidate node corresponding to the next service function node. The current physical node and the next physical node may form a physical link set
Remaining bandwidth between the current physical node and the next physical node @>
Greater than a predetermined bandwidth.
Step S335, perform logistic regression on the high-order representation of the current service function node, the high-order representation of the current candidate node, the high-order representation of the next service function node, and the high-order representation of the next candidate node, to obtain a second probability that a link from the current service function node to the next service function node is mapped to a link from the current candidate node to the next candidate node.
For example: splicing the high-order characteristics of the current service function node and the high-order characteristics of the current candidate node (for example, increasing the row number of vectors corresponding to the high-order characteristics) to obtain a first spliced characteristic; splicing the high-order representation of the next service function node and the high-order representation of the next candidate node to obtain a second splicing representation; and performing logistic regression processing on the first splicing representation and the second splicing representation by using a Sigmoid function to obtain a second probability that the link from the current service function node to the next service function node is mapped to the link from the current candidate node to the next candidate node. For example: the next candidate node comprises a second physical node and a third physical node, the second probability that the link from the current service function node to the next service function node is mapped to the link from the current candidate node to the second physical node is 0.8, the second probability that the link from the current service function node to the next service function node is mapped to the link from the current candidate node to the third physical node is 0.6, and the like.
In some embodiments, the second probability may be determined using the following equation:
in the formula (10), the first and second groups of the chemical reaction are shown in the formula,
represents a second probability of being greater than or equal to>
Representing a service function node->
Is higher order characterized by->
Indicating a physical node pick>
Is higher order characterized by->
Representing a service function node->
Is higher order characterized by->
Representing physical nodes
High order characterization of (1).
In some embodiments, the step S3074 may include the following step S336:
step S336, determining a mapping relationship between each service function node and each physical node based on the second probability.
For example: the first candidate physical node is determined as a current candidate node, the second probability that the link from the current service function node to the next service function node is mapped to the link from the current candidate node to the second physical node is 0.8, and the second probability that the link from the current service function node to the next service function node is mapped to the link from the current candidate node to the third physical node is 0.6, then it may be determined that the first service function node may be deployed to the first physical node, the second service function node may be deployed to the second physical node, and so on.
In the embodiment of the disclosure, a current candidate node corresponding to a current service function node is determined according to a first probability that the current service function node is deployed to each candidate physical node; and then determining the next candidate node corresponding to the next service function node through the second probability of mapping the link from the current service function node to the next service function node to the link from the current candidate node to the next candidate node, which is beneficial to rapidly and accurately realizing the deployment between each service function node and each physical node.
An embodiment of the present disclosure provides a node deployment method, as shown in fig. 4, the method includes the following steps S401 to S409:
step S401, a service function node set and a physical node set to be deployed are obtained.
Step S402, determining a first weighted graph corresponding to the service function node set based on the execution duration corresponding to each service function node and the transmission efficiency between the service function nodes.
Step S403, determining a second weighted graph corresponding to the physical node set based on the remaining resource amount corresponding to each physical node and the remaining bandwidth between the physical nodes.
Steps S401 to S403 correspond to steps S101 to S103, respectively, and the detailed implementation of steps S101 to S103 can be referred to.
Step S404, obtaining the trained deployment model.
Here, the deployment model includes at least a first characterization learning network, a second characterization learning network, a higher-order learning network, and a probabilistic prediction network, and different networks may implement different functions, e.g., the first characterization learning network may determine a vector characterization of each service function node, and so on.
Step S405, the first weighted graph is processed by utilizing the first representation learning network, and the vector representation of each service function node is obtained.
Here, the first characterization learning network may refer to a preset, trained machine learning model, such as a neural network model for performing characterization learning. For example: and inputting the first weighted graph into a first representation learning network to obtain the vector representation of each service function node.
And step S406, processing the second weighted graph by using the second characterization learning network to obtain the vector characterization of each physical node.
Here, the second characterization learning network may refer to a preset, trained machine learning model, such as a neural network model for performing characterization learning. For example: and inputting the second weighted graph into a second representation learning network to obtain the vector representation of each physical node.
Step S407, using the high-order learning network to process the vector characterization of each service function node and the vector characterization of each physical node, respectively, so as to obtain the high-order characterization of each service function node and the high-order characterization of each physical node.
Here, the higher-order learning network may refer to a preset, trained machine learning model, such as a neural network model for performing feature extraction. For example: inputting the vector representation of each service function node into a high-order learning network to obtain the high-order representation of each service function node; and inputting the vector representation of each service function node into a high-order learning network to obtain the high-order representation of each physical node.
Step S408, predicting the high-order characteristics of each service function node and the high-order characteristics of each physical node by using the probability prediction network, so as to obtain a mapping probability of each service function node deploying the corresponding physical node.
Here, the mapping probability may include a first probability and a second probability. The probabilistic predictive network may refer to a pre-set, trained machine learning model, such as a neural network model for performing probabilistic predictions. For example: inputting the high-order representation of each service function node and the high-order representation of each physical node into a probability prediction network to obtain a first probability of each service function node being deployed to the corresponding physical node; determining a current candidate node corresponding to the current service function node based on the first probability; and then obtaining a second probability that the link from the current service function node to the next service function node is mapped to the link from the current candidate node to the next candidate node.
In some embodiments, the first probability and the second probability may be used to construct an optimization function; and under the condition that the function value of the optimization function is minimum, updating the parameters of the untrained deployment model to obtain the trained deployment model. For example: acquiring a sample service function chain and a sample mapping relation deployed to corresponding physical nodes, a first sample weighted graph corresponding to the sample service function chain, the sample service function chain and a second sample weighted graph deployed to corresponding at least two physical nodes; inputting the first sample weighted graph and the second sample weighted graph into an untrained deployment model to obtain all current first probabilities and second probabilities; inputting all the first probabilities and the second probabilities to a preset optimization function to obtain optimization function values corresponding to all possible mapping relations; adjusting parameters of the untrained deployment model, so that the optimization function value corresponding to the sample mapping relation is the minimum value of the optimization function values corresponding to all possible mapping relations under the condition of the sample mapping relation; and determining the untrained deployment model after iteration as the trained deployment model under the condition that the preset precision condition is met by the result of preset iteration times.
In some embodiments, the optimization function may be determined using the following formula:
in the formula (11), the first and second groups,
a function value representing an optimization function, is->
Indicating a processing delay of the serving function node at the candidate node, <' > or>
Represents the propagation delay of data traffic on the link, and>
represents a first probability, is>
Representing a second probability.
In some embodiments, the first weighted graph corresponding to the service function set and the second weighted graph corresponding to the physical node set may be input to the trained deployment model, so as to obtain a mapping relationship between each service function node and each physical node.
Step S409, determining a mapping relationship between each service function node and each physical node based on all the mapping probabilities.
For example: determining that the preset probability of the first service function node being deployed to the first physical node is 0.9, the preset probability of the first service function node being deployed to the second physical node is 0.6, and the preset probability of the first service function node being deployed to the third physical node is 0.5; determining that the first service function node is deployed to the first physical node because the preset probability that the first service function node is deployed to the first physical node is the highest in the preset probabilities between the first service function node and all the physical nodes; then determining that the preset probability that the second service function node is deployed to the third physical node is the highest, and determining that the second service function node is deployed to the third physical node; and determining that the preset probability of the third service function node to be deployed to the second physical node is the highest, and determining that the third service function node is deployed to the second physical node.
In the embodiment of the disclosure, the mapping relationship between each service function node and each physical node can be accurately and quickly determined through different networks; meanwhile, in the training process of the deployment model, the deployment model can be trained through a back propagation algorithm by utilizing an optimization function determined based on the first probability and the second probability, so that the training efficiency, the training accuracy and the like can be improved.
The application of the node deployment method provided by the embodiment of the present disclosure in an actual scene is described below, taking cross-domain deployment of at least two service function chains as an example. Most of the service function chain deployment methods in the related art are deployed on the premise of an Infrastructure network (single domain network) managed by a single Infrastructure Provider (InP), and the service function node deployment method for the single domain network is often not applicable to a multi-domain network. The deployment of cross-domain service functions becomes difficult and has the problems of privacy and safety and the like. In deployment across multi-domain networks, much work has considered centralized/distributed deployment architectures. In which, the problem of privacy disclosure may exist in collecting intra-domain information in a centralized manner, and the resources of each domain cannot be shared among domains in a distributed manner. Moreover, this greatly hinders the progress of SFC deployment since the infrastructure providers are reluctant to expose detailed topology information to third parties.
In the embodiment of the disclosure, a first weighted graph corresponding to a service function node set and a second weighted graph corresponding to a physical node set are determined by constructing the service function node network and the physical node network; learning vector representations of the service function nodes and vector representations of the physical nodes based on the first weighted graph and the second weighted graph, wherein the vector representations of the service function nodes and the vector representations of the physical nodes can embody local and global characteristics of the network; and then, based on the vector representation of the service function node and the vector representation of the physical node, the mapping relation between each service node and each physical node is quickly and accurately determined.
In some embodiments, the SFC mapping process may be divided into three layers of a service function topology network, a request mapping network, and an underlying physical network. According to the embodiment of the disclosure, the request mapping network (i.e., the mapping relationship between each service node and each physical node) can be determined based on the preset service function topology network and the basic physical network.
The service function topology network may be a weighted topology graph (i.e., a first weighted graph) and may determine at least two service function chains based on the SFC request; and fusing service function nodes with the same execution function in different service function chains to obtain a fused first weighted graph. First weighted graph
Can be expressed as->
，/>
Expressed as a service function node, <' > or>
Represented as a service function chain.
As shown in fig. 5, the service function set may include a first service function chain SFC1, a second service function chain SFC2, and a third service function chain SFC3, where the SFC1 includes a first service function node 501, a second service function node 502, a third service function node 503, and a fourth service function node 504; the execution time of each service function node is 1 second, 5 seconds and 1 second respectively, and the transmission time delay between each service function node is 2 seconds, 2 seconds and 2 seconds respectively. SFC2 includes a fourth service function node 504, a second service function node 502, a first service function node 501; the execution time of each service function node is 2 seconds, 1 second and 2 seconds respectively, and the transmission time delay between each service function node is 5 seconds and 5 seconds respectively. The SFC3 comprises a second service function node 502, a fifth service function node 505, a fourth service function node 504, a first service function node 501; the execution time of each service function node is 1 second, 3 seconds, 2 seconds and 1 second respectively, and the transmission time delay between each service function node is 2 seconds, 3 seconds and 1 second respectively. The execution duration corresponding to each service function node may be determined as a first weight of each service function node; and determining the transmission delay between the service function nodes as a second weight between the service function nodes.
As shown in fig. 6, the first weights corresponding to the first service function node 601, the second service function node 602, the third service function node 603, the fourth service function node 604, and the fifth service function node 605 in the first weighted graph are 4, 3, 5, and 3, respectively.
The base layer physical network may be a weighted topology map (i.e., a second weighted map), and may determine the remaining resource amount corresponding to each physical node as a third weight of the physical node, and determine the remaining bandwidth between the physical nodes as a fourth weight between the physical nodes. Second weighted graph
Can be expressed as->
，/>
Represented as a set of physical nodes, and>
represented as physical links, etc. As shown in fig. 7, the second weighted graph includes a first physical node 701, a second physical node 702, a third physical node 703, a fourth physical node 704, a fifth physical node 705, and a sixth physical node 706. The residual resource amount of each physical node is respectively 3 million, 8 million, 15 million, 5 million, 11 million and 2 million; the remaining bandwidth between physical nodes is 7 million per second, 5 million per second, 4 million per second, etc., respectively.
The request mapping network may be constructed as a bipartite graph, i.e., if there is a request deployment, there is an edge, otherwise there is no edge; can use
Represents a mapping and->
Represents a set of service function nodes, and>
representing a set of physical nodes.As shown in fig. 8, a first service function node 801 may be deployed to a second physical node 806, a second service function node 802 may be deployed to a first physical node 805, a third service function node 803 may be deployed to a fifth physical node 808, and a fourth service function node 804 may be deployed to a third physical node 807.
In the embodiment of the disclosure, aiming at the problem of collaborative deployment of cross-domain multiple SFCs, a complex heterogeneous information network of the SFCs and physical service nodes is constructed, local and global node feature expression vectors are learned by adopting a weighted graph convolution method, and then the mapping relation between each service node and each physical node is determined; in the training process of the model, the deployment model is trained by optimizing the network parameters through the minimum running time, so that the training precision and efficiency are improved.
Meanwhile, a node semantic feature extraction method based on weighted graph convolution is provided, specifically, the LSTM is used for capturing the dependency relationship among the service function nodes, and the embedded representation of the service function nodes or the physical nodes is initialized; for each service function node, constructing a degree-out neighbor set and a degree-in neighbor set, obtaining degree-out representation and degree-in representation of the service function node through weighted aggregation, and obtaining vector representation of the service function node through linear aggregation; for a physical node, a weighted aggregation function can be used to learn node characterization on an undirected graph (a second weighted graph) to obtain a vector characterization of the physical node. In order to excavate the internal deployment relationship between the service function node and the physical node, a graph convolution network based on a meta-path is also constructed, and an attention mechanism is designed to aggregate semantic representations corresponding to different paths, so that a high-order representation of the service function node and a high-order representation of the physical node are obtained.
Finally, with the minimized time consumption as an optimization target, finding a candidate physical node set meeting the computational resources required by the VNF and possible adjacent service function chain sets and physical chain sets from the physical nodes, and predicting second probabilities of service function nodes and service function chains to the physical nodes and the physical links by using node representation; and constructing an optimization function by utilizing the processing delay and the transmission delay of the physical nodes, optimizing the network model by reverse learning, and the like.
Compared with the related technology, in the embodiment of the disclosure, vector representations of the VNF and the physical nodes can be obtained through offline calculation, and for the SFC online deployment request, decision deployment can be made through the inner product of the calculation node representations, so that the inference module has small capacity and high inference speed, and can meet the requirement of large-scale and cross-domain SFC deployment; meanwhile, a graph convolution method can be adopted, vector representation of the nodes is obtained by mining the internal incidence relation between the VNF and the physical nodes, the local and global VNF adjacency relation can be fully utilized to make reasoning, and therefore deployment precision is high.
Based on the foregoing embodiments, the present disclosure provides a node deployment apparatus, where the apparatus includes units and modules included in the units, and may be implemented by a processor in a computer device; of course, the implementation can also be realized through a specific logic circuit; in the implementation process, the Processor may be a Central Processing Unit (CPU), a Microprocessor Unit (MPU), a Digital Signal Processor (DSP), a Field Programmable Gate Array (FPGA), or the like.
Fig. 9 is a schematic structural diagram of a node deployment apparatus provided in the embodiment of the present disclosure, and as shown in fig. 9, the node deployment apparatus 900 includes: an obtaining module 910, a first determining module 920, a second determining module 930, and a third determining module 940, wherein:
an obtaining module 910, configured to obtain a service function node set and a physical node set to be deployed; the service function node set carries execution time corresponding to each service function node and transmission efficiency among the service function nodes; the physical node set carries the residual resource amount corresponding to each physical node and the residual bandwidth between the physical nodes; a first determining module 920, configured to determine a first weighted graph corresponding to the service function node set based on an execution duration corresponding to each service function node and transmission efficiency between the service function nodes; a second determining module 930, configured to determine, based on the remaining resource amount corresponding to each physical node and the remaining bandwidth between the physical nodes, a second weighted graph corresponding to the set of physical nodes; a third determining module 940, configured to determine, based on the first weighted graph and the second weighted graph, a mapping relationship between each service function node and each physical node; and the mapping relation is used for deploying each service function node to a corresponding physical node.
In some embodiments, the set of service function nodes guarantees at least two service function chains; the first determining module is further configured to: determining an execution function of a service function node in each service function chain; fusing all the service function chains based on the execution functions and the execution time corresponding to each service function node to obtain the execution time after the fusion of each service function node; and determining a first weighted graph corresponding to the service function node set based on the execution duration of each service function node after fusion and the transmission efficiency among the service function nodes.
In some embodiments, the third determining module is further configured to: determining a first adjacency matrix corresponding to the first weighted graph and determining a second adjacency matrix corresponding to the second weighted graph; determining a vector representation for each of the serving function nodes based on the first adjacency matrix; determining a vector characterization for each of the physical nodes based on the second adjacency matrix; and determining the mapping relation between each service function node and each physical node based on the vector representation of each service function node and the vector representation of each physical node.
In some embodiments, the third determining module is further configured to: determining transmission delay between the service function nodes as element values of non-diagonal elements in the first adjacency matrix, and determining execution duration corresponding to each service function node as element values of diagonal elements in the first adjacency matrix; the transmission time delay is determined based on the transmission efficiency and a preset data volume to be transmitted; and determining the residual bandwidth between the physical nodes as the element values of the non-diagonal elements in the second adjacent matrix, and determining the residual resource amount corresponding to each physical node as the element value of the diagonal element in the second adjacent matrix.
In some embodiments, the third determining module is further configured to: normalizing the element value of each element in the first adjacent matrix to obtain a normalized first adjacent matrix; performing feature extraction on the normalized first adjacency matrix by using a preset long-short term memory network to obtain an initial representation of each service function node; determining an in-degree node set and an out-degree node set corresponding to each service function node; aggregating the initial representations of the entry nodes in the entry node set to obtain the entry vector representations of the corresponding service function nodes; performing aggregation processing on the initial representations of the out-degree nodes in the out-degree node set to obtain out-degree vector representations of the corresponding service function nodes; and determining the vector representation of the corresponding service function node based on the in-degree vector representation and the out-degree vector representation of each service function node.
In some embodiments, the third determining module is further configured to: normalizing the element value of each element in the second adjacent matrix to obtain a normalized second adjacent matrix; performing feature extraction on the normalized second adjacency matrix by using a preset long-short term memory network to obtain an initial representation of each physical node; determining an adjacent node set corresponding to each physical node; and carrying out iterative aggregation processing on the initial characterization of each physical node and the initial characterization of the adjacent node in the corresponding adjacent node set to obtain the vector characterization of the corresponding physical node.
In some embodiments, the third determining module is further configured to: extracting the characteristics of the vector characterization of the service function node to obtain the high-order characterization of each service function node; carrying out feature extraction on the vector characterization of the physical nodes to obtain a high-order characterization of each physical node; determining mapping probability of each service function node for deploying corresponding physical nodes based on the high-order characteristics of each service function node and the high-order characteristics of each physical node; and determining the mapping relation between each service function node and each physical node based on all the mapping probabilities.
In some embodiments, the third determining module is further configured to: determining a first meta-path between each of the service function nodes and a corresponding neighboring node; performing linear transformation on the vector representation of the service function node corresponding to each first unitary path to obtain the semantic representation of the corresponding service function node; and carrying out normalization processing on the semantic representation of each service function node to obtain the high-order representation of the corresponding service function node.
In some embodiments, the third determining module is further configured to: determining a second element path between each physical node and the corresponding adjacent node; performing linear transformation on the vector representation of the physical node corresponding to each second element path to obtain the semantic representation of the corresponding physical node; and carrying out normalization processing on the semantic representation of each physical node to obtain the high-order representation of the corresponding physical node.
In some embodiments, the mapping probability comprises a first probability and a second probability; the third determining module is further configured to: determining a current serving function node from the set of serving function nodes, and determining at least two candidate physical nodes from the set of physical nodes that match the previous serving function node; performing logistic regression processing on the high-order characterization of the current service function node and the high-order characterization of each candidate physical node to obtain a first probability that the current service function node is deployed to each candidate physical node; determining a current candidate node from at least two of the candidate physical nodes based on the first probability; acquiring a next service function node and a next candidate node corresponding to the next service function node; performing logistic regression processing on the high-order representation of the current service function node, the high-order representation of the current candidate node, the high-order representation of the next service function node and the high-order representation of the next candidate node to obtain a second probability that a link from the current service function node to the next service function node is mapped to a link from the current candidate node to the next candidate node; the third determining module is further configured to: and determining the mapping relation between each service function node and each physical node based on the second probability.
In some embodiments, the third determining module is further configured to: acquiring a trained deployment model; the deployment model at least comprises a first representation learning network, a second representation learning network, a high-order learning network and a probability prediction network; processing the first weighted graph by using the first characterization learning network to obtain the vector characterization of each service function node; processing the second weighted graph by using the second representation learning network to obtain the vector representation of each physical node; respectively processing the vector representation of each service function node and the vector representation of each physical node by using the high-order learning network to obtain the high-order representation of each service function node and the high-order representation of each physical node; predicting the high-order representation of each service function node and the high-order representation of each physical node by using the probability prediction network to obtain the mapping probability of each service function node for deploying the corresponding physical node; wherein the mapping probability comprises a first probability and a second probability, the first probability and the second probability being used to construct an optimization function; updating parameters of an untrained deployment model under the condition that the function value of the optimization function is minimum to obtain the trained deployment model; and determining the mapping relation between each service function node and each physical node based on all the mapping probabilities.
The above description of the apparatus embodiments, similar to the above description of the method embodiments, has similar beneficial effects as the method embodiments. In some embodiments, functions of or modules included in the apparatuses provided in the embodiments of the present disclosure may be used to perform the methods described in the above method embodiments, and for technical details not disclosed in the embodiments of the apparatuses of the present disclosure, please refer to the description of the method embodiments of the present disclosure for understanding.
It should be noted that, in the embodiment of the present disclosure, if the node deployment method is implemented in the form of a software functional module and is sold or used as a standalone product, the node deployment method may also be stored in a computer-readable storage medium. Based on such understanding, the technical solutions of the embodiments of the present disclosure may be embodied in the form of a software product, which is stored in a storage medium and includes several 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 methods described in the embodiments of the present disclosure. 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 magnetic disk, or an optical disk. Thus, embodiments of the present disclosure are not limited to any specific hardware, software, or firmware, or any combination thereof.
The embodiment of the present disclosure provides a computer device, which includes a memory and a processor, where the memory stores a computer program that can be run on the processor, and the processor implements some or all of the steps of the above method when executing the program.
The disclosed embodiments provide a computer-readable storage medium having stored thereon a computer program that, when executed by a processor, performs some or all of the steps of the above-described method. The computer readable storage medium may be transitory or non-transitory.
The disclosed embodiments provide a computer program comprising computer readable code, where the computer readable code runs in a computer device, a processor in the computer device executes some or all of the steps for implementing the above method.
The disclosed embodiments provide a computer program product comprising a non-transitory computer readable storage medium storing a computer program that when read and executed by a computer performs some or all of the steps of the above method. The computer program product may be embodied in hardware, software or a combination thereof. In some embodiments, the computer program product is embodied in a computer storage medium, and in other embodiments, the computer program product is embodied in a Software product, such as a Software Development Kit (SDK), or the like.
Here, it should be noted that: the foregoing description of the various embodiments is intended to highlight various differences between the embodiments, which are the same or similar and all of which are referenced. The above description of the apparatus, storage medium, computer program and computer program product embodiments is similar to the description of the method embodiments above, with similar advantageous effects as the method embodiments. For technical details not disclosed in the embodiments of the disclosed apparatus, storage medium, computer program and computer program product, reference is made to the description of the embodiments of the method of the present disclosure for understanding.
It should be noted that fig. 10 is a schematic diagram of a hardware entity of a computer device in an embodiment of the present disclosure, and as shown in fig. 10, the hardware entity of the computer device 1000 includes: a processor 1001, a communication interface 1002, and a memory 1003, wherein:
the processor 1001 generally controls the overall operation of the computer apparatus 1000.
The communication interface 1002 may enable the computer device to communicate with other terminals or servers via a network.
The Memory 1003 is configured to store instructions and applications executable by the processor 1001, and may also buffer data (e.g., image data, audio data, voice communication data, and video communication data) to be processed or already processed by the processor 1001 and modules in the computer apparatus 1000, and may be implemented by a FLASH Memory (FLASH) or a Random Access Memory (RAM). Data transmission between the processor 1001, the communication interface 1002, and the memory 1003 can be performed via the bus 1004.
It should be appreciated that reference throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, structure or characteristic described in connection with the embodiment is included in at least one embodiment of the present disclosure. Thus, the appearances of the phrases "in one embodiment" or "in an embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. It should be understood that, in various embodiments of the present disclosure, the sequence numbers of the above steps/processes do not mean the execution sequence, and the execution sequence of each step/process should be determined by the function and the inherent logic thereof, and should not constitute any limitation on the implementation process of the embodiments of the present disclosure. The above-mentioned serial numbers of the embodiments of the present disclosure are merely for description and do not represent the merits of the embodiments.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising a … …" does not exclude the presence of another identical element in a process, method, article, or apparatus that comprises the element.
In the several embodiments provided in the present disclosure, it should be understood that the disclosed apparatus and method may be implemented in other ways. The above-described device embodiments are merely illustrative, for example, the division of the unit is only a logical functional division, and there may be other division ways in actual implementation, such as: multiple units or components may be combined, or may be integrated into another system, or some features may be omitted, or not implemented. In addition, the coupling, direct coupling or communication connection between the components shown or discussed may be through some interfaces, and the indirect coupling or communication connection between the devices or units may be electrical, mechanical or other forms.
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; can be located in one place or 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, all the functional units in the embodiments of the present disclosure may be integrated into one processing unit, or each unit may be separately regarded as one unit, or two or more units may be integrated into one unit; the integrated unit may be implemented in the form of hardware, or in the form of hardware plus a software functional unit.
Those of ordinary skill in the art will understand that: all or part of the steps for realizing the method embodiments can be completed by hardware related to program instructions, the program can be stored in a computer readable storage medium, and the program executes the steps comprising the method embodiments when executed; and the aforementioned storage medium includes: various media that can store program codes, such as a removable Memory device, a Read Only Memory (ROM), a magnetic disk, or an optical disk.
Alternatively, the integrated unit of the present disclosure may be stored in a computer-readable storage medium if it is implemented in the form of a software functional module and sold or used as a separate product. Based on such understanding, the technical solutions of the present disclosure may be substantially or partially embodied in the form of a software product stored in a storage medium, and include several instructions to enable a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the methods described in the embodiments of the present disclosure. And the aforementioned storage medium includes: a removable storage device, a ROM, a magnetic or optical disk, or other various media that can store program code.
The methods disclosed in the several method embodiments provided in this disclosure may be combined arbitrarily without conflict to arrive at new method embodiments.
If the disclosed embodiment relates to personal information, a product applying the disclosed embodiment explicitly informs personal information processing rules and obtains personal self-approval before processing the personal information. If the disclosed embodiment relates to sensitive personal information, the product applying the disclosed embodiment obtains individual consent before processing the sensitive personal information, and simultaneously meets the requirement of 'explicit consent'.
The above description is only an embodiment of the present disclosure, but the scope of the present disclosure is not limited thereto, and any person skilled in the art can easily conceive of changes or substitutions within the technical scope of the present disclosure, and all the changes or substitutions should be covered by the scope of the present disclosure.
Claims (14)
1. A node deployment method, comprising:
acquiring a service function node set and a physical node set to be deployed; the service function node set carries execution time corresponding to each service function node and transmission efficiency among the service function nodes; the physical node set carries the residual resource amount corresponding to each physical node and the residual bandwidth between the physical nodes;
determining a first weighted graph corresponding to the service function node set based on the execution duration corresponding to each service function node and the transmission efficiency among the service function nodes;
determining a second weighted graph corresponding to the physical node set based on the residual resource amount corresponding to each physical node and the residual bandwidth between the physical nodes;
determining a mapping relationship between each service function node and each physical node based on the first weighted graph and the second weighted graph; and the mapping relation is used for deploying each service function node to a corresponding physical node.
2. The method of claim 1, wherein the set of service function nodes guarantees at least two service function chains; the determining a first weighted graph corresponding to the service function node set based on the execution duration corresponding to each service function node and the transmission efficiency between the service function nodes includes:
determining an execution function of a service function node in each service function chain;
fusing all the service function chains based on the execution functions and the execution duration corresponding to each service function node to obtain the execution duration fused with each service function node;
and determining a first weighted graph corresponding to the service function node set based on the execution duration of each service function node after fusion and the transmission efficiency among the service function nodes.
3. The method of claim 1, wherein determining a mapping relationship between each service function node and each physical node based on the first weighted graph and the second weighted graph comprises:
determining a first adjacency matrix corresponding to the first weighted graph and determining a second adjacency matrix corresponding to the second weighted graph;
determining a vector representation for each of the serving function nodes based on the first adjacency matrix;
determining a vector characterization for each of the physical nodes based on the second adjacency matrix;
and determining the mapping relation between each service function node and each physical node based on the vector representation of each service function node and the vector representation of each physical node.
4. The method of claim 3, wherein determining a first adjacency matrix to which the first weighted graph corresponds and determining a second adjacency matrix to which the second weighted graph corresponds comprises:
determining transmission delay between the service function nodes as element values of non-diagonal elements in the first adjacency matrix, and determining execution duration corresponding to each service function node as element values of diagonal elements in the first adjacency matrix; the transmission time delay is determined based on the transmission efficiency and a preset data volume to be transmitted;
and determining the residual bandwidth between the physical nodes as the element values of the non-diagonal elements in the second adjacent matrix, and determining the residual resource amount corresponding to each physical node as the element value of the diagonal element in the second adjacent matrix.
5. The method of claim 3, wherein said determining a vector characterization for each of said serving function nodes based on said first adjacency matrix comprises:
normalizing the element value of each element in the first adjacent matrix to obtain a normalized first adjacent matrix;
performing feature extraction on the normalized first adjacency matrix by using a preset long-short term memory network to obtain an initial representation of each service function node;
determining an in-degree node set and an out-degree node set corresponding to each service function node;
aggregating the initial representations of the entry nodes in the entry node set to obtain the entry vector representations of the corresponding service function nodes;
performing aggregation processing on the initial representations of the out-degree nodes in the out-degree node set to obtain out-degree vector representations of the corresponding service function nodes;
and determining the vector representation of the corresponding service function node based on the in-degree vector representation and the out-degree vector representation of each service function node.
6. The method of claim 3, wherein said determining a vector characterization for each of said physical nodes based on said second adjacency matrix comprises:
normalizing the element value of each element in the second adjacent matrix to obtain a normalized second adjacent matrix;
performing feature extraction on the normalized second adjacency matrix by using a preset long-short term memory network to obtain an initial representation of each physical node;
determining an adjacent node set corresponding to each physical node;
and carrying out iterative aggregation processing on the initial characterization of each physical node and the initial characterization of the adjacent node in the corresponding adjacent node set to obtain the vector characterization of the corresponding physical node.
7. The method of claim 3, wherein the determining a mapping relationship between each service function node and each physical node based on the vector characterization of each service function node and the vector characterization of each physical node comprises:
extracting the characteristics of the vector characterization of the service function node to obtain the high-order characterization of each service function node;
carrying out feature extraction on the vector characterization of the physical nodes to obtain a high-order characterization of each physical node;
determining the mapping probability of each service function node for deploying the corresponding physical node based on the high-order characteristics of each service function node and the high-order characteristics of each physical node;
and determining the mapping relation between each service function node and each physical node based on all the mapping probabilities.
8. The method of claim 7, wherein said extracting features from said vector representations of said service function nodes to obtain a higher-order representation of each said service function node comprises:
determining a first meta-path between each of the service function nodes and a corresponding neighboring node;
performing linear transformation on the vector representation of the service function node corresponding to each first unitary path to obtain the semantic representation of the corresponding service function node;
and carrying out normalization processing on the semantic representation of each service function node to obtain the high-order representation of the corresponding service function node.
9. The method of claim 7, wherein said extracting features from said vector representations of said physical nodes to obtain a higher-order representation of each of said physical nodes comprises:
determining a second element path between each physical node and the corresponding adjacent node;
performing linear transformation on the vector representation of the physical node corresponding to each second element path to obtain the semantic representation of the corresponding physical node;
and carrying out normalization processing on the semantic representation of each physical node to obtain the high-order representation of the corresponding physical node.
10. The method of claim 7, wherein the mapping probability comprises a first probability and a second probability; the determining, based on the high-order characterization of each service function node and the high-order characterization of each physical node, a mapping probability that each service function node deploys a corresponding physical node includes:
determining a current serving function node from the set of serving function nodes, and determining at least two candidate physical nodes from the set of physical nodes that match the previous serving function node;
performing logistic regression processing on the high-order characterization of the current service function node and the high-order characterization of each candidate physical node to obtain a first probability that the current service function node is deployed to each candidate physical node;
determining a current candidate node from at least two of the candidate physical nodes based on the first probability;
acquiring a next service function node and a next candidate node corresponding to the next service function node;
performing logistic regression processing on the high-order representation of the current service function node, the high-order representation of the current candidate node, the high-order representation of the next service function node and the high-order representation of the next candidate node to obtain a second probability that a link from the current service function node to the next service function node is mapped to a link from the current candidate node to the next candidate node;
the determining a mapping relationship between each service function node and each physical node based on all the mapping probabilities includes:
and determining the mapping relation between each service function node and each physical node based on the second probability.
11. The method according to claim 1, wherein the determining a mapping relationship between each service function node and each physical node based on the first weighted graph and the second weighted graph comprises:
acquiring a trained deployment model; the deployment model at least comprises a first representation learning network, a second representation learning network, a high-order learning network and a probability prediction network;
processing the first weighted graph by using the first representation learning network to obtain the vector representation of each service function node;
processing the second weighted graph by using the second characterization learning network to obtain the vector characterization of each physical node;
respectively processing the vector representation of each service function node and the vector representation of each physical node by using the high-order learning network to obtain the high-order representation of each service function node and the high-order representation of each physical node;
predicting the high-order characteristics of each service function node and the high-order characteristics of each physical node by using the probability prediction network to obtain the mapping probability of each service function node for deploying the corresponding physical node; wherein the mapping probability comprises a first probability and a second probability, the first probability and the second probability being used to construct an optimization function; updating parameters of an untrained deployment model under the condition that the function value of the optimization function is minimum to obtain the trained deployment model;
and determining the mapping relation between each service function node and each physical node based on all the mapping probabilities.
12. A node deployment apparatus, comprising:
the system comprises an acquisition module, a configuration module and a configuration module, wherein the acquisition module is used for acquiring a service function node set and a physical node set to be deployed; the service function node set carries execution time corresponding to each service function node and transmission efficiency among the service function nodes; the physical node set carries the residual resource amount corresponding to each physical node and the residual bandwidth between the physical nodes;
a first determining module, configured to determine, based on an execution duration corresponding to each service function node and transmission efficiency between the service function nodes, a first weighted graph corresponding to the service function node set;
a second determining module, configured to determine, based on a remaining resource amount corresponding to each physical node and a remaining bandwidth between the physical nodes, a second weighted graph corresponding to the set of physical nodes;
a third determining module, configured to determine, based on the first weighted graph and the second weighted graph, a mapping relationship between each service function node and each physical node; and the mapping relation is used for deploying each service function node to a corresponding physical node.
13. A computer device comprising a memory and a processor, the memory storing a computer program operable on the processor, wherein the processor implements the steps of the method of any one of claims 1 to 11 when executing the program.
14. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 11.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310140237.9A CN115842744B (en) | 2023-02-21 | 2023-02-21 | Node deployment method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310140237.9A CN115842744B (en) | 2023-02-21 | 2023-02-21 | Node deployment method, device, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115842744A true CN115842744A (en) | 2023-03-24 |
| CN115842744B CN115842744B (en) | 2023-07-14 |
Family
ID=85579963
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310140237.9A Active CN115842744B (en) | 2023-02-21 | 2023-02-21 | Node deployment method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115842744B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108259238A (en) * | 2018-01-10 | 2018-07-06 | 重庆邮电大学 | The method that network function service chaining maps in data center |
| US20180227244A1 (en) * | 2017-02-03 | 2018-08-09 | Fujitsu Limited | Efficient message forwarding for distributed resource orchestration |
| CN110134493A (en) * | 2019-05-05 | 2019-08-16 | 西安邮电大学 | A kind of service function chain Deployment Algorithm avoided based on resource fragmentation |
| CN113708972A (en) * | 2021-08-31 | 2021-11-26 | 广东工业大学 | Service function chain deployment method and device, electronic equipment and storage medium |
| CN113904923A (en) * | 2021-09-27 | 2022-01-07 | 重庆电子工程职业学院 | Service function chain joint optimization method based on software defined network |
| CN114205317A (en) * | 2021-10-21 | 2022-03-18 | 北京邮电大学 | Service function chain SFC resource allocation method based on SDN and NFV and electronic equipment |
-
2023
- 2023-02-21 CN CN202310140237.9A patent/CN115842744B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180227244A1 (en) * | 2017-02-03 | 2018-08-09 | Fujitsu Limited | Efficient message forwarding for distributed resource orchestration |
| CN108259238A (en) * | 2018-01-10 | 2018-07-06 | 重庆邮电大学 | The method that network function service chaining maps in data center |
| CN110134493A (en) * | 2019-05-05 | 2019-08-16 | 西安邮电大学 | A kind of service function chain Deployment Algorithm avoided based on resource fragmentation |
| CN113708972A (en) * | 2021-08-31 | 2021-11-26 | 广东工业大学 | Service function chain deployment method and device, electronic equipment and storage medium |
| CN113904923A (en) * | 2021-09-27 | 2022-01-07 | 重庆电子工程职业学院 | Service function chain joint optimization method based on software defined network |
| CN114205317A (en) * | 2021-10-21 | 2022-03-18 | 北京邮电大学 | Service function chain SFC resource allocation method based on SDN and NFV and electronic equipment |
Non-Patent Citations (1)
| Title |
|---|
| 李丹 等: "基于最优加权图匹配的服务功能链部署方法" * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115842744B (en) | 2023-07-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Alaiz-Moreton et al. | Multiclass classification procedure for detecting attacks on MQTT-IoT protocol | |
| US10262019B1 (en) | Distributed management optimization for IoT deployments | |
| Kondratenko et al. | Multi-criteria decision making for selecting a rational IoT platform | |
| WO2022171066A1 (en) | Task allocation method and apparatus based on internet-of-things device, and network training method and apparatus | |
| Bahutair et al. | Multi-perspective trust management framework for crowdsourced IoT services | |
| Wang et al. | A multitask learning-based network traffic prediction approach for SDN-enabled industrial internet of things | |
| CN111368973A (en) | Method and apparatus for training a hyper-network | |
| CN111695046A (en) | User portrait inference method and device based on spatio-temporal mobile data representation learning | |
| JP2021060982A (en) | Data analysis system diagnostic method, data analysis system optimization method, device, and medium | |
| CN115034379A (en) | Causal relationship determination method and related equipment | |
| US20240095529A1 (en) | Neural Network Optimization Method and Apparatus | |
| CN114240506A (en) | Modeling method of multi-task model, promotion content processing method and related device | |
| Putrada et al. | Edgesl: Edge-computing architecture on smart lighting control with distilled knn for optimum processing time | |
| WO2023143570A1 (en) | Connection relationship prediction method and related device | |
| CN116668351A (en) | Quality of service prediction method, device, computer equipment and storage medium | |
| CN115842744B (en) | Node deployment method, device, equipment and storage medium | |
| Sountharrajan et al. | On-the-go network establishment of iot devices to meet the need of processing big data using machine learning algorithms | |
| Mahapatra et al. | Next Generation Task Offloading Techniques in Evolving Computing Paradigms: Comparative Analysis, Current Challenges, and Future Research Perspectives | |
| CN115759183A (en) | Related method and related device for multi-structure text graph neural network | |
| Afrasiabi et al. | Reinforcement learning-based optimization framework for application component migration in nfv cloud-fog environments | |
| WO2022252596A1 (en) | Method for constructing ai integrated model, and inference method and apparatus of ai integrated model | |
| CN115618065A (en) | Data processing method and related equipment | |
| Kurniawan et al. | Mobile computing and communications-driven fog-assisted disaster evacuation techniques for context-aware guidance support: A survey | |
| CN114866438A (en) | Abnormal hidden danger prediction method and system under cloud architecture | |
| CN116090504A (en) | Training method and device for graphic neural network model, classifying method and computing equipment |
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 |