CN112953761B - Virtual-real resource mapping method for virtual network construction in multi-hop network - Google Patents
Virtual-real resource mapping method for virtual network construction in multi-hop network Download PDFInfo
- Publication number
- CN112953761B CN112953761B CN202110114947.5A CN202110114947A CN112953761B CN 112953761 B CN112953761 B CN 112953761B CN 202110114947 A CN202110114947 A CN 202110114947A CN 112953761 B CN112953761 B CN 112953761B
- Authority
- CN
- China
- Prior art keywords
- virtual
- virtual network
- network
- link
- topology
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/12—Discovery or management of network topologies
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/142—Network analysis or design using statistical or mathematical methods
Landscapes
- Engineering & Computer Science (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Mathematical Physics (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Algebra (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Pure & Applied Mathematics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Probability & Statistics with Applications (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention provides a resource virtual-real mapping method for virtual network construction in a multi-hop network, which comprises the following steps: s1: abstracting a physical network, and establishing an initialized virtual network topological structure connected in pairs; s2: optimizing the virtual network topology by adopting a small world theory and through a virtual network evaluation index to obtain an optimized topology connection relation; s3: sequentially endowing each link of the virtual network topology with equal communication resources, and distributing different communication resources for each link by the residual resources according to the parallel service transmission number; s4: and (3) finishing end-to-end virtual link planning by adopting a path selection method based on reinforcement learning, and mapping virtual links among the virtual network user nodes into actual links. The invention can improve the link utilization rate and reduce the network bandwidth overhead under the condition of meeting the requirement of congestion-free communication among multiple users.
Description
Technical Field
The invention relates to the technical field of communication, in particular to a resource virtual-real mapping method for virtual network construction in a multi-hop network.
Background
At present, because the situation of the communication environment has the problems of complexity, variability, a large number of communication nodes and complicated resource management, a virtual network is generally constructed to effectively manage a complex physical network. The network virtualization is to construct a virtual network on the existing network to meet the requirements of diversified services on the premise of keeping the existing internet architecture; the method is an important scheme for dealing with the current network rigidity, and is an important paradigm that a plurality of heterogeneous isolation virtual networks share a bottom physical resource network in the future.
Virtual network mapping is a key to instantiate a virtual network to achieve network resource allocation. The virtual network mapping algorithm facing the differentiated service request provides a virtual node mapping algorithm based on node saturation and a virtual link mapping algorithm based on node saturation by defining the node saturation, and deals with the situation that the virtual node resource request and the virtual link resource request in the virtual network request are different greatly, so that the network resource is distributed unevenly.
The network virtualization hierarchical model comprises a 3-layer structure as follows: an infrastructure layer, a virtualization overlay layer, and a service layer. Transparent isolation is formed between an application layer and a bottom layer basic network through a virtualization layer, and a mapping relation is established by adopting a dynamic resource allocation and control technology, so that the internet capability is improved and promoted, and the requirements of end-to-end service quality, safety, manageability and the like are met.
In one of the prior art, a virtual network mapping algorithm for differentiated service requests maps a virtual node to a physical node that satisfies the resource requirement and has a node saturation closest to the virtual node saturation, determines a physical node bearing the virtual node and a part of physical links on a physical path bearing the virtual link, and maps the virtual link to a physical path composed of physical links satisfying the resource request, as shown in fig. 1.
Firstly, defining the resource ratio of a node to a connected link as node saturation, including the saturation of a physical node and the saturation of a virtual node; secondly, mapping the virtual nodes on the residual resources by applying a virtual node mapping algorithm based on the node saturation to meet the virtual node resource request, and determining the physical nodes bearing the virtual nodes on the physical nodes of which the node saturation is closest to the virtual node saturation in the virtual network request; and finally, determining a part of physical links on the physical path for bearing the virtual links, applying a virtual link mapping algorithm based on the node saturation, selecting one physical path with the node saturation closest to the node saturation requested by the virtual network to bear the virtual links, and instantiating the virtual network.
However, one of the above technical solutions in the prior art is based on the combination of virtual links, and the virtual links are built between every two nodes to implement the building of the network virtual network form, although the optimization strategy based on the node saturation can make the network resource distribution more balanced. However, under larger-scale network conditions, the method causes the problems of high complexity, multiple redundant paths, low resource utilization rate and the like.
In the second prior art, the hierarchical model for network virtualization includes the following 3 layers: infrastructure layer, virtualization overlay layer, and service layer, as shown in fig. 2. The infrastructure layer is made up of a multitude of infrastructure providers (InP), each InP comprising a large number of programmable virtualization-capable nodes, such as routers and end servers. The details of the bottom layer resources are shielded by the InP through a unified description mode of the virtualized resources, and unified virtual resource abstraction is provided for the upper layer. The virtualization overlay is based on a distributed intelligent multi-agent system and consists of a series of distributed domain managers. Each domain manager corresponds to a respective InP. The domain manager is responsible for detecting and abstracting resources, distributing and scheduling virtual resources, and establishing and instantiating a virtual network. The service layer is composed of a plurality of virtual networks which are isolated from each other. These virtual networks are driven by requests initiated by users or user agents, and are scheduled, distributed, created by domain managers to accomplish specific functions. These specific functions mainly include two major aspects: the novel network protocol, the test verification of the architecture and the specific virtual private network topology are realized aiming at specific service requirements.
For the coordination and interaction problem among the domain managers in the virtualization layer, a centralized or distributed management mode can be adopted to interact information according to a specific protocol so as to obtain the resource information of the related domain managers. Each domain manager is internally composed of a series of corresponding functional modules: resource discovery, context modeling, resource mapping, a specific policy library, and the like. The resource discovery mainly determines the topology of the management network and the state of the corresponding network elements, further, the adjacent domain managers also share the state information, links are established among the networks, and the construction of the inter-domain virtual network is realized. The virtual resource mapping is mainly to allocate virtual nodes and virtual links to corresponding physical nodes and paths according to the constraints in the policy base.
In the network virtualization hierarchical model, topology construction and resource allocation of a virtual network are completed by a domain manager. However, when the network scale is increased, no appropriate scheme is proposed for inter-domain management of the domain manager, and the workload of the corresponding domain manager is also significantly increased, which may increase the information synchronization overhead. However, there is no uniform method for mapping virtual resources, and it is difficult to implement adaptive allocation of resources.
Disclosure of Invention
The invention provides a resource virtual-real mapping method for virtual network construction in a multi-hop network, which aims to solve the problems of high complexity, multiple redundant paths, low resource utilization rate and the like caused by the fact that most of traditional virtual network construction methods are based on combination of virtual links and construction of network virtual network forms is realized by constructing the virtual links between every two users.
In order to solve the technical problems, the technical scheme of the invention is as follows: a virtual-real mapping method of resources constructed for a virtual network in a multi-hop network comprises the following steps:
s1: abstracting a physical network, and establishing an initialized virtual network topological structure;
s2: optimizing the virtual network topology by adopting a small world theory and through the virtual network evaluation index to obtain an optimized topology connection relation;
s3: sequentially endowing each link of the virtual network topology with equal communication resources, and distributing different communication resources for each link by the residual resources according to the parallel service transmission number;
s4: and (3) finishing end-to-end virtual link planning by adopting a path selection method based on reinforcement learning, and mapping virtual links among the virtual network user nodes into actual links.
Preferably, in step S1, the physical location of the node in the physical network is specifically confirmed, and the connection relationship between all nodes is confirmed, so as to perform abstraction to establish the initialized virtual network topology.
Furthermore, all the logic nodes in the initial virtual network are fully connected; after the virtual network is established, a physical resource pool is called, resource information is abstracted and constructed in several ways including combination, deletion and filling, and a virtual resource pool database is filled by adopting a normalization output result.
Furthermore, resource information in the virtual resource pool is abstracted into a logic network expressed by a graph, and a network topology relation is established by a maximum flow algorithm in a graph theory, namely, the virtual link capacity between virtual user nodes is acquired by physical resource topology through the maximum flow algorithm; and finishing the combing of the logic relationship between the virtual user nodes by adopting a link combination mode between the virtual user nodes so as to finish establishing a primary virtual network topology.
And further, the virtual network evaluation index comprises a virtual network weighted path length, a virtual network weighted efficiency and a virtual network weighted betweenness.
Further, the virtual network weighted path length is obtained by weighting the actual physical path length in the virtual network, and the expression is as follows:
in the formula (d) i,j Representing the distance d between a virtual user node i and a virtual user node j in a virtual network i,j The method is used for describing the number of edges on the shortest path between two virtual user nodes;
representing the physical path lengths in the logical links that belong between virtual user node i and virtual user node j in the actual physical topology.
Still further, the expression of the virtual network weighting efficiency is as follows:
where N represents the number of virtual network nodes.
Further, step S2, a small world theory is adopted, and the virtual network topology is optimized through the virtual network evaluation index, which specifically includes:
s201: initial virtual network information is obtained through the virtual resource pool,
s202: calculating the weighted betweenness of all the virtual networks in the current virtual network according to a shortest path algorithm;
s203: arranging all edges in the virtual network from small to large according to the weighted betweenness of the corresponding virtual networks, and setting the communication link bandwidth of the first edge to be 0, namely disconnecting the communication link bandwidth;
s204: and calculating the virtual network weighting efficiency in the current virtual network according to the virtual network weighting path length, finishing the virtual network topology optimization if the virtual network weighting efficiency exceeds a threshold value, and otherwise returning to the step S202 to continue executing.
Still further, in step S3, the specific steps are as follows:
s301: acquiring the total number of allocable virtual resources of the current virtual network;
s302: calculating the minimum resource quantity for maintaining the virtual network, and sequentially endowing each link of the virtual network topology with equal communication resources;
s303: counting the parallel service transmission number of each virtual link in the current virtual network;
s304: and distributing different communication resources for each link according to the parallel service transmission number by using the residual resources in the link resource pool.
Compared with the prior art, the technical scheme of the invention has the beneficial effects that:
1. the method and the device have the advantages that the maximum link capacity between the nodes is obtained under a limited scene based on the virtual resource pool abstracted from the bottom physical resource pool, enough bandwidth is reserved for the virtual network user, and the virtual network request of the user is ensured.
2. The invention utilizes the small world theory when optimizing the virtual network topology, improves the link utilization rate and reduces the virtual network control and bandwidth overhead on the premise of ensuring the virtual network node communication.
3. The virtual network has the function of self-adapting service distribution, and the communication resources of the links can be dynamically adjusted according to the parallel service transmission number of each virtual link after the network is established.
4. When the virtual network performs virtual-real mapping, the invention adopts a path selection method based on reinforcement learning to complete end-to-end virtual link planning and map virtual links between virtual network nodes into actual links, and the time complexity of the path selection method is relatively low.
5. The virtual network has good extensibility, and the virtual network user nodes can dynamically join and leave the virtual network.
Drawings
Fig. 1 is a network model of a differentiated service request oriented virtual network mapping algorithm in the prior art.
FIG. 2 is a layered model of two network virtualization technologies in the prior art.
Fig. 3 is a functional diagram of a virtual network construction process in embodiment 1.
Fig. 4 is a flowchart of a virtual network construction process in embodiment 1.
Fig. 5 is a physical topology diagram of the zone of embodiment 1.
Fig. 6 is a virtual network topology diagram of embodiment 1.
Fig. 7 is a typical virtual network topology of embodiment 1.
Fig. 8 is a comparison graph of the virtual network topology before and after optimization in example 1.
Fig. 9 is a flow of the small world network algorithm of embodiment 1.
Fig. 10 is an initial allocation of virtual network resources according to embodiment 1.
Fig. 11 is a comparison between before and after the adaptive optimization of virtual network resources according to example 1.
Fig. 12 is a flow chart of the virtual network resource adaptive optimization algorithm of embodiment 1.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and are used for illustration only and should not be construed as limiting the patent. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The technical solution of the present invention is further described below with reference to the accompanying drawings and examples.
Example 1
Due to the problem that the communication environment situation is complex, the number of communication nodes is large, the resource management is complex, and a virtual network can be constructed to effectively manage a complex physical network. After the virtual network is constructed, corresponding resources are reserved successfully in the physical network, so that the communication requirement between any two nodes needing to initiate information transmission in the network can be quickly met. Aiming at a virtual network construction request related to multi-user interconnection and intercommunication, a virtual network construction process is mainly divided into three processes of constructing a virtual network topology, optimizing a small world topology and mapping virtual and real resources, so that a process that a communication network is from real to virtual and then from virtual to real is realized, as shown in fig. 3.
Specifically, the present embodiment provides a virtual-real resource mapping method for virtual network construction in a multi-hop network, and as shown in fig. 4, the method includes the following steps:
s1: the virtual network request node reports the virtual network request to a cluster head of a distributed management center, the cluster head of the distributed management center carries out node combination in a physical network, a link is simplified, the physical network is abstracted, and an initialization virtual network topological structure connected in pairs is established;
s2: optimizing the virtual network topology by adopting a small world theory and through a virtual network evaluation index to obtain an optimized topology connection relation;
s3: sequentially endowing each link of the virtual network topology with equal communication resources, and distributing different communication resources to each link by the residual resources according to the parallel service transmission number, thereby completing self-adaptive distribution;
s4: and finishing end-to-end virtual link planning by adopting a path selection method based on reinforcement learning, and mapping virtual links among the virtual network user nodes into actual links.
The embodiment is influenced by the massive amount of physical communication equipment and the diversification of communication network resources, so that the mobile communication network is full of massive and complicated physical network information. Effective information is directly and accurately identified in massive physical information, and the rapid and accurate transmission of the effective information in a communication network is realized, which can hardly be realized by the prior art. And due to the influence of complex environment situation, physical information such as geographic position, electromagnetic environment, available bandwidth and the like has strong timeliness, and the information management complexity is further aggravated by the dynamicity of resource information. Therefore, it is not realistic to directly manage the physical network information resources.
The virtual network is a logical network containing virtual network links, and a connection relationship topology which is more convenient to manage is obtained by abstracting the internal logical relationship of an actual physical network, wherein the virtual network links refer to physical connection relationships which replace entity nodes through network virtualization among a plurality of communication devices. Compared with the traditional mode of directly managing the physical communication network, the virtual network can more flexibly and effectively extract the physical information which is needed, the physical information becomes a logic network variable to carry out unified management, and the complexity of the information management process of the communication network is greatly reduced.
Step S1, the physical position of the node in the physical network is confirmed concretely, the connection relation among all the nodes is confirmed, and therefore the initialized virtual network topological structure is established in an abstract mode. The present embodiment uses the underlying physical topology to describe the relationship between the underlying physical nodes and the wireless links, as shown in fig. 5. The underlying physical topology is described by graph theory G (V, E), where V is the set of physical nodes and E is the set of edges connecting the physical nodes. Figure 5 is a schematic diagram of the physical topology of any one zone,there are 11 physical nodes in the graph, and edges connecting the 11 physical nodes. Each edge corresponds to a link in the communication, if two nodes v i And v j There is a communication link between them, there is a corresponding edge e in the edge set ij And if no communication link exists between the two nodes, no corresponding edge exists in the edge set.
The physical resource pool information corresponding to the bottom-layer physical topology is shown in table 1. The row and column numbers respectively correspond to the numbers of the nodes, the specific elements in the table represent bandwidth information of the communication link between the corresponding nodes, for example, the value of the third column in the second row of the table is 9Mbps, the bandwidth of the communication link between the node 1 and the node 2 is 9Mbps, the value of the fifth column in the sixth row is 11Mbps, the bandwidth of the communication link between the node 5 and the node 4 is 11Mbps, and the value of the fourth column in the second row is 0, which indicates that no direct communication link exists between the node 1 and the node 3.
TABLE 1 physical resource pool
After the virtual network is established, the virtual network calls a physical resource pool in reported physical information, abstractly constructs the virtual resource pool by adopting a plurality of modes including combination, deletion and filling of the required resource information, and fills a resource pool database by adopting a normalization output result. The conversion of the networking request to the virtual network is shown in fig. 5 and 6. The number of nodes in the domain is 11, the number of virtual network users is 5, after resource abstraction and virtualization, the matrix dimension of the physical resource pool is reduced from the original dimension (11,11) of the table 1 to the dimension (5,5) of the table 2, and the complexity of a subsequent algorithm can be obviously reduced. The virtual network topology is shown in fig. 6, wherein the node 1, the node 3, the node 8, the node 9 and the node 10 are virtual network user nodes, and dotted lines among the virtual network user nodes represent logical links among users. When the virtual network topology is initialized, a logical link is established between each virtual network user, that is, the virtual network users are fully connected.
TABLE 2 virtual resource pool
| 1 | 3 | 8 | 9 | 10 | ||
| 1 | 0 | 2Mbps | | 2Mbps | 2Mbps | |
| 3 | 2Mbps | 0 | | 2Mbps | 2Mbps | |
| 8 | 2Mbps | 2Mbps | 0 | | 2Mbps | |
| 9 | 2Mbps | 2Mbps | 2Mbps | 0 | |
|
| 10 | 2Mbps | 2Mbps | 2Mbps | 2Mbps | 0 |
In the expression of the virtual network topology, the method of using a graph composed of "nodes" and "edges" is common and intuitive. In the figure, the attributes of interest in the communication network, such as network connectivity and end-to-end connection relationship, can be more intuitively expressed. Therefore, there is a need to further abstract the resource information in the virtual resource pool into a logical network expressed by a graph. In the virtual resource pool, because there are a plurality of connection relations between physical nodes, a network topology relation can be established by a maximum flow method in graph theory, that is, in fig. 6, virtual link capacity between nodes represented by a dotted line is obtained by a physical resource topology through a maximum flow algorithm. And the quick combing of the logic relation between the nodes is effectively finished by adopting modes such as link combination and the like between the two virtual user nodes, thereby finishing the preliminary virtual network topology establishment work.
After network virtualization, an undirected graph G (V, E, B, H) is shown, where B represents the virtual link bandwidth capability between virtual nodes and H represents the minimum hop count of the physical links to which the virtual links map. In fig. 7, assuming that bandwidth or hop count requirements cannot be met between node No. 8 and node No. 9, the virtual network topology is shown by a dotted line.
In a specific embodiment, the purpose of virtual network topology optimization is to achieve an optimal balance between the number of available resources and the network efficiency. The evaluation indexes of the virtual network topology optimization are as follows:
The path length is defined as the distance d between two nodes i and j in the network i,j It is used to describe the number of edges on the shortest path between the two nodes. However, in the virtual network, the original multi-hop may be virtualized as one-hop representation, so that the attribute needs to be subjected to virtual network weighting, that is, the actual physical path length is weighted in the virtual network, so that the distance parameter in the virtual network is reflected more objectively and comprehensively; the expression of the virtual network weighted path length is as follows:
wherein the content of the first and second substances,
representing the physical path length in a logical link that is subordinate to virtual node i and virtual node j in the actual physical topology. Similarly, the average path length L also needs to be modified by virtual network weighting to
Wherein N is the number of virtual network nodes.
Evaluation index 2: virtual network weighting efficiency
In the process of calculating the network topology efficiency, the influence of the actual physical nodes is added. The traditional network topology evaluation index can be adopted when the distance calculation is carried out, but the distance between the nodes i and j is corrected to be the virtual network weighted path length
The virtual net weighted efficiency is modified into
Wherein N is the number of virtual network nodes.
Evaluation index 3: virtual network weighted betweenness
In an actual network, the betweenness refers to the number of times that all shortest paths in the network pass through the node of the element in the graph (if the element in the graph is a node, the betweenness refers to the betweenness of the node, and if the element in the graph is an edge, the betweenness of the edge). However, when calculating the betweenness in the virtual network, the betweenness is weighted by the actual virtual network
Replace the original d i,j By the correction method, the betweenness in the virtual network and the virtual network weighting betweenness may not be equal, because the actual physical hop count is different from the virtual network weighting coefficient.
The main task of virtual network topology optimization is to rationalize the network topology structure and simplify the topology structure on the premise of ensuring the network communication efficiency, and the purpose is to realize the optimal balance between the available resource number and the network efficiency. Therefore, if the topology structure can be optimized through the relevant attributes of the network topology, the whole virtual network is traversed from the aspect of the original pairwise connected topology, the link relation with relatively high efficiency is reserved, and the link relation with low efficiency is eliminated, so that the limited resources are more reasonably distributed to the virtual network. As shown in fig. 8, the steps of virtual network topology optimization are specifically as follows:
s201: initial virtual network information may be obtained through the established virtual resource pool first,
s202: calculating the weighted betweenness of all the virtual networks in the current virtual network according to a shortest path algorithm;
s203: arranging all edges in the virtual network from small to large according to the weighted betweenness of the corresponding virtual networks, and setting the communication link bandwidth of the first edge to be 0, namely disconnecting the communication link bandwidth;
s204: and calculating the virtual network weighting efficiency in the current virtual network according to the virtual network weighting path length, finishing the virtual network topology optimization if the virtual network weighting efficiency exceeds a threshold value, and otherwise returning to the step S202 to continue executing.
In a specific embodiment, the initial establishment of the virtual network refers to a state of the virtual resource pool after the topology of the virtual network is completed and before parallel traffic is transmitted in each virtual link. At this time, all nodes in the virtual network are not fully connected, and only the edge with higher bandwidth utilization rate is reserved. The virtual network allocates a fixed bandwidth, e.g., 2Mbps, to each edge when initially established, as shown in fig. 10.
The main task of the virtual network adaptive optimization is to continuously correct the network resource allocation by using historical statistical data after completing the initial establishment of the virtual network link structure, as shown in fig. 11, the specific steps are as follows:
s301: acquiring the total number of the distributable virtual resources of the current virtual network;
s302: calculating the minimum resource quantity for maintaining the virtual network, and sequentially endowing each link of the virtual network topology structure with equal communication resources;
s303: counting the parallel service transmission number of each virtual link in the current virtual network;
s304: and distributing different communication resources for each link according to the number of parallel service transmissions by using the residual resources in the link resource pool.
In the method, firstly, a minimum initialized resource quantity capable of guaranteeing communication is calculated and obtained for a constructed topological structure, equal communication resources are sequentially given to each link of the initialized topology, and each virtual link is supposed to obtain virtual resources with 2M bandwidth. And then, the number of the parallel services in the link is adaptively adjusted continuously, more communication resources are endowed to the high-value virtual link, and the statistical result is used as input and fed back to the resource control module, so that the reserved resource allocation method of the next batch is guided, namely the number of the reserved resources is in direct proportion to the number of the parallel services in the link.
The purpose of the adaptive adjustment strategy adopted in the embodiment is to ensure that both the normal service and the parallel mode can be ensured in the network. The virtual network link resources at the initialization moment are configured to meet the minimum networking requirement, then the residual resources are flexibly distributed based on the parallel service transmission number to gradually adapt to the current service condition, and finally the optimization of the whole virtual network is successfully realized.
In a specific embodiment, after the virtual network topological structure is optimized, networking requests of pairwise communication among multiple users are converted into requests of construction of multiple virtual links with indirect structures, and complexity of an initial networking problem is remarkably reduced. Each virtual link in the virtual network topology can be realized through a virtual link construction algorithm based on reinforcement learning. And finally, the construction of the whole virtual network is completed, the communication requirement of intercommunication between every two users is realized, and the utilization rate of network resources is obviously improved.
The physical network topology is represented by graph G (V, E), where i E V represents the Mesh node in the backbone network. Definition of
The neighbor node set of the node i is the set of all nodes in the communication range of the node i, and j represents the neighbor node of the node i. The edge ij epsilon E represents the wireless link from the node i to the node j. Suppose that the starting node of a certain service flow rho is s, the target node is d, and the bandwidth requirement is w thr . The obtained routing path needs to satisfy two types of QoS indexes at the same time, and meanwhile, in order to improve the utilization rate of the whole network resources, the load balance of the network needs to be considered in the required path. To solve this problem we propose the following: the bandwidth index is expressed as a constraint form, when the path residual bandwidth is greater than the bandwidth requirement, a path with lower time delay and higher residual bandwidth is selected from the path residual bandwidth, and when the path residual bandwidth is less than the bandwidth requirement, the service request is rejected; and for the indexes of time delay and residual bandwidth, a multi-objective optimization problem is proposed and modeled.
The problem of path selection is mapped to a Markov Decision Process (MDP) E = < X, a, P, R >, where X represents a set of states, a represents a set of actions, P represents a probability transition matrix, and R represents a reward function. The correspondence between the routing problem of the backbone network and the markov decision process is shown in table 3:
TABLE 3 correspondence of routing problem to Markov decision process
| Dynamic game model | Backbone net |
| State set S | Formed by each node in the graph G (V, E) |
| Action set A | Selecting neighbor nodes as next hop nodes |
| Transition probability P | Transition probability p =1 for selected actions |
| Prize r | Link parameter index d ij ，w ij ，e ij |
In this embodiment, the four-tuple of the markov decision process corresponding to the routing problem is known, and can be solved by using a reinforcement learning median iterative algorithm. Function of state value V π (x) Representing the maximum cumulative reward attainable by the strategy pi in the current state, the state-action value function Q π (x, a) represents the maximum accumulated reward that can be obtained by continuing through policy π after performing action a in the current state.
T steps V under accumulated reward π (x) Can defineIs the following formula:
under the condition of T-step accumulated reward
Can be defined as the following equation:
since the Markov process model is known, a full probability expansion can be performed:
in the formula (I), the compound is shown in the specification,
representing an instantaneous award, V representing a cumulative award for a total period of time (T-1 steps); p represents transition probability, and the transition state is considered to have occurred when the action is selected, i.e. P is 1, and unselected actions P are 0
For a given convergence threshold θ, the median iterative algorithm process is as follows:
the embodiment can award the instant prize
Instead, two types of indicators are used, and the action with the maximum weighted sum of the two types of indicators is selected when the action is selected. Assuming that the action set is a, in the state x, the action a is to select x' as the next hop node, and then the corresponding iterative update formula for obtaining the maximum path residual bandwidth is as follows:
when q is w (x,a)≥w thr And updating the time delay and the error rate index. Because the two types of indicators are updated in different ways, the accumulated award cannot be calculated by a single addition. The optimal action a 'for the next state x' is selected first:
q(x′,a′)＝min(q(x′,a * )) (6)
wherein q (x ', a') represents the normalized weighted average of the two types of optimization targets;
and then calculating the q value under the current state-action pair:
q w (x,a)＝min(w xx′ ,q w (x′,a′)) (7)
q d (x,a)＝d xx′ +q d (x′,a′) (8)
q (x, a) will be updated next * ) The value of (c):
qMax d ＝max(q d (x,a * )) (9)
when the q value matrix is converged, the updating is finished, and q can be obtained w (x,a)，q d (x,a)，q e (x, a), q (x, a) four q-value tables with state on the abscissa and action on the ordinate, each row in the table representing the optimal prize value available for selecting the respective action in the current state. q (x, a) is normalized for two types of optimization objectivesAnd transforming the weighted average value, sequentially selecting the minimum q value according to q (x, a) to obtain a strategy pi (namely routing output), and distributing path resources according to the strategy pi to realize virtual and real resource mapping.
The algorithm time complexity based on reinforcement learning is O (n) 2 ). Some heuristic algorithms are as follows: the ant colony algorithm and the like have great advantages in time complexity when the node magnitude is very large. However, in this embodiment, the backbone network adopts a distributed domain-divided network management architecture, and the global routing problem can be decomposed into sub-problems of routing problems in several domains, and the management architecture of this division and treatment method reduces the order of magnitude of each sub-problem node by a factor of about 30 nodes in each domain. In this case, the heuristic algorithm has too slow algorithm processing time due to too many iterations, and the time complexity is O (n) 2 ) The algorithm has greater advantages. According to the good distributed characteristic of the value iteration algorithm, the algorithm can be accelerated in a parallel computing mode, and therefore time complexity is reduced.
It should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.
Claims (9)
1. A virtual-real resource mapping method for virtual network construction in a multi-hop network is characterized in that: the method comprises the following steps:
s1: abstracting a physical network, and establishing an initialized virtual network topological structure;
s2: optimizing the virtual network topology by adopting a small world theory and through a virtual network evaluation index to obtain an optimized topology connection relation;
s3: sequentially endowing each link of the virtual network topology with equal communication resources, and distributing different communication resources for each link by the residual communication resources according to the parallel service transmission number;
s4: and (3) finishing end-to-end virtual link planning by adopting a path selection method based on reinforcement learning, and mapping virtual links among virtual network nodes into actual links.
2. The virtual-real resource mapping method for virtual network construction in a multi-hop network according to claim 1, wherein: step S1, the physical position of the node in the physical network is confirmed concretely, the connection relation among all the nodes is confirmed, and therefore the initialized virtual network topological structure is established in an abstract mode.
3. The virtual-real resource mapping method for virtual network construction in a multi-hop network according to claim 2, wherein: all the logic nodes in the initial virtual network are fully connected; after the virtual network is established, the physical resource pool is called, the resource information is abstracted and constructed in several modes including combination, deletion and filling, and the normalized output result is adopted to fill the virtual resource pool database.
4. The virtual-real resource mapping method for virtual network construction in a multi-hop network according to claim 3, wherein: abstracting resource information in a virtual resource pool into a logic network expressed by a graph, and establishing a network topology relationship by a maximum flow algorithm in a graph theory, namely acquiring virtual link capacity among virtual user nodes by physical resource topology through the maximum flow algorithm; and finishing the combing of the logic relationship between the virtual user nodes by adopting a link combination mode between the virtual user nodes so as to finish establishing a primary virtual network topology.
5. The virtual-real resource mapping method for virtual network construction in a multi-hop network according to claim 4, wherein: the virtual network evaluation index comprises a virtual network weighting path length, a virtual network weighting efficiency and a virtual network weighting betweenness.
6. The virtual-real resource mapping method for virtual network construction in a multi-hop network according to claim 5, wherein: the virtual network weighted path length means that the actual physical path length is weighted in the virtual network, and the expression is as follows:
in the formula (d) i,j Representing the distance d between a virtual user node i and a virtual user node j in a virtual network i,j The method is used for describing the number of edges on the shortest path between two virtual user nodes;
representing the physical path lengths in the logical links that are subordinate in the actual physical topology between virtual user node i and virtual user node j.
7. The virtual-real resource mapping method for virtual network construction in a multi-hop network according to claim 6, wherein: the expression of the virtual network weighting efficiency is as follows:
where N represents the number of virtual network nodes.
8. The virtual-real resource mapping method for virtual network construction in a multi-hop network according to claim 7, characterized in that: s2, optimizing the virtual network topology by adopting a small world theory and through the virtual network evaluation index, wherein the method specifically comprises the following steps:
s201: initial virtual network information is obtained through the virtual resource pool,
s202: calculating the weighted betweenness of all the edges in the current virtual network according to a shortest path algorithm;
s203: arranging all edges in the virtual network from small to large according to the weighted betweenness of the corresponding virtual networks, and setting the communication link bandwidth of the first edge to be 0, namely disconnecting the communication link bandwidth;
s204: and calculating the virtual network weighting efficiency in the current virtual network according to the virtual network weighting path length, finishing the virtual network topology optimization if the virtual network weighting efficiency exceeds a threshold value, and otherwise returning to the step S202 to continue executing.
9. The virtual-real resource mapping method for virtual network construction in a multi-hop network according to claim 8, wherein: and step S3, the specific steps are as follows:
s301: acquiring the total number of allocable virtual resources of the current virtual network;
s302: calculating the minimum resource quantity for maintaining the virtual network, and sequentially endowing each link of the virtual network topology structure with equal communication resources;
s303: counting the parallel service transmission number of each virtual link in the current virtual network;
s304: and distributing different communication resources for each link according to the parallel service transmission number by using the residual resources in the link resource pool.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110114947.5A CN112953761B (en) | 2021-01-26 | 2021-01-26 | Virtual-real resource mapping method for virtual network construction in multi-hop network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110114947.5A CN112953761B (en) | 2021-01-26 | 2021-01-26 | Virtual-real resource mapping method for virtual network construction in multi-hop network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112953761A CN112953761A (en) | 2021-06-11 |
| CN112953761B true CN112953761B (en) | 2022-10-25 |
Family
ID=76238313
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110114947.5A Active CN112953761B (en) | 2021-01-26 | 2021-01-26 | Virtual-real resource mapping method for virtual network construction in multi-hop network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112953761B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113726565B (en) * | 2021-08-24 | 2023-08-15 | 哈尔滨工程大学 | Method for designing monitoring information transmission topological structure in strip-shaped non-mobile network area |
| CN114268371B (en) * | 2021-11-02 | 2023-03-31 | 北京邮电大学 | Quantum channel resource allocation method and device and electronic equipment |
| CN116192720B (en) * | 2021-11-26 | 2024-06-04 | 中国联合网络通信集团有限公司 | Link optimization method and device, electronic equipment and storage medium |
| CN115484169A (en) * | 2022-09-09 | 2022-12-16 | 山石网科通信技术股份有限公司 | Method, device and system for constructing network topology structure |
| CN115941506A (en) * | 2022-09-29 | 2023-04-07 | 重庆邮电大学 | Multi-type service resource arrangement method based on strategy network reinforcement learning |
| CN115987794B (en) * | 2023-03-17 | 2023-05-12 | 深圳互联先锋科技有限公司 | Intelligent shunting method based on SD-WAN |
| CN116582451B (en) * | 2023-07-13 | 2023-11-21 | 三峡科技有限责任公司 | Communication operation method combining elastic communication network with virtualization technology |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104144135A (en) * | 2014-07-25 | 2014-11-12 | 电子科技大学 | Resource distribution method and survivability resource distribution method used for multicast virtual network |
| CN107566194A (en) * | 2017-10-20 | 2018-01-09 | 重庆邮电大学 | A kind of method for realizing the mapping of cross-domain virtual network network |
| CN110365514A (en) * | 2019-05-24 | 2019-10-22 | 北京邮电大学 | SDN multistage mapping method of virtual network and device based on intensified learning |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6664812B2 (en) * | 2016-05-10 | 2020-03-13 | 国立研究開発法人情報通信研究機構 | Automatic virtual resource selection system and method |
-
2021
- 2021-01-26 CN CN202110114947.5A patent/CN112953761B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104144135A (en) * | 2014-07-25 | 2014-11-12 | 电子科技大学 | Resource distribution method and survivability resource distribution method used for multicast virtual network |
| CN107566194A (en) * | 2017-10-20 | 2018-01-09 | 重庆邮电大学 | A kind of method for realizing the mapping of cross-domain virtual network network |
| CN110365514A (en) * | 2019-05-24 | 2019-10-22 | 北京邮电大学 | SDN multistage mapping method of virtual network and device based on intensified learning |
Non-Patent Citations (2)
| Title |
|---|
| 面向收益最大化的虚拟网跨域映射策略;江逸茗等;《工程科学与技术》;20180321(第02期);第118-125页 * |
| 面向边云混合环境的虚拟资源分配与管理研究;付霄元;《中国优秀博硕士学位论文全文数据库（博士） 信息科技辑》;20210115(第01期);第I139-5页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112953761A (en) | 2021-06-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112953761B (en) | Virtual-real resource mapping method for virtual network construction in multi-hop network | |
| CN111010294B (en) | Electric power communication network routing method based on deep reinforcement learning | |
| CN106411749B (en) | A kind of routing resource for software defined network based on Q study | |
| CN110365514A (en) | SDN multistage mapping method of virtual network and device based on intensified learning | |
| CN104202183B (en) | The method and apparatus that a kind of solution SDN stream ranks configuration conformance updates | |
| CN110191148A (en) | A kind of statistical function distribution execution method and system towards edge calculations | |
| CN110365568A (en) | A kind of mapping method of virtual network based on deeply study | |
| WO2023024219A1 (en) | Joint optimization method and system for delay and spectrum occupancy in cloud-edge collaborative network | |
| CN107196806B (en) | Topological proximity matching virtual network mapping method based on sub-graph radiation | |
| CN111010295A (en) | SDN-MEC-based power distribution and utilization communication network task migration method | |
| CN110061881A (en) | A kind of energy consumption perception virtual network mapping algorithm based on Internet of Things | |
| Lu et al. | A dynamic and collaborative multi-layer virtual network embedding algorithm in SDN based on reinforcement learning | |
| CN114301794B (en) | LEOMEO double-layer satellite constellation-oriented interlayer link topology design method | |
| CN106817256A (en) | A kind of distributed system network resource operation management reliability method for improving | |
| Nguyen et al. | Efficient virtual network embedding with node ranking and intelligent link mapping | |
| Hu et al. | Dynamic task offloading in MEC-enabled IoT networks: A hybrid DDPG-D3QN approach | |
| Zheng et al. | Minimizing the latency of embedding dependence-aware SFCs into MEC network via graph theory | |
| Alvizu et al. | Machine-learning-based prediction and optimization of mobile metro-core networks | |
| CN103618674A (en) | A united packet scheduling and channel allocation routing method based on an adaptive service model | |
| WO2023222061A1 (en) | Intent-driven wireless network resource conflict resolution method and apparatus | |
| Zhang | Reliable virtual network mapping algorithm with network characteristics and associations | |
| Pan et al. | Satellite network load balancing strategy for SDN/NFV collaborative deployment | |
| CN115941506A (en) | Multi-type service resource arrangement method based on strategy network reinforcement learning | |
| Teja Sree et al. | Mobile Edge Computing Architecture Challenges, Applications, and Future Directions | |
| Mao et al. | Ant colony based online learning algorithm for service function chain deployment |
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 |