CN105337834A - Mapping algorithm adopted in wireless network virtualization environment - Google Patents

Abstract

The invention discloses a mapping algorithm adopted in a wireless network virtualization environment. The mapping algorithm aims to solve the problem of existing wireless network virtualization mapping algorithms that public resources including bandwidth and frequency spectrum which can be shared between different basic facilities within a certain range and private resources including power and computing capacity which only belong to fixed hardware equipment and can not be shared are not well defined and distinguished. The mapping algorithm is capable of realizing joint optimization of power resources and frequency spectrum resources, interaction between power resources and frequency spectrum resources is comprehensively considered, the cost function of virtual network mapping is established by defining unit resource cost for reflecting the physical network loading level, mapping cost minimization is taken as the optimization object, physical network load balancing is achieved, path separation is adopted during virtual network mapping, and the utilization efficiency of physical resources is improved and the success rate of virtual network mapping is increased.

Description

Mapping algorithm in wireless network virtualization environment

Technical Field

The invention belongs to the research of mapping algorithms in a wireless network virtualization environment, belongs to the resource allocation category, and particularly relates to the research of resource allocation combining sharable public resources (such as frequency spectrums and the like) and unshared private resources (such as power and the like).

Background

With the increasingly mature of various wireless communication technologies and the mass emergence of diversified mobile services, a wireless network presents a diversified form of intensive deployment, various services and coexistence of heterogeneous networks in the future. In a complex network environment, the compatibility of multiple wireless network technologies, the selection of users to different wireless access networks, the switching between heterogeneous networks and other problems are new challenges for the development of wireless networks. The wireless network virtualization technology provides an effective management mode for the heterogeneous wireless network, and coexistence and fusion of the heterogeneous wireless network are realized through abstraction and uniform representation of network resources, resource sharing and efficient multiplexing. The wireless network virtualization can decouple complex and diverse network management and control functions from hardware, and extract an upper layer to perform unified coordination and management, so that the network management cost is reduced, and the network management and control efficiency is improved. Centralized control allows service providers without wireless network infrastructure to provide differentiated services to subscribers as well.

The biggest difference between network virtualization environments and the current Internet is that the current Internet consists of only one role of Internet service provider, whereas network virtualization environments consist of two different roles, i.e. infrastructure providers (InP) and Service Providers (SP). The infrastructure provider is responsible for the construction, operation and maintenance of physical network facilities, virtualizes and abstracts physical resources, monitors and manages the physical network resources, can establish virtual resources as required, and provides physical resources such as physical links, programmable routers and the like for the service provider; and the service provider leases resources from one or more infrastructure providers to build its own virtual network, and deploys a customized protocol to provide end-to-end transport services for virtual network users. Under the network virtualization environment, based on a design mechanism that an infrastructure provider is separated from a service provider, multiple heterogeneous virtual networks can share bottom-layer public physical network resources, so that the construction and operation and maintenance costs are shared, and the high maintenance cost of physical network facilities is reduced.

The resource allocation work for establishing the virtual network is called virtual network mapping, and is a key problem of network virtualization research. Virtual Network Mapping (VNM), or Virtual Network Embedding (VNE), is a request (including elements such as topology, resource demand, and location restriction) for establishing a virtual network (virtual network, VN) according to a Service Provider (SP), and on the premise of not breaking the constraint of underlying resources, how to map multiple virtual networks with different topologies into a shared infrastructure at the same time, and ensure the effective utilization of the underlying resources. In order to map the entire virtual network, the virtual network mapping needs to find a suitable physical node in the physical network for the virtual node and needs to satisfy the resource request of the node, and at the same time, a physical path is found on the physical network for the virtual link and the remaining resources of each path need to satisfy the request of the virtual link mapping. The underlying physical network is composed of multiple infrastructure providers in a cooperative and competitive relationship, requiring cooperation between the multiple infrastructure providers if the virtual network is mapped across the multiple infrastructure providers.

When virtual network mapping is performed, if multiple problems such as resource constraint, admission control, online request, topology diversity and the like need to be considered at the same time, the virtual network mapping problem becomes extremely complex, and the optimization problem caused by the problem is generally NP-hard. Even ignoring several of these factors, the virtual network mapping problem is still complex. The virtual network mapping problem, which considers only node and link resource constraints in the most general case, is NP-hard. Even if only the link resource constraint is considered, the virtual network mapping problem meeting the bandwidth constraint can be reduced to the path separation problem under the condition that the virtual node is mapped, and the problem is still an NP difficult problem. Because the virtual network mapping problem is generally NP-hard and cannot find the optimal solution within polynomial time, when solving the virtual network mapping problem, the search space is often reduced by simplifying the problem or adding assumptions to obtain an approximately optimal solution.

At present, virtual network mapping algorithm research faces many challenges, and particularly, the problems faced in a wireless network environment are more prominent. The wireless network environment has self architectural characteristics and specific constraints, and the design of the optimization algorithm is more worthy of further intensive research. Lv in (see literature: XiaoLv, AoXiong, Shunlizhang, and XuueSongQiu, VCG-basedBandwidth HallocationSchemevorNetworkVirtulation [ C ], incedingsof 2012 IEEESympossingcomputerisation communication (ISCC), pp.744-749, July2012.) proposes a resource allocation mechanism based on VCG (Vickrey-Clarke-Grove) mechanism that maximizes the total benefit of SP by suppressing SP selfishness. Then, a Q learning bidding selection algorithm is designed, so that the SP obtains an optimal bidding strategy; yun proposes a mapping algorithm for wireless multi-hop networks, which effectively utilizes physical layer resources (see documents: Donggyu Yun, Jungseulok, Bongjihnshin Shin, Soobumpark, and Yungyi, Embelling of virtual network requestrequestSoverdustwireless multi-hop networks [ J ], computer networks, vol.57, pp.1139-1152, April 2013.). Fu (ref: FuF, kozatuc. "wireless network virtual availability au (al)" C, ieee info com2010.) proposes a system framework of a virtual wireless network, in which a network controller (network kopersor: NO) is responsible for resource allocation, SPs bid for network resources according to the behavior of users (EU), and NO auctions by executing Vickrey-Clarke-groves (vcg) mechanism, models it as a steinberg game and finds an optimal solution.

In a wireless environment, the research on the mapping algorithm is very little, and the mapping algorithm in a wired environment is not applicable to a wireless environment. Firstly, in the existing mapping algorithm for wireless network virtualization, public resources such as bandwidth and spectrum, which can be shared and allocated among different infrastructures within a certain range, and private resources such as power and computing power, which only belong to fixed hardware devices and cannot be scheduled and utilized mutually, are not clearly defined and distinguished, and differences, complementarity and alternatives among resources with different properties, cost required for resource allocation and adverse effects (such as interference) generated at the same time, and comprehensive effects on system performance are not considered in the resource allocation. For example: meanwhile, when a certain type of resource (such as power) is in shortage and another type of resource (such as spectrum) is in surplus, on the premise of meeting the service requirement (such as capacity), how to balance the configuration scheme and how to realize flexible resource joint optimization decision through what data scheduling is also an urgent problem to be solved.

Secondly, in the construction process of the virtual network, the load pressure of the physical nodes and the load pressure of the links need to be considered at the same time, the occupation of the bottom layer physical resources by each virtual network is coordinated uniformly and evenly, and the phenomenon that the excessive use of some local resources is caused because only the minimum requirement of the resources is considered and the reasonable use of the resources is not considered when the virtual network is mapped is avoided, and other local resources are possibly idle and cannot be used. Efficient resource utilization involves load balancing issues.

Wireless networks are then largely distinguished from wired networks in that wireless transmission links are susceptible to environmental effects that attenuate the signal. Because the wireless link has a broadcast property, a radio signal sent by one node can be acquired by other nodes, and therefore, the influence of an interference signal in the network on the mapping performance needs to be fully considered in the mapping process of the virtualization of the wireless network.

Finally, for a virtual request, under the condition of meeting the speed requirement, one virtual link is mapped to a plurality of physical paths, and compared with a method for mapping one virtual link to one physical path, the method can undoubtedly bring more flexibility to the scheduling and configuration of the virtual network, thereby reducing or even eliminating the constraint condition of the virtual network mapping problem and meeting the virtual network requests as much as possible.

Disclosure of Invention

Aiming at the following existing defects, a mapping algorithm under a wireless network virtualization environment is provided, which realizes the efficient utilization of resources and simultaneously improves the probability of successful mapping of virtual requests. . The technical scheme of the invention is as follows: a mapping algorithm in a wireless network virtualization environment, comprising the steps of:

step 1): constructing a wireless network virtualization mapping model;

step 2): calculating available resources and extended resources in the current network according to a wireless network virtualization mapping model, determining an interference model, and then constructing a cost function of virtual network mapping by defining a resource unit price as an optimization target, wherein the target is to minimize mapping cost, and the mapping cost comprises the cost of node mapping and the cost of link mapping;

step 3): the method comprises the steps of firstly realizing optimized mapping of virtual nodes, then carrying out virtual link mapping to find a path based on the positioned virtual nodes, synthesizing a path separation algorithm, and optimizing the flow request of the virtual link by using the principle of minimum mapping cost when a virtual network is constructed.

Further, the step 1 of constructing the wireless network virtualization mapping model includes a physical network model and a wireless virtual network request model, where the physical network model is: using an undirected graph G with weights^S＝(N^S，L^S) Represents the underlying physical network, where N^SRepresenting a set of underlying physical nodes, L^SRepresenting a set of underlying links; for each physical node n^S∈N^SIn other words, the node power p (n) is also included^S) And the location information loc (n) of the node^S) For each physical link l^S∈L^SIn other words, the link bandwidth information b (l) is included^S)，b(l^S) Represents the physicsA maximum value of available bandwidth for the link; furthermore, with p^STo represent a physical path, P, in a physical network^SRepresenting a set of physical paths; the wireless virtual network request model is as follows: defining an undirected graph G with weight^V＝(N^V,L^V) As a request for a wireless virtual network, N^VRepresenting a set of virtual network requesting nodes, each node n^V∈N^VContaining location information loc (n)^V)，L^VSet of links representing virtual network requests, for each virtual link l^V∈L^VIt contains the transmission rate R (l) that the virtual link needs to reach^V) The physical network needs to allocate a certain bandwidth and power for the virtual request.

Further, the objective function in step 2 is

$\min \left(C\right(P^S\left)\right)=\min (\underset{n^V\→n^S,n^S\∈P^S}{\Σ}\mathrm{\α}\left(n^S\right)p\left(n^V\right)+\underset{l^V\→l^S,l^S\∈P^S}{\Σ}\mathrm{\β}\left(l^S\right)b\left(l^V\right))$

s.t.p(n^V)≤A_N(n^S)

dis(loc(n^V),loc(n^S))≤D^V

$b\left(l^V\right){\log }_2(1+\frac{p\left(n^V\right)G\left(l^V\right)}{{\mathrm{\σ}}^2+{\mathrm{\Σ}}_{n^S\∈N^S\backslash n^V\→n^S}p\left(n^S\right)G\left(l^S\right)})\≥R\left(l^V\right)$

Wherein,A_N(n^S) And A_L(l^S) Representing the available resources of the node and link, respectively, the calculation is as follows:

${\mathrm{Capacity}}_N\left(n^S\right)=A_N\left(n^S\right)=p\left(n^S\right)-\underset{\∀n^V\→n^S}{\Σ}p\left(n^V\right),{\mathrm{Capacity}}_L\left(l^S\right)=A_L\left(l^S\right)=b\left(l^S\right)-\underset{\∀l^V\→l^S}{\Σ}b\left(l^V\right)$

G(l^V) Representing the channel gain, σ, of the mapped physical link²Representing white gaussian noise in the channel,representing other link pairsThe interference generated by the link.

Further, step 3, the path found by virtual link mapping is found by using a K shortest path algorithm.

Further, the step 3 implements optimized mapping of the virtual nodes, including the steps of: (1) firstly, a physical node set which can be mapped in respective range is found for two virtual nodes, namely dis (loc (n) is satisfied^V),loc(n^S) D (D represents the mapping radius of the virtual node in the virtual network request, dis (loc (n))^V),loc(n^S) Represents a virtual node n^STo physical node n^VDistance of (d); 2) and respectively calculating the expansion resources of the nodes in the two physical node sets, and selecting the node with the maximum expansion resource to map the virtual node.

6. The mapping algorithm in a wireless network virtualization environment according to claim 1,

the link mapping process of the step 3 is realized by the following steps:

(1) respectively calculating the interference coefficient of each link in the physical network;

(2) distributing the node mapping cost to the links corresponding to the physical nodes, namely, the weight of each physical link is the sum of the resource cost of the corresponding node and the link interference coefficient, and finding K shortest paths P between two physical nodes by adopting a K shortest path algorithm₁,P₂,...P_kThe shortest path comprehensively considers the size of resources and the interference between links;

(3) first, the speed request passes through the path P with lowest cost₁If the total data Δ R transmitted has not yet reached the requested data size and is on path P₁Cost of upstream transmission and on path P₂When the cost of the upper transmission is equal, the data is shunted to the path P₂Continuously increasing the transmitted data if DeltaR < R (l)^V) And is Δ C₂＝ΔC₃Then the data is shunted to path P₃By analogy, data can be shunted to the pathP_kIf Δ R < R (l) at this time^V) If so, link mapping fails; at any time during this process, if Δ R ≧ R (l)^V) Then the mapping is successful and the algorithm is terminated.

The invention has the following advantages and beneficial effects:

the invention provides a self-adaptive mapping algorithm comprehensively considering the joint matching of frequency spectrum resources and power resources, which constructs a cost function of virtual network mapping by defining the unit resource cost reflecting the load level of a physical network, and realizes the self-adaptive efficient resource balanced utilization by taking the minimum mapping cost as an optimization target. In the process of link mapping, a link segmentation algorithm is integrated, interference among wireless links is considered, and when a virtual network is constructed, a flow request of the virtual link is segmented on the principle of minimum mapping cost, so that the utilization efficiency of physical resources is improved, and the probability of successful mapping of the virtual network is increased.

Drawings

FIG. 1 is a diagram of the present invention providing a preferred network virtualization architecture;

fig. 2 is a mapping diagram of path splitting according to the present invention.

Detailed Description

The invention is further described below with reference to the accompanying drawings:

as shown in fig. 1, which depicts two virtual networks that may be heterogeneous, VN1, VN2 are managed by service providers SP1 and SP2, respectively, and SP1 builds virtual network VN1 using the resources of two different basic service providers InP1 and InP2, as in fig. 1. SP2, on the other hand, deploys the virtual network VN2 in conjunction with child virtual network resources provided from the infrastructure providers InP1 and service provider SP 1. The mapping problem of network virtualization is to solve how InP1 and InP2 allocate reasonable resources according to VN1 and VN2 requests.

1 network model

1.1 Wireless physical network model

Using an undirected graph G with weights^S＝(N^S，L^S) Representing the underlying physical network, consider a one-slot system. Wherein N is^SRepresenting a set of underlying physical nodes, L^SRepresenting a set of underlying links. For each physical node n^S∈N^SIn other words, the node power p (n) is also included^S) And the location information loc (n) of the node^S). For each physical link l^S∈L^SIn other words, the link bandwidth information b (l) is included^S)，b(l^S) Representing the maximum value of the bandwidth available for the physical link. Furthermore, with p^STo represent a physical path, P, in a physical network^SRepresenting a collection of physical paths.

1.2 Wireless virtual network request model

Each virtual network request arrives dynamically, stays on and then leaves when the service is completed. Defining an undirected graph G with weight^V＝(N^V,L^V) As a request for a wireless virtual network. N is a radical of^VRepresenting a set of virtual network requesting nodes, each node n^V∈N^VContaining location information loc (n)^V)，L^VSet of links representing virtual network requests, for each virtual link l^V∈L^VIt contains the transmission rate R (l) that the virtual link needs to reach^V) The physical network needs to allocate a certain bandwidth and power for the virtual request.

$b\left(l^V\right){\log }_2(1+\frac{p\left(n^V\right)G\left(l^V\right)}{{\mathrm{\&sigma;}}^2+{\mathrm{\&Sigma;}}_{n^S\&Element;N^S\backslash n^V\&RightArrow;n^S}p\left(n^S\right)G\left(l^S\right)})\&GreaterEqual;R\left(l^V\right)---\left(1\right)$

P (n) in the formula^V) And b (l)^V) Respectively representing the power size allocated to the node and the bandwidth size allocated to the link in the mapping process，G(l^V) Representing the channel gain, σ, of the virtual link²Representing white gaussian noise in the channel,indicating the interference that other links cause to the link.

2 System analysis

2.1 Wireless network available resources

1) Available resources of a node

Since the underlying physical node may carry one or more virtual nodes, n for each node^S∈N^SThe available resources are

${\mathrm{Capacity}}_N\left(n^S\right)=A_N\left(n^S\right)=p\left(n^S\right)-\underset{\&ForAll;n^V\&RightArrow;n^S}{\&Sigma;}p\left(n^V\right)---\left(2\right)$

2) Resources available to the link

Since the underlying physical links may also carry one or more virtual links, for each physical link l^S∈L^SThe available resources are

${\mathrm{Capacity}}_L\left(l^S\right)=A_L\left(l^S\right)=b\left(l^S\right)-\underset{\&ForAll;l^V\&RightArrow;l^S}{\&Sigma;}b\left(l^V\right)---\left(3\right)$

2.2 extended resources of a node

In the mapping process, a physical node mapping with a large amount of available resources is generally selected preferentially, which may cause a situation that a link connected to the physical node mapping has few available resources, resulting in a high resource cost in a link mapping stage or failing to find a suitable link mapping. The invention thus defines the extended resource of a physical node as the sum of the power resource of the node and the available bandwidth resources of all links connected to the node, i.e. the sum of the power resources of the node and the available bandwidth resources of all links connected to the node

$A_N\left(n^S\right)+{\mathrm{\&Sigma;}}_{l\&Element;L\left(n\right)}^S{A}_L\left(l^S\right)---\left(4\right)$

Wherein L isⁿFor a set of links through node n, the more extended resources of a physical node, the preference for that node mapping.

2.3 interference model

The invention considers the interference between links in a physical network aiming at a wireless network environment, and only considers the interference between adjacent links for one physical link to define a link interference coefficient

$d_I\left(l\right)=\frac{d_l+1}{A_L\left(l^S\right)}---\left(5\right)$

Wherein d is_lNumber of links that may interfere with other adjacent links, link interference factor and d_lIs in direct proportion. The link weight needs to be marked in the link mapping stage to find the shortest path, the link interference coefficient is used as a part of the coefficient, the shortest path algorithm ensures that the selected path interference is small, and simultaneously the available resources of the link are more, which is consistent with the aim of minimizing the mapping cost.

2.4 virtual network mapping problem

When a wireless virtual network request arrives, the physical network must decide whether to accept the request, if so, the physical network must select a proper node and path from the nodes and paths of the physical network, and when the wireless virtual network finishes the service departure, the occupied resources are released.

For each virtualized resource allocation of the wireless network, use one G^VTo G^SMapping of subsets, Map G^V→G^S(N^*,P^*) Wherein

1) node mapping Map^N:N^V→N^*

For a virtual network request, each virtual node is mapped to a different physical node, i.e. each virtual node of the same virtual network request monopolizes a physical node in the underlying network. Notably, if multiple virtual nodes are owned by different virtual network requests, they may coexist on the same physical node in the underlay network. Because the node resource limitation comprises CPU capacity, power resource size, memory size, disk space, geographical position and the like, the invention only considers two factors of the power resource size and the geographical position limitation.

For then^v→n^SThe node mapping process needs to satisfy a certain limiting condition, the size of the power resource allocated to the virtual node cannot be larger than that of the underlying physical node, and meanwhile, the node requested by the wireless virtual network cannot exceed the communication range of the physical node from the node, namely

$p\left(n^V\right)\&le;A_N\left(n^S\right),(\&ForAll;n^v\&RightArrow;n^S)---\left(6\right)$

dis(loc(n^V),loc(n^S))≤D^V(7)

Wherein D is^VThe representation is the mapping radius of the virtual node in the virtual network request.

2) Link Map^L:L^V→P^*

For a virtual request, in the link mapping stage, one virtual link can be mapped to one physical path one to one, the mapping algorithm has large resource constraint, and if one virtual link can be mapped to a plurality of physical paths under the condition of meeting the bandwidth requirement, more flexibility can be undoubtedly brought to the scheduling and configuration of the virtual network, so that the constraint condition of the virtual network mapping problem is reduced or even eliminated, and the virtual network request as much as possible is met. In the invention, a virtual link is mapped to a plurality of physical paths through shunting, and only bandwidth factors are considered on the link.

For theP^SThe method is characterized in that a set of mapped physical paths is shown, a certain limiting condition also needs to be met in the link mapping process, and the size of the bandwidth allocated to the virtual link cannot be larger than the available bandwidth of the physical link.

$b\left(l^V\right)\&le;A_L\left(l^S\right),(\&ForAll;l^V\&RightArrow;P^S,l^S\&Element;P^S)---\left(8\right)$

2.5 cost function

In the virtual network mapping process, in order to ensure efficient utilization of resources, load pressures of physical nodes and links need to be considered at the same time, and the problem that part of the physical nodes or physical links are overloaded and other nodes or links are in a light load condition, which is the load balancing problem to be considered, is avoided.

The invention constructs the cost function of virtual network mapping by defining the unit resource cost (namely the unit resource price) reflecting the resource pressure, takes the minimum mapping cost as an optimization target, and supports the multipath mapping of the virtual link to effectively realize the balanced utilization of the resources. In general, the more resources available, the lower the unit price of the resource, and vice versa, the present invention assumes that the price per unit resource is in an inverse relationship with the available resources.

For any physical path, the cost of transmitting a signal over that path includes the cost of power required by the node and the cost of bandwidth required by the link.

$Cost\left(P^S\right)=C\left(P^S\right)=\underset{n^V\&RightArrow;n^S,n^S\&Element;P^S}{\&Sigma;}\mathrm{\&alpha;}\left(n^S\right)p_{n^S}^{n^V}+\underset{l^V\&RightArrow;l^S,l^S\&Element;P^S}{\&Sigma;}\mathrm{\&beta;}\left(l^S\right)b_{n^S}^{n^V}---\left(9\right)$

$\mathrm{\&alpha;}\left(n^S\right)=\frac{1}{A_N\left(n^S\right)},\mathrm{\&beta;}\left(l^S\right)=\frac{1}{A_L\left(l^S\right)}$

α (n)^S) Denoted as node n^SThe price of the unit power resource,to representPower allocated to the node during the mapping process, β (l)^S) Represents a link l^SThe price per unit of the bandwidth resource,indicating the amount of bandwidth allocated for the link during the mapping process.

2.6 objective function

The solution of the virtual network mapping problem is to map a virtual network with virtual node and virtual link constraints to an actual physical network through a certain virtual network mapping algorithm, wherein the virtual node is mapped to a physical node, the virtual link is mapped to a physical path, and the constraints of the virtual node and the virtual link on resource requirements need to be met. The resources referred to herein are concerned with the available power of the node and the bandwidth resources of the link.

The invention mainly aims at minimizing resource cost, and provides a novel wireless virtual network mapping algorithm under the condition of meeting a virtual request, namely, an objective function is

$\min \left(C\right(P^S\left)\right)=\min (\underset{n^V\&RightArrow;n^S,n^S\&Element;P^S}{\&Sigma;}\mathrm{\&alpha;}\left(n^S\right)p\left(n^V\right)+\underset{l^V\&RightArrow;l^S,l^S\&Element;P^S}{\&Sigma;}\mathrm{\&beta;}\left(l^S\right)b\left(l^V\right))$

s.t.p(n^V)≤A_N(n^S)

dis(loc(n^V),loc(n^S))≤D^V(10)

$b\left(l^V\right)\&le;A_L\left(l^S\right),(\&ForAll;l^V\&RightArrow;P^S,l^S\&Element;P^S)$

$b\left(l^V\right){\log }_2(1+\frac{p\left(n^V\right)G\left(l^V\right)}{{\mathrm{\&sigma;}}^2+{\mathrm{\&Sigma;}}_{n^S\&Element;N^S\backslash n^V\&RightArrow;n^S}p\left(n^S\right)G\left(l^S\right)})\&GreaterEqual;R\left(l^V\right)$

Since the objective function cannot find an optimal solution, a mapping algorithm is determined according to the objective function, and mapping cost is minimized as much as possible. The mapping algorithm can fully consider the balanced utilization of the underlying physical network when the virtual network is constructed, reduce the physical link consumption of the virtual network, and reduce the mapping bearing capacity of a single physical node and a link, thereby balancing the load in the physical network. The physical network resource is optimized according to the principle of balanced utilization of the physical network resource and minimum mapping cost, and the utilization efficiency of the physical resource is improved.

3 mapping algorithm basic steps

In the mapping algorithm provided by the invention, firstly, a physical node mapping is selected for each virtual node according to the geographical position and the size of node expansion resources, then, the node mapping cost in the path is distributed to the link corresponding to the physical node, the mutual interference among the links is considered, K paths are found by using a K shortest path algorithm, and the path separation algorithm is combined to carry out optimization on the principle of minimum mapping cost. For example, fig. 2 is a schematic diagram of path separation, where a numeral represents a link rate, and a rate request is shunted by using a path separation method, so that the physical resource utilization efficiency is improved, and the successful mapping probability of a virtual network is improved. In the present invention, the wireless virtual network request is exemplified by a data flow from a source node to a destination node.

3.1 node mapping procedure

(1) Firstly, a physical node set which can be mapped in respective range is found for two virtual nodes, namely dis (loc (n) is satisfied^V),loc(n^S))≤D；

(2) And respectively calculating the expansion resources of the nodes in the two physical node sets, and selecting the node with the maximum expansion resource to map the virtual node.

3.2 Link mapping procedure

(1) And respectively calculating the interference coefficient of each link in the physical network.

(2) Distributing the node mapping cost to the links corresponding to the physical nodes, namely, the weight of each physical link is the sum of the resource cost of the corresponding node and the link interference coefficient, and finding K shortest paths P between two physical nodes by adopting a K shortest path algorithm₁,P₂,...P_kThe shortest path comprehensively considers the size of the resource and the interference between the links.

(3) First, the speed request passes through the path P with lowest cost₁Following path P₁Addition of data to be transmitted, path P₁The increase of resources consumed and the increase of cost per unit resource lead to the increase of the path P₁The total cost of the upper transmission increases if the total data transmitted Δ R has not yet reached the requested data size (Δ R < R (l)^V) And on path P)₁Cost of upstream transmission and on path P₂Equal cost of uplink transmission (Δ C)₁＝ΔC₂) Then the data is shunted to path P₂Continuously increasing the transmitted data if DeltaR < R (l)^V) And is Δ C₂＝ΔC₃Then the data is shunted to path P₃By analogy, data can be shunted to path P_kIf Δ R < R (l) at this time^V) Then the link mapping fails. At any time during this process, if Δ R ≧ R (l)^V) Then the mapping is successful and the algorithm is terminated.

4 algorithm analysis

This section mainly develops analysis for the path separation algorithm at the link mapping stage. Suppose to be at P₁The rate of uplink transmission is Δ R₁Path P₁Each physical link l_1iAll transmission rates above are Δ R₁When P is present₁The ith link rate of (1) is Δ R₁When the temperature of the water is higher than the set temperature,

${\mathrm{\&Delta;R}}_1={\mathrm{\&Delta;b}}_{1i}{\log }_2(1+\frac{{\mathrm{\&Delta;p}}_{1i}{G}_{1i}}{{\mathrm{\&sigma;}}^2+{\mathrm{\&Sigma;}}_{n^S\&Element;N^S\backslash n_{1i}}p\left(n^S\right)G\left(l^S\right)}),(\&ForAll;n_{1i}\&Element;P_1,l_{1i}\&Element;P_1)---\left(11\right)$

${\mathrm{\&Delta;C}}_{11}=(\frac{1}{A_N\left(n^S\right)}{\mathrm{\&Delta;p}}_{1i}+\frac{1}{A_L\left(l^S\right)}{\mathrm{\&Delta;b}}_{1i}),({\mathrm{\&Delta;p}}_{1i}\&le;A_N(n^S),{\mathrm{\&Delta;b}}_{1i}\&le;A_L(l^S\left)\right)---\left(12\right)$

wherein G is_1iFor the channel gain, σ, of the link from the ith node to the next node on the first path²Is the noise on that channel. For a certain link, the link gain and the received interference can be calculated, and then the resource cost mapped to the certain link is obtained as a function of the power according to the two formulas.

$\begin{array}{c}{\mathrm{\&Delta;C}}_{11}=\frac{1}{A_N\left(n^S\right)}{\mathrm{\&Delta;p}}_{1i}+\frac{1}{A_L\left(l^S\right)}{\mathrm{\&Delta;b}}_{1i}\\ =\frac{1}{A_N\left(n^S\right)}{\mathrm{\&Delta;p}}_{1i}+\frac{1}{A_L\left(l^S\right)}.\frac{{\mathrm{\&Delta;R}}_1}{{\log }_2(1+\frac{{\mathrm{\&Delta;p}}_{1i}{G}_{1i}}{{\mathrm{\&sigma;}}^2+{\mathrm{\&Sigma;}}_{n^S\&Element;N^S\backslash n_{1i}}p\left(n^S\right)G\left(l^S\right)})},(\&ForAll;n_{1i}\&Element;P_1,l_{1i}\&Element;P_1)\end{array}---\left(13\right)$

Obviously, the cost function is a concave function, and the derivative of equation (13) is derived, and the point where the derivative is zero is the point where the cost is the minimum. At this time, path P₁The required power of the physical node i is delta p'_1iThen calculates the physical link l_1iThe bandwidth required on is the bandwidth size Δ b'_1iTo obtain

$min\left({\mathrm{\&Delta;C}}_{11}\right)=\frac{1}{A_N\left(n^i\right)}{\mathrm{\&Delta;p}}_{1i}^{\&prime;}+\frac{1}{A_L\left(l^i\right)}{\mathrm{\&Delta;b}}_{1i}^{\&prime;}---\left(14\right)$

At this time, the path P can be obtained₁The rate of uplink transmission is Δ R₁The cost required for the data of (1) is Δ C₁

${\mathrm{\&Delta;C}}_1=\underset{n_{1i}\&Element;P_1,l_{1i}\&Element;P_1}{\&Sigma;}{\mathrm{\&Delta;C}}_{1i}=\underset{n^i\&Element;N\left(P_1\right)}{\&Sigma;}\frac{1}{A_N\left(n^i\right)}{\mathrm{\&Delta;p}}_{1i}^{\&prime;}+\underset{l^i\&Element;P_1}{\&Sigma;}\frac{1}{A_L\left(l^i\right)}{\mathrm{\&Delta;b}}_{1i}^{\&prime;}---\left(15\right)$

Similarly, the result is at path P₂The rate of uplink transmission is Δ R₁The cost required for the data of (1) is Δ C₂

${\mathrm{\&Delta;C}}_2=\underset{n_{1i}\&Element;P_2,l_{1i}\&Element;P_2}{\&Sigma;}{C}_{2i}=\underset{n^i\&Element;N\left(P_2\right)}{\&Sigma;}\frac{1}{A_N\left(n^i\right)}{\mathrm{\&Delta;p}}_{2i}^{\&prime;}+\underset{l^i\&Element;P_2}{\&Sigma;}\frac{1}{A_L\left(l^i\right)}{\mathrm{\&Delta;b}}_{2i}^{\&prime;}---\left(16\right)$

Comparison of Δ C₁And Δ C₂If Δ C₁＜ΔC₂Then choose to continue on path P₁Up transmission, when the path P is on₁Up to Δ R₁A certain amount of resources are consumed, and the size of the network resources and the corresponding price of the unit resources need to be updated. When node i consumes Δ p_1iAfter the power of (A), the available power resource is A_N(nⁱ)-Δp′_1iThe price per unit power of node i isWhen link l consumes Δ b_1iAfter the bandwidth of (2), the available bandwidth resource is A_L(lⁱ)-Δb′_1iThe price per bandwidth resource of the link isTherefore, the conditions for the split flow are

$\begin{array}{c}\underset{i\&Element;N\left(p_1\right)}{\&Sigma;}\&Integral;\frac{1}{A_N\left(n^i\right)-p_{1i}}{\mathrm{dp}}_{1i}+\underset{l^i\&Element;P_1}{\&Sigma;}\&Integral;\frac{1}{A_L\left(l^i\right)-b_{1i}}{\mathrm{db}}_{1i}\&GreaterEqual;\underset{i\&Element;N\left(p_2\right)}{\&Sigma;}\&Integral;\frac{1}{A_N\left(n^i\right)-p_{2i}}{\mathrm{dp}}_{2i}+\underset{l^i\&Element;P_2}{\&Sigma;}\&Integral;\frac{1}{A_L\left(l^i\right)-b_{2i}}{\mathrm{db}}_{2i}\\ {\mathrm{\&Delta;b}}_{1i}{\log }_2(1+\frac{{\mathrm{\&Delta;p}}_{1i}{G}_{1i}}{{\mathrm{\&sigma;}}^2+{\mathrm{\&Sigma;}}_{n^S\&Element;N^S\backslash n_{1i}}p\left(n^S\right)G\left(l^S\right)})={\mathrm{\&Delta;b}}_{2i}{\log }_2(1+\frac{{\mathrm{\&Delta;p}}_{2i}{G}_{1i}}{{\mathrm{\&sigma;}}^2+{\mathrm{\&Sigma;}}_{n^S\&Element;N^S\backslash n_{2i}}p\left(n^S\right)G\left(l^S\right)})\end{array}---\left(17\right)$

Increasing the transmission rate, and calculating the total transmission rate DeltaR ═ DeltaR₁＜R(l^V) To obtain a new Δ C₁And Δ C₂Comparison of Δ C₁And Δ C₂If Δ C₁＜ΔC₂Then choose to continue on path P₁Up-transmission, repeating the operation until Δ C₁≥ΔC₂Then the request is shunted to path P₂The above.

Same on path P₁And P₂The data request is continuously added and the mapping is updated, if the total transmission rate Δ R ═ Δ R₁+ΔR₂＜R(l^V) Similarly, assume that path P is selected₃The mapping cost Δ C can be obtained₃Comparison of Δ C₂And Δ C₃If Δ C₂＜ΔC₃Then choose to continue on path P₁And path P₂Up transmission Δ R₂Get an updated mapping cost Δ C₁And Δ C₂,ΔC₃Repeating the operation until Δ C₂≥ΔC₃Then the request is shunted to path P₃And (3) the following conditions are met:

$\begin{array}{c}\underset{i\&Element;N\left(p_2\right)}{\&Sigma;}{\&Integral;}_0^{A_N\left(n^i\right)}\frac{1}{A_N\left(n^i\right)-p_{2i}}{\mathrm{dp}}_{2i}+\underset{l^i\&Element;P_3}{\&Sigma;}{\&Integral;}_0^{A_L\left(l^i\right)}\frac{1}{A_L\left(l^i\right)-b_{2i}}{\mathrm{db}}_{2i}\\ \&GreaterEqual;\underset{i\&Element;N\left(p_3\right)}{\&Sigma;}{\&Integral;}_0^{A_N\left(n^i\right)}\frac{1}{A_N\left(n^i\right)-p_{3i}}{\mathrm{dp}}_{3i}+\underset{l^i\&Element;P_3}{\&Sigma;}{\&Integral;}_0^{A_L\left(l^i\right)}\frac{1}{A_L\left(l^i\right)-b_{3i}}{\mathrm{db}}_{3i}\end{array}---\left(18\right)$

calculating the total rate of transmission Δ R, if Δ R < R (l)^V) Then the above operations are repeated until the rate of the dummy request is satisfied. If the request is mapped to path P_kIf the request is still not met, the mapping is failed, if the mapping process is terminated as long as the rate of the virtual request is met in the process, and the result at this time is the result of virtual network mapping.

The above examples are to be construed as merely illustrative and not limitative of the remainder of the disclosure. After reading the description of the invention, the skilled person can make various changes or modifications to the invention, and these equivalent changes and modifications also fall into the scope of the invention defined by the claims.

Claims (6)

1. A mapping algorithm in a wireless network virtualization environment, comprising the steps of:

step 1): constructing a wireless network virtualization mapping model;

step 2): calculating available resources and extended resources in the current network according to a wireless network virtualization mapping model, determining an interference model, and then constructing a cost function of virtual network mapping by defining a resource unit price as an optimization target, wherein the target is to minimize mapping cost, and the mapping cost comprises the cost of node mapping and the cost of link mapping;

step 3): the method comprises the steps of firstly realizing optimized mapping of virtual nodes, then carrying out virtual link mapping to find a path based on the positioned virtual nodes, synthesizing a path separation algorithm, and optimizing the flow request of the virtual link by using the principle of minimum mapping cost when a virtual network is constructed.

2. The mapping algorithm in a wireless network virtualization environment according to claim 1,

the step 1 of constructing the wireless network virtualization mapping model comprises a physical network model and a wireless virtual network request model, wherein the physical network model comprises: using an undirected graph G with weights^S＝(N^S，L^S) Represents the underlying physical network, where N^SRepresenting a set of underlying physical nodes, L^SRepresenting a set of underlying links; for each physical node n^S∈N^SIn other words, the node power p (n) is also included^S) And the location information loc (n) of the node^S) For each physical link l^S∈L^SIn other words, the link bandwidth information b (l) is included^S)，b(l^S) Represents the maximum value of the available bandwidth of the physical link; furthermore, with p^STo represent a physical path, P, in a physical network^SRepresenting a set of physical paths; the wireless virtual network request model is as follows: defining an undirected graph G with weight^V＝(N^V,L^V) As a request for a wireless virtual network, N^VRepresenting a set of virtual network requesting nodes, each node n^V∈N^VContaining location information loc (n)^V)，L^VSet of links representing virtual network requests, for each virtual link l^V∈L^VIt contains the transmission rate R (l) that the virtual link needs to reach^V) The physical network needs to allocate a certain bandwidth and power for the virtual request.

3. The mapping algorithm in a virtualized wireless network environment according to claim 1 or 2, wherein the objective function in step 2 is

$min\left(C\right(P^S\left)\right)=min\left(\underset{n^V\&RightArrow;n^S,n^S\&Element;P^S}{\mathrm{\&Sigma;}}\mathrm{\&alpha;}\right(n^S\left)\right(n^V)+\underset{l^V\&RightArrow;l^S,l^S\&Element;P^S}{\mathrm{\&Sigma;}}\mathrm{\&beta;}(l^S\left)b\right(l^V\left)\right)$

s.t.p(n^V)≤A_N(n^S)

dis(loc(n^V),loc(n^S))≤D^V

$b\left(l^V\right)\&le;A_L\left(l^S\right)(\&ForAll;l^V\&RightArrow;P^S,l^S\&Element;P^S)$

$b\left(l^V\right){\log }_2(1+\frac{p\left(n^V\right)G\left(l^V\right)}{{\mathrm{\&sigma;}}^2+{\mathrm{\&Sigma;}}_{n^S\&Element;N^S\backslash n^V\&RightArrow;n^S}p\left(n^S\right)G\left(l^S\right)})\&GreaterEqual;R\left(l^V\right)$

Wherein,A_N(n^S) And A_L(l^S) Representing the available resources of the node and link, respectively, the calculation is as follows:

${\mathrm{Capacity}}_N\left(n^S\right)=A_N\left(n^S\right)=p\left(n^S\right)-\underset{\&ForAll;n^V\&RightArrow;n^S}{\mathrm{\&Sigma;}}p\left(n^V\right),{\mathrm{Capacity}}_L\left(l^S\right)=A_L\left(l^S\right)-\underset{\&ForAll;l^V\&RightArrow;l^S}{\mathrm{\&Sigma;}}p\left(l^V\right)$ G(l^V) Representing the channel gain, σ, of the mapped physical link²Representing white gaussian noise in the channel,indicating the interference that other links cause to the link.

4. The mapping algorithm in the virtualized environment of wireless network according to claim 1 or 2, wherein the step 3 of performing virtual link mapping to find paths uses K shortest paths algorithm to find K paths.

5. The mapping algorithm in a wireless network virtualization environment according to claim 1,the step 3 of implementing the optimized mapping of the virtual nodes includes the steps of: (1) firstly, a physical node set which can be mapped in respective range is found for two virtual nodes, namely dis (loc (n) is satisfied^V),loc(n^S) D, D represents the mapping radius of the virtual node in the virtual network request, dis (loc (n)^V),loc(n^S) Represents a virtual node n^STo physical node n^VThe distance of (d); 2) and respectively calculating the expansion resources of the nodes in the two physical node sets, and selecting the node with the maximum expansion resource to map the virtual node.

6. The mapping algorithm in a wireless network virtualization environment according to claim 1,

the link mapping process of the step 3 is realized by the following steps:

(1) respectively calculating the interference coefficient of each link in the physical network;

(2) distributing the node mapping cost to the links corresponding to the physical nodes, namely, the weight of each physical link is the sum of the resource cost of the corresponding node and the link interference coefficient, and finding K shortest paths P between two physical nodes by adopting a K shortest path algorithm₁,P₂,...P_kThe shortest path comprehensively considers the size of resources and the interference between links;

(3) first, the speed request passes through the path P with lowest cost₁If the total data Δ R transmitted has not yet reached the requested data size and is on path P₁Cost of upstream transmission and on path P₂When the cost of the upper transmission is equal, the data is shunted to the path P₂Continuously increasing the transmitted data if DeltaR < R (l)^V) And is Δ C₂＝ΔC₃Then the data is shunted to path P₃By analogy, data can be shunted to path P_kIf Δ R < R (l) at this time^V) If so, link mapping fails; at any time during this process, if Δ R ≧ R (l)^V) Then the mapping is successful and the algorithm is terminated.