CN104243258A - Virtual network mapping method and system based on classification - Google Patents

Abstract

The invention provides a virtual network mapping method and system based on classification. The virtual network mapping method comprises the steps that according to network topology information of each region, virtual nodes in a request for a virtual network are classified, the requested virtual network is partitioned into small-sized virtual network sub-requests according to a classification result, the virtual network sub-requests are dispatched to management entities in all the regions, and concurrent mapping processing is conducted; when the management entities in all the regions map the virtual network sub-requests, a resource equalization mechanism is introduced, equalization processing is conducted on demand limits of the virtual network sub-requests and available resources of a bottom layer physical network of the regions, and mapping of the virtual nodes and mapping of virtual links are coordinated and integrated; after the management entities in all the regions complete mapping, a global management entity integrates mapping results of the management entities in all the regions and feeds back the mapping results to a user. Compared with the prior art, the virtual network mapping method and system based on classification has the advantages that computation complexity is low, the mapping rate is high, and bottom layer physical network loads are better equalized.

Description

A kind of mapping method of virtual network based on classification and system

Technical field

The present invention relates to network virtualization technology, specifically, relate to a kind of mapping method of virtual network based on classification and system.

Background technology

Network virtualization technology is the core technology of Future Internet, and its allows to support multiple virtual network in a shared bottom-layer network resource, can have the different network architectures, also can carry dissimilar service between these virtual networks.

A key issue of network virtualization technology is virtual network mapping problems.By the dummy node in virtual network requests and virtual link are mapped in bottom-layer network resource, to meet node and the link requirements of virtual network requests.Virtual network mapping problems is proved to be very scabrous technical problem, and its computation complexity is higher.Existing method does not propose effective measures to reduce computation complexity, and make to map the processing time longer, the utilance of bottom-layer network resource is lower.

Summary of the invention

The object of the invention is to, in order to overcome the problems referred to above, the invention provides a kind of mapping method of virtual network based on classification and system.

To achieve these goals, the invention provides a kind of mapping method of virtual network based on classification, described method comprises:

Step 101) when receiving virtual network requests, the dummy node involved by virtual network requests is classified according to criteria for classification, the dummy node belonged to a different category is split in different virtual net string bag requests;

Wherein, namely dummy node is categorized into from the bottom physical network in its nearest region according to dummy node to the demand information in geographical position by described criteria for classification; All dummy nodes in the request of same virtual net string bag belong to same classification, and namely this classification is the classification belonging to any one dummy node in this virtual net string bag request;

Step 102) be sent to each virtual net string bag request belonging to it bottom physical network in certain region;

Step 103) according to the equally loaded ability of each node in the equally loaded demand of each node in the request of virtual net string bag and region, the request of virtual net string bag to be mapped in region on bottom physical layer network; Wherein,

Described equally loaded requirement representation when load balancing, the loading demand of each dummy node; The resource that described loading demand is defined as dummy node request and the product of resource of all link request being connected to described dummy node;

When described equally loaded capability list is shown in load balancing, the load capacity of each physical node in region; The available resources that described load capacity is defined as physical node and the product of available resources of all links being connected to described physical node.

According to the geographical position demand information of dummy node dummy node is categorized in the bottom physical network in nearest region, the geographical position that requires from it when classification analysis carries out to the dummy node in virtual network requests.

Above-mentioned steps 103) comprise further:

Step 103-1) detect the current unmapped dummy node n with maximum equally loaded demand in the request of virtual net string bag ^v;

Step 103-2) find dummy node n in this region ^vk candidate's bottom physical node, k described both candidate nodes is current also unmapped and can meet described dummy node n in this region ^vthe bottom physical node of front k maximum equally loaded ability of demand;

Step 103-3) to each candidate's bottom physical node, calculate the path sum between this candidate's bottom physical node and the bottom physical node corresponding to each dummy node mapped, by described dummy node n ^vbe mapped to and have in the both candidate nodes of shortest path length, and the virtual link between the dummy node mapped is mapped.

Above-mentioned equally loaded ability adopts following acquisition:

Step 201) physical node of the regional that bottom physical network comprises and the resource of physical link are normalized respectively;

Wherein, the memory capacity of normalized and the node of physical node, magnetic disc i/o, CPU quantity are relevant with memory size; The normalized of physical link is relevant to the bandwidth of link, delay and delay jitter;

Step 202) calculate the load capacity of the node in bottom physical network in regional;

Wherein, the described load capacity available resources that are defined as physical node and the product of available resources of all links being connected to described physical node;

Step 203) calculate the equally loaded matrix of regional in bottom physical network;

Wherein, each element in described equally loaded matrix is: the weighted value of the mutual contribution of the load balance ability between each physical node in certain region, and the weighted value of the mutual contribution of described load balance ability is specially: as physical node n ^sbe exactly physical node m ^sitself, i.e. n ^s=m ^s, or physical node n ^swith physical node m ^sbetween when having physical link to connect, physical node n ^sto physical node m ^sequally loaded ability have contribution, now weighted value is the positive number of a non-zero; Otherwise, physical node n ^sto physical node m ^sthe contribution weight of equally loaded ability be 0; n ^sn-th of representing matrix ^soK, m ^sthe m of representing matrix ^srow, the line number of matrix and columns are all total numbers of physical node, and what footmark s represented is bottom physical network;

Step 204) according to equally loaded matrix, iterative computation goes out the equally loaded ability of each physical node in region, i.e. physical node n ^sload balance ability be the contribution sum of the equally loaded ability of all physical nodes in this region.

Above-mentioned equally loaded demand is adopted and is obtained with the following method:

Step 301) demand of the dummy node in each virtual net string bag request and virtual link is normalized respectively;

Wherein, the memory capacity of normalized and the node of dummy node, magnetic disc i/o, CPU quantity are relevant with memory size demand; The normalized of virtual link and the bandwidth of link, delay are relevant with delay jitter demand;

Step 302) calculate the loading demand of each dummy node in each virtual net string bag request;

Wherein, the described loading demand demand that is defined as dummy node and the product of demand of all links being connected to described dummy node;

Step 303) calculate the equally loaded matrix of each virtual net string bag request;

Wherein, each element in described equally loaded matrix is: the interactional weighted value of the equally loaded demand between each dummy node in the request of virtual net string bag, and this weighted value is specially: as dummy node n ^vbe exactly dummy node m ^vitself, i.e. n ^v=m ^v, or dummy node n ^vwith dummy node m ^vbetween when having virtual link to connect, definition weighted value is the numerical value of a non-zero; Otherwise n ^vto m ^vthe weighing factor of equally loaded demand be 0;

N ^vn-th of representing matrix ^voK, m ^vthe m of representing matrix ^vrow, the line number of matrix and columns are all total numbers of dummy node, and what footmark v represented is virtual network requests;

Step 304) according to equally loaded matrix, iterative computation goes out the equally loaded demand of each dummy node in region, i.e. dummy node n ^vload balancing demand be the equally loaded demand of all dummy nodes in this virtual net string bag request affect sum.

In order to realize said method, the invention provides a kind of virtual network mapped system based on classification, described system comprises:

Global administration's module, during for receiving virtual network requests, dummy node involved by virtual network requests is classified according to criteria for classification, by in the dummy node segmentation to different virtual net string bag request that belongs to a different category, described criteria for classification refers to that a dummy node will belong to the classification at place, region nearest from it in bottom physical network; And certain region each virtual subnetwork request is sent in the bottom physical network belonging to it; With

District management module, for the equally loaded ability according to each node in the equally loaded demand of each node in the request of virtual net string bag and region, to be mapped in region on bottom physical layer network by the request of virtual net string bag;

Wherein, described equally loaded requirement representation when load balancing, the loading demand of each dummy node; The resource that described loading demand is defined as dummy node request and the product of resource of all link request being connected to described dummy node;

When described equally loaded capability list is shown in load balancing, the load capacity of each physical node in region; The available resources that described load capacity is defined as physical node and the product of available resources of all links being connected to described physical node.

Dummy node is categorized in the nearest region bottom physical network in the geographical position that requires from it according to the geographical position demand information of dummy node when classification analysis carries out to the dummy node in virtual network requests.

Above-mentioned zone administration module comprises further:

Equally loaded ability or equally loaded demand obtain submodule, for obtaining the equally loaded ability value of each node in each region of bottom physical network and obtaining the equally loaded requirements of each dummy node of each virtual net string bag request;

Detection sub-module, for detecting the current unmapped dummy node n with maximum equally loaded demand in the request of virtual net string bag ^v;

Candidate's bottom physical node chooser module, for finding dummy node n in certain region ^vk candidate's bottom physical node, k described both candidate nodes is current also unmapped and can meet described dummy node n in this region ^vthe bottom physical node of front k maximum equally loaded ability of demand; With

Node and link maps submodule, for each candidate's bottom physical node, calculate the path sum between this candidate's bottom physical node and the bottom physical node corresponding to each dummy node mapped, by described dummy node n ^vbe mapped to and have in the both candidate nodes of shortest path length, and the virtual link between the dummy node mapped is mapped.

Above-mentioned equally loaded ability obtains submodule and comprises further:

First normalized module, is normalized respectively for the physical node of regional that comprises bottom physical network and the resource of physical link;

Wherein, the memory capacity of normalized and the node of physical node, magnetic disc i/o, CPU quantity are relevant with memory size; The normalized of physical link is relevant to the bandwidth of link, delay and delay jitter;

First node load capacity acquisition module, for calculating the load capacity of the node in bottom physical network in regional;

Wherein, the described load capacity available resources that are defined as physical node and the product of available resources of all links being connected to described physical node;

Equally loaded matrix acquisition module, for calculating the equally loaded matrix of regional in bottom physical network;

Wherein, the mutual contribution weighted value of the equally loaded ability between each physical node in the described equally loaded defined matrix region of bottom physical network, and if only if physical node n ^sbe exactly physical node m ^sitself, i.e. n ^s=m ^s, or n ^sand m ^sbetween have physical link to connect, now weighted value is the positive number of a non-zero; Otherwise physical node n ^sto physical node m ^sthe contribution weight of equally loaded ability be 0; With

Equally loaded capacity calculation module, for according to equally loaded matrix, iterative computation goes out the equally loaded ability of each physical node in region, i.e. physical node n ^sload balance ability be the contribution sum of the equally loaded ability of all physical nodes in this region.

Above-mentioned equally loaded demand obtains submodule and comprises further:

Second normalized module, for being normalized respectively the demand of the dummy node in each virtual net string bag request and virtual link;

Wherein, the memory capacity of normalized and the node of dummy node, magnetic disc i/o, CPU quantity are relevant with memory size demand; The normalized of virtual link and the bandwidth of link, delay are relevant with delay jitter demand;

Second loading demand computing module, for calculating the loading demand of each dummy node in each virtual net string bag request;

Wherein, the described loading demand demand that is defined as dummy node and the product of demand of all links being connected to described dummy node;

Second equally loaded matrix, for calculating the equally loaded matrix of each virtual net string bag request;

Wherein, the interactional weighted value of the equally loaded demand between each dummy node in the virtual net string bag request of described equally loaded defined matrix, and if only if dummy node n ^vbe exactly dummy node m ^vitself, i.e. n ^v=m ^v, or dummy node n ^vwith dummy node m ^vbetween have virtual link to connect, weighted value is now the positive number of a non-zero; Otherwise dummy node n ^vto dummy node m ^vthe weighing factor of equally loaded demand be 0; With

Equally loaded Requirement Acquisition module, for according to equally loaded matrix, iterative computation goes out the equally loaded demand of each dummy node in region, i.e. dummy node n ^vload balancing demand be the equally loaded demand of all dummy nodes in this virtual net string bag request affect sum.

In a word, the invention provides a kind of mapping method of virtual network based on classification, described method comprises: according to the network topological information in each region, classification analysis is carried out to the dummy node in virtual network requests, according to the virtual net string bag request that the virtual network of request is divided into scale less by classification results, the management entity tasking each region is divided to carry out the mapping process walked abreast the request of virtual net string bag; The management entity in each region is when the request of maps virtual network, introduce resources balance mechanism, respectively equilibrium treatment is carried out to the demand restriction of virtual net string bag request and the available resources of region bottom physical network, coordinate to integrate the mapping of dummy node and the mapping of virtual link simultaneously; After management entity in all regions all completes mapping, integrated the mapping result of the management entity in each region by the management entity of the overall situation and feed back to user.Described mapping method of virtual network processes the mapping of virtual network requests in a distributed manner on the management entity of the overall situation, the management entity in region processes virtual network requests centralizedly and maps.Described classification mainly utilizes the demand restricted informations such as the geographical position of each dummy node in virtual network requests, distance between the management entity calculating each dummy node and described each region, and carry out classification analysis to determine the region belonging to dummy node according to described distance.The dummy node belonging to same category is divided in identical virtual net string bag request by described segmentation, and the virtual link between the request of each virtual net string bag is reduced to a virtual link.When the management entity of described mapping method of virtual network in each region carries out the mapping of virtual net string bag request, introduce resources balance mechanism.When the management entity of described mapping method of virtual network in each region carries out the mapping of virtual net string bag request, coordinate to incorporate the mapping of dummy node and the mapping of virtual link.Described resources balance mechanism is realized the impact of this node by the resource load of aggregators self and the resource load of its neighbor node.The load capacity of described node is relevant with the link circuit resource being connected to this node to the resource of node.

Compared with prior art, technical advantage of the present invention is:

By technique scheme, virtual network requests by sub-request less for the scale that is split into, and is mapped concomitantly by regional management entity, thus reduces computation complexity, reduces mapping time.The management entity of regional is when the request of maps virtual network, machine-processed by introducing resources balance, to realize the load balancing of bottom-layer network resource.The extensive distribution avoiding node mapping is integrated in the coordination that dummy node maps and virtual link maps, and decreases the bottom physical path length of link maps.

Accompanying drawing explanation

Fig. 1 is the workflow diagram based on global management entity in the mapping method of virtual network of classification provided by the invention;

Fig. 2 is the workflow diagram based on district management entity in the mapping method of virtual network of classification provided by the invention;

Fig. 3 is the bottom physical network configuration diagram of the embodiment of the mapping method of virtual network based on classification provided by the invention;

Fig. 4 is the schematic diagram that the virtual network of global management entity in the embodiment of the mapping method of virtual network based on classification provided by the invention maps;

Fig. 5 is the load-balancing mechanism schematic diagram in the embodiment of the mapping method of virtual network based on classification provided by the invention; In this figure each near nodal little square frame in the normalized node resource of numeral, by memory capacity, magnetic disc i/o, CPU, the impacts such as internal memory; The normalized link circuit resource of numeral in this figure on lines, by the impact of bandwidth, delay, shake;

Fig. 6-1 is the virtual net string bag request schematic diagram that the embodiment of the present invention provides;

Fig. 6-2 is the schematic diagrames before the bottom physical network in the region that provides of the embodiment of the present invention maps;

Fig. 6-3 be in the region that provides of the embodiment of the present invention bottom physical network map after schematic diagram.

Embodiment

Below in conjunction with accompanying drawing, technical scheme of the present invention is described in detail.

A kind of mapping method of virtual network based on classification of the present invention comprises:

According to the network topological information in each region, classification analysis is carried out to the dummy node in virtual network requests, according to the virtual net string bag request that the virtual network of request is divided into scale less by classification results, the management entity tasking each region is divided to carry out the mapping process walked abreast the request of virtual net string bag; The management entity in each region is when the request of maps virtual network, introduce resources balance mechanism, respectively equilibrium treatment is carried out to the demand restriction of virtual net string bag request and the available resources of region bottom physical network, coordinate to integrate the mapping of dummy node and the mapping of virtual link simultaneously; After management entity in all regions all completes mapping, integrated the mapping result of the management entity in each region by the management entity of the overall situation and feed back to user.

In order to clearly set forth content of the present invention, what the present invention mapped virtual network is described as follows:

A. bottom physical network

Definition bottom physical network is a non-directed graph set G ^s(N ^s, E ^s), i-th region representation of bottom physical network is non-directed graph wherein represent the node set in the i of region, represent the link set in the i of region.Defined node n ^sneighbor node set be N ^s(n ^s), definition is connected to node n ^slink set be E ^s(n ^s).Node in bottom physical network and link are all associated with their resource restriction.We define in bottom physical network, and node resource restricted information comprises geographical position L ^s, memory capacity S ^s, magnetic disc i/o speed I ^s, CPU ability C ^swith memory size M ^s, link circuit resource restriction comprises bandwidth B ^s, postpone D ^swith delay jitter J ^s(e.g., physical node n ^snode resource restricted information: geographical position represents position coordinates L ^s(n ^s)=(10,20), memory capacity S ^s(n ^s)=500MB, magnetic disc i/o speed I ^s(n ^s)=100bps, CPU ability C ^s(n ^s)=4 core and memory size M ^s(n ^s)=2GB, link circuit resource limits: bandwidth B ^s(n ^s)=100MB, delay D ^s(n ^s)=200ms and delay jitter J ^s(n ^s)=20ms).Table 1 has itemized out the resource restriction of bottom physical network in the present invention, and is each resource restriction definition weights influence factor, limits with the normalization resource of computing node and link.Therefore, bottom physical network nodes n ^snormalization resource representation be: $R_N^s\left(n^s\right)=S^s(n^s{)}^{*}{w}_S^n+I^s(n^s{)}^{*}{w}_I^n+C^s(n^s{)}^{*}{w}_C^n+M^s(n^s{)}^{*}{w}_M^n;$ Link e ^snormalization resource representation be: $R_E^s\left(e^s\right)=B^s(e^s{)}^{*}{w}_B^e+D^s(e^s{)}^{*}{w}_D^e+J^s(e^s{)}^{*}{w}_J^e.$ Wherein geographical position is mainly used in classification analysis, need not consider in the normalization of node resource represents.

The resource restriction of table 1 bottom physical network

B. virtual network requests

Virtual networks request of the present invention is a non-directed graph G ^v(N ^v, E ^v), i-th son request of virtual network requests is expressed as non-directed graph wherein represent the node set of i-th virtual net string bag request, represent the link set of i-th virtual net string bag request.Defined node n ^vneighbor node set be N ^v(n ^v), definition is connected to node n ^vlink set be E ^v(n ^v).Node in virtual network requests and link are all associated with their demand restriction.In our virtual networks request, the demand restricted information that dummy node proposes comprises: geographical position L ^v, memory capacity S ^v, magnetic disc i/o speed I ^v, CPU quantity C ^vwith memory size M ^v, the demand restriction of virtual link comprises bandwidth B ^v, postpone D ^vwith delay jitter J ^v(e.g., dummy node n ^vnode resource restricted information: geographical position represents position coordinates L ^v(n ^v)=(30,20), memory capacity S ^v(n ^v)=5MB, magnetic disc i/o speed I ^v(n ^v)=50bps, CPU ability C ^v(n ^v)=2 core and memory size M ^v(n ^v)=200MB, link circuit resource limits: bandwidth B ^v(n ^v)=10MB, delay D ^v(n ^v)=200ms and delay jitter J ^v(n ^v)=20ms).Table 2 is demand restrictions of virtual network requests in the present invention, and wherein the weights influence factor of each demand restriction is identical with the weights influence factor that resource in bottom physical network limits.Therefore, dummy node n in virtual network requests ^vnormalization resource representation be: $R_N^v\left(n^v\right)=S^v(n^v{)}^{*}{w}_S^n+I^v(n^v{)}^{*}{w}_I^n+C^v(n^v{)}^{*}{w}_C^n+M^v(n^v{)}^{*}{w}_M^n;$ Virtual link e ^vnormalization resource representation be: $R_E^v\left(e^v\right)=B^v(e^v{)}^{*}{w}_B^e+D^v(e^v{)}^{*}{w}_D^e+J^v(e^v{)}^{*}{w}_J^e.$ Wherein geographical position is mainly used in classification analysis, need not consider in the normalization of node resource demand represents.

The demand restriction of table 2 virtual network requests

C. the place-centric of region bottom physical network

In order to better virtual network requests is mapped in bottom physical network, the present invention makes full use of the topology information of virtual network requests, according to the geographical position demand restriction of dummy node, classification analysis is carried out to the dummy node in virtual network requests, to be mapped to by dummy node in its nearest region bottom physical network.In definition bottom physical network, the place-centric set of regional is O ^s, wherein for the place-centric of region i, it is defined as the geometric center of all nodes in the i of region, namely

$O_i^s=\frac{1}{\left|N_i^s\right|}\underset{\∀n^s\∈N_i^s}{\mathrm{\Σ}}{L}^s\left(n^s\right)$

Wherein, represent the node number in the i of region.

D. load-balancing mechanism

In order to improve resource utilization and the load balancing of bottom physical network, the present invention adopts load-balancing mechanism to make full use of the resource of node self and the resource of its neighbor node.In bottom physical network, the load capacity of node is relevant with the link circuit resource being connected to node with node resource, and link circuit resource is shared by its two end points.Therefore, bottom physical network interior joint n is defined ^sload capacity R ^s(n ^s) as follows:

$R^s\left(n^s\right)=R_N^s\left(n^s\right)\·\underset{e^s\∈E^s\left(n^s\right)}{\mathrm{\Σ}}\frac{1}{2}{R}_E^s\left(e^s\right)$

On this basis, bottom physical network interior joint n is defined ^sequally loaded ability as follows:

$\stackrel{\‾}{R^s\left(n^s\right)}=w_1\·\stackrel{\‾}{R^s\left(n^s\right)}+w_2\underset{m^s\∈N^s\left(n^s\right)}{\mathrm{\Σ}}\frac{R^s\left(n^s\right)}{\underset{h^s\∈N^s\left(n^s\right)}{\mathrm{\Σ}}{R}^s\left(h^s\right)}\·\stackrel{\‾}{R^s\left(m^s\right)}$

That is, the equally loaded ability of equally loaded ability and node self of node is relevant with the equally loaded ability of the neighbor node of node, wherein w ₁, w ₂be the relevant weights influence factor, meet w ₁+ w ₂=1, for node m ^sto node n ^simpact.Thus, the equally loaded matrix of definable region i as follows:

$A_i^s(m^s,n^s)=\left\{\begin{array}{ccc}{w}_1& {\mathrm{ifm}}^s=n^s,m^s\∈N_i^s& \\ \frac{w_2\·R^s\left(n^s\right)}{\underset{h^s\∈N^s\left(m^s\right)}{\mathrm{\Σ}}}& \mathrm{if}(m^s,n^s)\∈E_i^s& \\ 0& \mathrm{other}& \end{array}\right.$

The equally loaded ability of region i interior nodes is expressed as follows:

In like manner, the equally loaded demand of the node in definable virtual network requests the wherein loading demand R of dummy node in virtual network requests ^v(n ^v) meet:

$R^v\left(n^v\right)=R_N^v\left(n^v\right)\·\underset{e^v\∈E^v\left(n^v\right)}{\mathrm{\Σ}}\frac{1}{2}{R}_E^v\left(e^v\right)$

Dummy node n ^vequally loaded demand meet:

$\stackrel{\‾}{R^v\left(n^v\right)}=w_1\·\stackrel{\‾}{R^v\left(n^v\right)}+w_2\·\underset{m^v\∈N^v\left(n^v\right)}{\mathrm{\Σ}}\frac{R^v\left(n^v\right)}{\underset{h^v\∈N^v\left(n^v\right)}{\mathrm{\Σ}}{R}^v\left(h^v\right)}\·\stackrel{\‾}{R^v\left(m^v\right)}$

That is, the equally loaded of the equally loaded demand and node self of dummy node needs the equally loaded demand of the neighbor node of summing junction relevant, wherein, for node m ^vto node n ^vimpact.

The equally loaded matrix A of virtual network requests ^vbe defined as follows:

$A^v(m^v,n^v)=\left\{\begin{array}{ccc}{w}_1& {\mathrm{ifm}}^v=n^v,m^v\∈N^v& \\ \frac{w_2\·R^v\left(n^v\right)}{\underset{h^v\∈N^v\left(m^v\right)}{\mathrm{\Σ}}{R}^v\left(h^v\right)}& \mathrm{if}(m^v,n^v)\∈E^v& \\ 0& \mathrm{other}& \end{array}\right.$

Thus, the equally loaded demand of definable i-th virtual net string bag request interior joint

In order to more clearly set forth the object, technical solutions and advantages of the present invention, below in conjunction with the accompanying drawing in the embodiment of the present invention, technical scheme in the embodiment of the present invention is clearly and completely described, obviously, described embodiment is a part of embodiment of the present invention, instead of whole embodiments.

As shown in Figure 1, the invention provides a kind of as follows based on the workflow of global management entity in the mapping method of virtual network of classification: when 1) starting, the arrival of global management entity wait virtual network requests; 2) when virtual network requests arrives, described global management entity carries out classification analysis to the dummy node in virtual network requests, determines the classification belonging to each dummy node and region; 3) according to described classification results, virtual network requests is divided into multiple virtual subnet request; 4) each described virtual net string bag request dispatching is processed respectively to the management entity of regional; 5) when the management entity of regional has all processed, merge the result of the management entity of regional and feed back to user.

When carrying out classification analysis to the dummy node in virtual network requests, main considering the demand restrictions such as the geographical position of dummy node, dummy node being categorized in the nearest region bottom physical network in the geographical position that requires from it, therefore, definable dummy node n ^vaffiliated classification $\mathrm{\φ}\left(n^{\text{v}}\right)={\arg \min }_{i\∈[0,|O^s\left|\right)}\left\{\mathrm{dist}(L^v\left(n^v\right),O_i^s)\right\}.$

According to the classification results of virtual network requests, the dummy node belonging to different classification is divided in different virtual net string bag requests, and simplifies the link between each height request.Belong to the request of i-th virtual net string bag dummy node n ^vmeet φ (n ^v)=i, the virtual link between all leap virtual net string bag request i and virtual net string bag request j is reduced to a link e (i, j), and it satisfies condition: namely the normalization demand of the virtual link of this simplification crosses over the maximum in all link normalization demands of virtual net string bag request i and virtual net string bag request j, if the result that virtual network maps can meet simplify link e (i, j) demand, then all will be met for all virtual link demands of crossing over virtual net string bag request i and virtual net string bag request j.After virtual network requests being divided into the request of multiple virtual net string bag, by each virtual net string bag request dispatching in suitable region bottom physical network, for virtual net string bag request i, namely to be assigned in i-th bottom physical network and further map process.

Fig. 3 is the bottom physical network configuration diagram of a kind of mapping method of virtual network embodiment based on classification of the present invention, and as shown in Figure 3, the bottom physical network in the present embodiment is double-layer structure, i.e. management level and service layer.Wherein, management level comprise global management entity and district management entity.Nodal information in this region of district management entity assembles and Link State, and calculate the place-centric in this region and the logical topology of region interior nodes, then send global management entity to by upgrading the result calculated.Global management entity directly and district management entity communication, keeps the last state of whole physical network.Wherein, the information of each node and link in district management entity periodic collection region, only has when the network state in region changes, and district management entity initiatively transmits lastest imformation, to reduce communication overhead to global management entity.In addition, global management entity generally belongs to passive receiving area network state information, but still can initiatively interrogation management entity, and asks for relevant information.

Fig. 4 is the schematic diagram that the virtual network of global management entity in the embodiment of a kind of mapping method of virtual network based on classification of the present invention maps.As shown in Figure 4, when virtual network requests VNR arrives, first global management entity carries out classification analysis to the topology of virtual network requests, and according to the result of classification, this virtual network requests is divided into the less virtual net string bag request VNR of scale _sub1and VNR _sub2.According to the classification belonging to each virtual net string bag request, they are assigned in suitable region bottom physics the centralized virtual network of carrying out walking abreast and map.As shown in Figure 4, virtual net string bag request VNR _sub1be assigned to region 1 to map, virtual net string bag request VNR _sub2be assigned to region 3 to map.

As shown in Figure 2, the invention provides a kind of as follows based on the workflow of district management entity in the mapping method of virtual network of classification: when 1) starting, the equally loaded ability of each node in the bottom-layer network node in district management entity zoning; 2) arrival of virtual net string bag request is waited for; 3) when the request of virtual net string bag arrives, the equally loaded demand of each dummy node in the request of virtual net string bag is calculated; 4) request of virtual net string bag is mapped on the bottom physical network in region.

During the equally loaded demand of the equally loaded ability of bottom physical network nodes and the request of virtual net string bag in zoning, calculate utilizing the aforementioned equally loaded mechanism mentioned.For the bottom physical network nodes in the i of region, its equally loaded ability is expressed as its algorithm false code is as follows:

The same algorithm of the present invention calculates the equally loaded demand of i-th virtual net string bag request

Fig. 5 is the load-balancing mechanism schematic diagram in the embodiment of a kind of mapping method of virtual network based on classification provided by the invention, and the present invention calculates equally loaded demand and equally loaded ability respectively to bottom physical network in the request of virtual net string bag and region.As shown in Figure 5, the present invention uses load-balancing mechanism to calculate equally loaded ability to bottom physical network, is similar to the calculating of the equally loaded demand of virtual net string bag request.The calculating of equally loaded demand and equally loaded ability depends on the equally loaded matrix (as shown in table 4) of each network topology and the load capacity (as shown in table 3) of each node.As can be seen from the result of calculation of such as table 5, the Node B in this region has maximum equally loaded ability, and node G has minimum load balance ability.

Table 3, the load capacity of each node

A	B	C	D	E	F	G
							175	387	85	247.5	90	160	72

Table 4, equally loaded matrix

?	A	B	C	D	E	F	G
								A	0.70	0.17	0	0	0	0.09	0.04
B	0.09	0.70	0.04	0.12	0.05	0	0
								C	0	0.16	0.70	0.14	0	0	0
D	0	0.19	0.06	0.70	0.06	0	0
								E	0	0.12	0	0.11	0.70	0.07	0
F	0.16	0	0	0	0.08	0.70	0.06
								G	0.16	0	0	0	0	0.14	0.70

Table 5, the load balance ability of each node

A	B	C	D	E	F	G
							182.5	344.3	91.19	229.5	125.5	108.4	48.48

In the present invention, district management entity needs the equally loaded ability of each node in bottom physical network in range of summation according to the equally loaded of each node in the request of virtual net string bag, adopts Greedy strategy the request of virtual net string bag to be mapped in region on bottom physical network.The present invention is in the process mapped, and the mapping of dummy node and virtual link is carried out simultaneously, and namely when mapping dummy node, the demand characteristic of combined with virtual link is to avoid the expense in longer path in bottom physical network.When mapping, first detect the current unmapped dummy node n with maximum equally loaded demand in the request of virtual net string bag ^v, then find dummy node n in this region ^vk candidate's bottom physical node, k described both candidate nodes is current also unmapped and can meet described dummy node n in this region ^vthe bottom physical node of front k maximum equally loaded ability of demand.To each candidate's bottom physical node, calculate the path sum between this candidate's bottom physical node and the bottom physical node corresponding to each dummy node mapped, by described dummy node n ^vbe mapped to and have in the both candidate nodes of shortest path length, and the virtual link between the dummy node mapped is mapped.

The mapping algorithm false code of district management entity to the request of virtual net string bag is described below:

Fig. 6-1,6-2 and 6-3 is the schematic diagram that the virtual network of district management entity in the embodiment of a kind of mapping method of virtual network based on classification of the present invention maps, the normalization resource of the numeric representation node in the little square frame in above-mentioned three width figure, by memory capacity, magnetic disc i/o, the impact of CPU and internal memory; The normalization resource of the numeric representation link on lines, by bandwidth, postpones, the impact of shake; Moment on wave (t1, t2 ...) time sequencing that is processed in the map of node pointed by expression or link; equally loaded ability/equally loaded the demand of the numeric representation node in icon.In the present embodiment, the number k=3 of both candidate nodes is got.As in fig. 6-2, before starting mapping, the node in the request of virtual net string bag with maximum equally loaded demand is a, meanwhile, the node in bottom physical network in region with maximum equally loaded ability is B, wherein, the resource requirement of node a is 12, the available resources of Node B are 14, and therefore Node B can meet the demand of node a, are mapped in Node B by node a.Now, in the request of virtual net string bag, the unmapped node with maximum equally loaded demand is dummy node c, the both candidate nodes of dummy node c in bottom physical network is D, A, E(reason is: the number k=3 getting both candidate nodes, and after removing the bottom physical node B that mapped, maximum 3 both candidate nodes of remaining k=3 equally loaded ability are D, A, E).Wherein, D, A, E tri-nodes are 1 to the path of the Node B mapped, and the equally loaded ability of D is maximum, is therefore mapped on D by dummy node c, the dummy node a will mapped again, virtual link <a between c, c> is mapped to physical link <B, on D>.Now, dummy node to be mapped is b, and the both candidate nodes of dummy node b is A, E, F.Wherein, both candidate nodes A is 4 to the path sum of the node mapped, both candidate nodes E is 2 to the path of the node mapped, both candidate nodes F is 4 to the path of the node mapped, therefore, dummy node b is mapped on both candidate nodes E, to reduce the expense of longer map paths, and then by virtual link <a, b> is mapped to bottom physical pathway <B, E>, by virtual link <c, b> is mapped to bottom physical pathway <D, E>.

In a kind of mapping method of virtual network based on classification of the present invention, when all virtual net string bag requests are all successfully mapped to the bottom physical network in each region, whole virtual network requests maps and just successfully be have mapped.Otherwise, refusal is mapped and Resourse Distribute, and result is returned to user.

In sum, a kind of mapping method of virtual network based on classification of the present invention is by carrying out classification analysis to the dummy node in virtual network requests, the virtual net string bag request being divided into scale less the virtual network of request according to classification results also divides the management entity tasking each region to carry out the mapping process walked abreast, thus reduce computation complexity, decrease mapping time.The management entity of regional is when the request of maps virtual network, assist mapping by the equally loaded ability introducing equally loaded demand and bottom physical network that resources balance mechanism calculates the request of virtual net string bag, thus achieve the load balancing of bottom-layer network resource.In district management entity, the extensive distribution avoiding node mapping is integrated in the coordination that dummy node maps and virtual link maps, and decreases the bottom physical path length of link maps.

It should be noted last that, above embodiment is only in order to illustrate technical scheme of the present invention and unrestricted.Although with reference to embodiment to invention has been detailed description, those of ordinary skill in the art is to be understood that, modify to technical scheme of the present invention or equivalent replacement, do not depart from the spirit and scope of technical solution of the present invention, it all should be encompassed in the middle of right of the present invention.

Claims (8)

1., based on a mapping method of virtual network for classification, described method comprises:

Step 101) when receiving virtual network requests, the dummy node involved by virtual network requests is classified according to criteria for classification, the dummy node belonged to a different category is split in different virtual net string bag requests;

Wherein, namely dummy node is categorized into from the bottom physical network in its nearest region according to dummy node to the demand information in geographical position by described criteria for classification; All dummy nodes in the request of same virtual net string bag belong to same classification, and namely this classification is the classification belonging to any one dummy node in this virtual net string bag request;

Step 102) be sent to each virtual net string bag request belonging to it bottom physical network in certain region;

Step 103) according to the equally loaded ability of each node in the equally loaded demand of each node in the request of virtual net string bag and region, the request of virtual net string bag to be mapped in region on bottom physical layer network; Wherein,

Described equally loaded requirement representation when load balancing, the loading demand of each dummy node; The resource that described loading demand is defined as dummy node request and the product of resource of all link request being connected to described dummy node;

When described equally loaded capability list is shown in load balancing, the load capacity of each physical node in region; The available resources that described load capacity is defined as physical node and the product of available resources of all links being connected to described physical node.

2. the mapping method of virtual network based on classification according to claim 1, is characterized in that, described step 103) comprise further:

Step 103-1) detect the current unmapped dummy node n with maximum equally loaded demand in the request of virtual net string bag ^v;

Step 103-2) find dummy node n in this region ^vk candidate's bottom physical node, k described both candidate nodes is current also unmapped and can meet described dummy node n in this region ^vthe bottom physical node of front k maximum equally loaded ability of demand;

Step 103-3) to each candidate's bottom physical node, calculate the path sum between this candidate's bottom physical node and the bottom physical node corresponding to each dummy node mapped, by described dummy node n ^vbe mapped to and have in the both candidate nodes of shortest path length, and the virtual link between the dummy node mapped is mapped.

3. the mapping method of virtual network based on classification according to claim 1 and 2, is characterized in that, described equally loaded ability adopts following acquisition:

Step 201) physical node of the regional that bottom physical network comprises and the resource of physical link are normalized respectively;

Wherein, the memory capacity of normalized and the node of physical node, magnetic disc i/o, CPU quantity are relevant with memory size; The normalized of physical link is relevant to the bandwidth of link, delay and delay jitter;

Step 202) calculate the load capacity of the node in bottom physical network in regional;

Wherein, the described load capacity available resources that are defined as physical node and the product of available resources of all links being connected to described physical node;

Step 203) calculate the equally loaded matrix of regional in bottom physical network;

Wherein, each element in described equally loaded matrix is: the weighted value of the mutual contribution of the load balance ability between each physical node in certain region, and the weighted value of the mutual contribution of described load balance ability is specially: as physical node n ^sbe exactly physical node m ^sitself, i.e. n ^s=m ^s, or physical node n ^swith physical node m ^sbetween when having physical link to connect, physical node n ^sto physical node m ^sequally loaded ability have contribution, now weighted value is the positive number of a non-zero; Otherwise, physical node n ^sto physical node m ^sthe contribution weight of equally loaded ability be 0; n ^sn-th of representing matrix ^soK, m ^sthe m of representing matrix ^srow, the line number of matrix and columns are all total numbers of physical node, and what footmark s represented is bottom physical network;

Step 204) according to equally loaded matrix, iterative computation goes out the equally loaded ability of each physical node in region, i.e. physical node n ^sload balance ability be the contribution sum of the equally loaded ability of all physical nodes in this region.

4. the mapping method of virtual network based on classification according to claim 1 or 4, is characterized in that, described equally loaded demand is adopted and obtained with the following method:

Step 301) demand of the dummy node in each virtual net string bag request and virtual link is normalized respectively;

Wherein, the memory capacity of normalized and the node of dummy node, magnetic disc i/o, CPU quantity are relevant with memory size demand; The normalized of virtual link and the bandwidth of link, delay are relevant with delay jitter demand;

Step 302) calculate the loading demand of each dummy node in each virtual net string bag request;

Wherein, the described loading demand demand that is defined as dummy node and the product of demand of all links being connected to described dummy node;

Step 303) calculate the equally loaded matrix of each virtual net string bag request;

Wherein, each element in described equally loaded matrix is: the interactional weighted value of the equally loaded demand between each dummy node in the request of virtual net string bag, and this weighted value is specially: as dummy node n ^vbe exactly dummy node m ^vitself, i.e. n ^v=m ^v, or dummy node n ^vwith dummy node m ^vbetween when having virtual link to connect, definition weighted value is the numerical value of a non-zero; Otherwise n ^vto m ^vthe weighing factor of equally loaded demand be 0;

N ^vn-th of representing matrix ^voK, m ^vthe m of representing matrix ^vrow, the line number of matrix and columns are all total numbers of dummy node, and what footmark v represented is virtual network requests;

Step 304) according to equally loaded matrix, iterative computation goes out the equally loaded demand of each dummy node in region, i.e. dummy node n ^vload balancing demand be the equally loaded demand of all dummy nodes in this virtual net string bag request affect sum.

5., based on a virtual network mapped system for classification, it is characterized in that, described system comprises:

Global administration's module, during for receiving virtual network requests, dummy node involved by virtual network requests is classified according to criteria for classification, by in the dummy node segmentation to different virtual net string bag request that belongs to a different category, described criteria for classification refers to that a dummy node will belong to the classification at place, region nearest from it in bottom physical network; And certain region each virtual subnetwork request is sent in the bottom physical network belonging to it; With

District management module, for the equally loaded ability according to each node in the equally loaded demand of each node in the request of virtual net string bag and region, to be mapped in region on bottom physical layer network by the request of virtual net string bag;

Wherein, described equally loaded requirement representation when load balancing, the loading demand of each dummy node; The resource that described loading demand is defined as dummy node request and the product of resource of all link request being connected to described dummy node;

When described equally loaded capability list is shown in load balancing, the load capacity of each physical node in region; The available resources that described load capacity is defined as physical node and the product of available resources of all links being connected to described physical node.

6. the virtual network mapped system based on classification according to claim 5, it is characterized in that, described district management module comprises further:

Equally loaded ability or equally loaded demand obtain submodule, for obtaining the equally loaded ability value of each node in each region of bottom physical network and obtaining the equally loaded requirements of each dummy node of each virtual net string bag request;

Detection sub-module, for detecting the current unmapped dummy node n with maximum equally loaded demand in the request of virtual net string bag ^v;

Candidate's bottom physical node chooser module, for finding dummy node n in certain region ^vk candidate's bottom physical node, k described both candidate nodes is current also unmapped and can meet described dummy node n in this region ^vthe bottom physical node of front k maximum equally loaded ability of demand; With

Node and link maps submodule, for each candidate's bottom physical node, calculate the path sum between this candidate's bottom physical node and the bottom physical node corresponding to each dummy node mapped, by described dummy node n ^vbe mapped to and have in the both candidate nodes of shortest path length, and the virtual link between the dummy node mapped is mapped.

7. the virtual network mapped system based on classification according to claim 6, is characterized in that, described equally loaded ability obtains submodule and comprises further:

First normalized module, is normalized respectively for the physical node of regional that comprises bottom physical network and the resource of physical link;

Wherein, the memory capacity of normalized and the node of physical node, magnetic disc i/o, CPU quantity are relevant with memory size; The normalized of physical link is relevant to the bandwidth of link, delay and delay jitter;

First node load capacity acquisition module, for calculating the load capacity of the node in bottom physical network in regional;

Wherein, the described load capacity available resources that are defined as physical node and the product of available resources of all links being connected to described physical node;

Equally loaded matrix acquisition module, for calculating the equally loaded matrix of regional in bottom physical network;

Wherein, the mutual contribution weighted value of the equally loaded ability between each physical node in the described equally loaded defined matrix region of bottom physical network, and if only if physical node n ^sbe exactly physical node m ^sitself, i.e. n ^s=m ^s, or n ^sand m ^sbetween have physical link to connect, now weighted value is the positive number of a non-zero; Otherwise physical node n ^sto physical node m ^sthe contribution weight of equally loaded ability be 0; With

Equally loaded capacity calculation module, for according to equally loaded matrix, iterative computation goes out the equally loaded ability of each physical node in region, i.e. physical node n ^sload balance ability be the contribution sum of the equally loaded ability of all physical nodes in this region.

8. the virtual network mapped system based on classification according to claim 6, is characterized in that, described equally loaded demand obtains submodule and comprises further:

Second normalized module, for being normalized respectively the demand of the dummy node in each virtual net string bag request and virtual link;

Wherein, the memory capacity of normalized and the node of dummy node, magnetic disc i/o, CPU quantity are relevant with memory size demand; The normalized of virtual link and the bandwidth of link, delay are relevant with delay jitter demand;

Second loading demand computing module, for calculating the loading demand of each dummy node in each virtual net string bag request;

Wherein, the described loading demand demand that is defined as dummy node and the product of demand of all links being connected to described dummy node;

Second equally loaded matrix, for calculating the equally loaded matrix of each virtual net string bag request;

Wherein, the interactional weighted value of the equally loaded demand between each dummy node in the virtual net string bag request of described equally loaded defined matrix, and if only if dummy node n ^vbe exactly dummy node m ^vitself, i.e. n ^v=m ^v, or dummy node n ^vwith dummy node m ^vbetween have virtual link to connect, weighted value is now the positive number of a non-zero; Otherwise dummy node n ^vto dummy node m ^vthe weighing factor of equally loaded demand be 0; With

Equally loaded Requirement Acquisition module, for according to equally loaded matrix, iterative computation goes out the equally loaded demand of each dummy node in region, i.e. dummy node n ^vload balancing demand be the equally loaded demand of all dummy nodes in this virtual net string bag request affect sum.