CN105262663B - A kind of cross-domain mapping method of mixing virtual network - Google Patents

A kind of cross-domain mapping method of mixing virtual network Download PDF

Info

Publication number
CN105262663B
CN105262663B CN201510510671.7A CN201510510671A CN105262663B CN 105262663 B CN105262663 B CN 105262663B CN 201510510671 A CN201510510671 A CN 201510510671A CN 105262663 B CN105262663 B CN 105262663B
Authority
CN
China
Prior art keywords
node
mapping
domain
cost
multicast
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Fee Related
Application number
CN201510510671.7A
Other languages
Chinese (zh)
Other versions
CN105262663A (en
Inventor
孙罡
杨广华
廖丹
向兵
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Electronic Science and Technology of China
China Mobile Group Design Institute Co Ltd
Original Assignee
University of Electronic Science and Technology of China
China Mobile Group Design Institute Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by University of Electronic Science and Technology of China, China Mobile Group Design Institute Co Ltd filed Critical University of Electronic Science and Technology of China
Priority to CN201510510671.7A priority Critical patent/CN105262663B/en
Publication of CN105262663A publication Critical patent/CN105262663A/en
Application granted granted Critical
Publication of CN105262663B publication Critical patent/CN105262663B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Landscapes

  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

The invention discloses a kind of cross-domain mapping methods of mixing virtual network, in the case where a known bottom multiple-domain network and mixing virtual network (HVN) are asked, by based on HVNMMD D-algorithms and based on the mapping method of pattern decomposition, the node computing resource and link bandwidth resource distributed in bottom-layer network give mixing virtual network requests, to have found a kind of mapping scheme not only having met QoS but also save mapping cost.Secondly, the particularity of multicast portions in multiple domain bottom-layer network and hybrid network has been considered in the present invention, optimization has been done to common bandwidth resources and computing resource to meet the development need of current mixing virtual network cross-domain mapping.

Description

A kind of cross-domain mapping method of mixing virtual network
Technical field
The invention belongs to Internet communication technology fields, more specifically, be related to it is a kind of mixing virtual network it is cross-domain Mapping method.
Background technology
Internet has been achieved for huge achievement as the acquisition of information of the world today and the effective means of exchange.In mistake In the decades gone, the continuous development of network technology and multifarious increase of the Internet, applications make network environment increasingly It is good, and confirm its value.But the fast development of internet also causes prodigious burden to network so that existing net Network framework cannot carry so many application well.And due to the multi-provider characteristic of network, each supply of the coordinating and unifying Quotient is difficult to do new adjustment to bottom-layer network structure.Therefore, network of today can only structure limited by multi-provider It can only carry out slowly and simply changing, extensive quickly innovation cannot be carried out.
Network virtualization technology is as the effective means for solving internet bottleneck problem, recently by the wide of scholar and enterprise General concern.Network virtualization technology is allowed in the network application that isomery is disposed in shared network bottom layer without changing network Bottom architecture.Therefore network application is enriched, while maintains the overall architecture of existing network.The essence of network virtualization is exactly By the computing resource of bottom physical network nodes and the bandwidth resources of link by being abstracted into the virtual resource that can be dynamically distributed, The physical network resource of isolation distributed to each network application and different virtual network frameworks is allow to share bottom.Network Resource has following features following characteristics under virtualized environment:
(1), isomerism:Virtual resource type in network virtualization environment is various, Various Functions, access configuration mode, Difference between local management system operation, shared rule is very big.
(2), distributivity:Virtual resource is distributed the different places on position in the ground, is under the jurisdiction of multiple infrastructure supplies Quotient.
(3), autonomy:Owner of the infrastructure supplier as virtual resource has five-star management to resource Permission, and there is autonomous managerial ability.
(4), scalability:On the one hand, due to the needs of Facilities Construction, existing infrastructure provider can increase new net Network device resource expands network size;On the other hand, due to the demand of business development, virtual net may may require that new base Infrastructure provider includes to management framework.
(5), dynamic:In network virtualization environment, virtual resource can be dynamically joined to or leave system, the institute of resource Locate the dynamic change over time such as position, service provision capacity, load, it is also possible to equipment physical fault occur and lead The situation for causing resource inaccessible.
Bottom-layer network resource it is shared extensively be network virtualization main purpose.Therefore the mapping of virtual network resource is The emphasis of network virtualization technology is also difficult point.Virtual resource mapping algorithm as network virtualization technology critical issue it One, it realizes the process on the physical resource that the virtual network requests of user are reasonably mapped to bottom physical network, In how efficiently to distribute physical network resource to meet link bandwidth and the joint behavior requirement of each virtual network, be virtual network The key of mapping problems.
In existing network application, communication form has the forms such as unicast, multicast, broadcast, mixing virtual network.It is wherein single It broadcasts and has been widely used in many real-time applications for needing high QoS with multicast, however in virtual network, existing majority Virtual resource mapping algorithm is only effective to unicast service or multicast service.Single unicast or multicast mapping algorithm is simultaneously not suitable for In mix virtual network the case where, while the resource of a bottom physical network supplier is possible to not carry entire virtual net Network asks or since position constraint needs several physical network suppliers cooperation to dispose virtual network requests, it is therefore desirable to carry out The considerations of multiple domain.
Invention content
It is an object of the invention to overcome the deficiencies of the prior art and provide a kind of cross-domain mapping sides of mixing virtual network Method, in the case of a known bottom multiple-domain network and mixing virtual network HVN request, find out it is a kind of meeting QoS, and subtract The mapping scheme of small mapping cost.
For achieving the above object, a kind of mixing virtual network cross-domain mapping side based on HVNMMD-D algorithms of the present invention Method, which is characterized in that include the following steps:
(1), HVN requests are split
According to the attribute of dummy node and virtual link in mixing virtual network, HVN requests are split into multicast network and are asked MVN and unicast networks is asked to ask UVN;
(2), the cross-domain mapping of multicast network request MVN
(2.1), mapping costs of the estimation MVN in each physical domain
(2.1.1), mapping cost in the domain of multicast root node is estimated
Cost(v→ns)=(cst (ns)+(MC-Con(ns)))*req(v) (1)
Wherein, Con (ns) it is physical node nsAdjacent node number, MC is maximum in all nodes of bottom-layer network in domain Con (ns), req (v) is resource quantity needed for MVN root nodes v, cst (ns) it is physical node nsResource unit price;
According to formula (1), candidate mappings node of the physical node of cost minimization as MVN root nodes v is chosen;
(2.1.2), mapping cost in the domain of multicast leaf node is estimated
Cost(vL→nk)+Cost(p) (2)
Wherein
Cost(vL→nk)=cst (nk)*req(vL) (3)
Wherein, cst (nk) it is physical node nkResource unit price, req (vL) it is multicast leaf node vLRequired number of resources Amount, Cost (vL→nk) it is multicast leaf node vLIn physical node nkOn mapping spend;
lvTo connect multicast root node v and multicast leaf node vLVirtual link, p be virtual link lvIn physical network In mapping path, cst (ls) it is physical link lsResource unit price, req (lv) it is resource needed for virtual link, Cost (p) is Virtual link lvMapping cost;
According to formula (2), the physical node n of mapping least cost is chosensAs MVN leaf nodes vLCandidate mappings section Point, the path p between corresponding node are virtual link lvCandidate mappings path;
(2.2), the cross-domain mapping of multicast network request MVN
(2.2.1), global view is built to multicast network
It is empty first to initialize global figure, then the both candidate nodes in each domain and path candidate are added in global figure, and root Increase cross-domain path candidate according to the virtual link connection relation in virtual network according to the both candidate nodes in not same area;
(2.2.2), optimization modeling is carried out according to global view
Optimization aim is:
Constraints is:
Wherein, ldFor the path candidate in global figure, nd、mdFor ldTwo endpoints.x(nd) it is binary variable, work as time Select node ndIt is 1 to be selected then, is otherwise 0, y (ld) it is binary variable, as path candidate ldIt is 1 to be selected then, is otherwise 0, cad(nd) it is to be mapped in ndOn dummy node, cad (ld) it is to be mapped in ldOn virtual link, f (nd) it is reflecting for both candidate nodes Penetrate cost f (nd)=Cost (cad (nd)→nd), f (ld) be path candidate mapping cost f (ld)=Cost (ld),It is more Dummy node is broadcast,For multicast virtual link,For the both candidate nodes set in global figure,For the candidate road in global figure Diameter set,For multicast node set,For multicast link set;
(2.2.3), model solution
First to the variable y (l in modeld)、x(nd) integer relaxation is carried out, recycle cplex to be solved, in the knot of solution In fruit, all-multicast dummy node is traversed, x (n are selected for each multicast virtual noded) maximum both candidate nodes ndAs final Mapping node, select ndBetween path candidate be final mapping path, and with this carry out multicast network ask MVN across Domain mapping;
(3), the cross-domain mapping of unicast networks request UVN
(3.1), unmapped unicast dummy node v in UVN is estimatedADomain in map cost
Unmapped unicast dummy node v is chosen from UVNA, then estimate unicast dummy node vADomain in map cost;
Cost(vA→nk)+Cost(D) (7)
Wherein
Cost(vA→nk) can be calculated according to formula (2);
Wherein, Cost (D) is unicast dummy node vAWith the dummy node being mapped in unicast networks in same physical domain Between virtual link set mapping spend, P is the mapping path set that virtual link is integrated into physical network, and p is one in P Path;
The physical node n of minimum mapping cost is chosen according to formula (7)kAs unicast dummy node vABoth candidate nodes, it is right It is the path candidate of virtual link to answer the path between node;
(3.2), the cross-domain mapping of unicast networks request UVN
(3.2.1), global view is built to unicast networks
It is empty first to initialize global figure, then the both candidate nodes in each domain selected in step (3.1) and path candidate are added It is added in global figure, and cross-domain according to the virtual link connection relation increase in virtual network according to the both candidate nodes in not same area Path candidate;
(3.2.2), optimization modeling is carried out according to global view
Optimization aim is:
Constraints is:
Wherein, ldFor the path candidate in global figure, nd、mdFor the both candidate nodes in global figure and it is ldTwo endpoints, x(nd) it is binary variable, as both candidate nodes ndIt is 1 to be selected then, is otherwise 0, y (ld) it is binary variable, work as path candidate ldIt is 1 to be selected then, is otherwise 0, cad (nd) it is to be mapped in ndOn dummy node, cad (ld) it is to be mapped in ldOn virtual chain Road, f (nd) be both candidate nodes mapping cost f (nd)=Cost (cad (nd)→nd), f (ld) be path candidate mapping cost f (ld)=Cost (ld),For unicast dummy node,For unicast virtual link,For the node set in global figure,For Set of paths in global figure.For unicast networks node set,For unicast networks link set;
(3.2.3), model solution
First to the variable y (l in modeld)、x(nd) integer relaxation is carried out, recycle cplex to be solved, in the knot of solution In fruit, all unicast dummy nodes are traversed, x (n are selected for each unicast dummy noded) maximum both candidate nodes ndAs final Mapping node, select ndBetween path candidate be final mapping path, and with this carry out unicast networks ask UVN across Domain mapping.
Further, the mixing virtual network cross-domain mapping method based on pattern decomposition that the present invention also provides a kind of, it is special Sign is, includes the following steps:
(1), the spectrum segmentation of HVN requests
(1.1), adjacency matrix is established to HVN requests
If An×nFor the adjacency matrix of single HVN requests, wherein n is dummy node number in HVN requests, and each virtual Nodal community, i.e. multicast or unicast are carried in node, A (i, j) indicates the void between i-th of dummy node and j-th of dummy node Quasi- link bandwidth demand, wherein i, j ∈ [1, n];
(1.2), matrix D is builtn×n, and meet:
(1.3), matrix B is builtn×n, Bn×n=Dn×n-An×n
(1.4), calculating matrix Bn×nPreceding k characteristic value and corresponding feature vector, wherein the maximum occurrences of k be matrix Bn×nCharacteristic value total number;A feature space is constituted with feature vector again, each vector in feature space is by it in feature The position occurred in vector represents some dummy node, is clustered to the vectorial K-means in feature space, clusters acquired results It is exactly the dummy node set of each subgraph;
(2), the candidate domain of subgraph is chosen
(2.1), subgraph is ranked up
The attribute for traversing all dummy nodes in subgraph, the subgraph containing multicast root node is made number one, containing more The subgraph for broadcasting leaf node comes thereafter successively, finally comes the subgraph for containing only unicast node finally successively;
(2.2), cost is mapped in the domain of estimation subgraph
If containing multicast root node in subgraph, its cost is estimated according to formula (1), then chooses mapping Least-cost Physical node is as mapping node in the domain of multicast root node;
If containing multicast leaf node in subgraph, estimate its domain in that mapping cost, link reflects in domain according to formula (2) Cost is penetrated, then chooses the physical node of mapping Least-cost as mapping node in the domain of multicast leaf node;
If containing unicast dummy node in subgraph, mapping cost in its domain first is calculated according to formula (7), recycles formula (10) current subgraph virtual unicast node v is calculateduWith mapped completion subgraph between link maps cost.
Wherein L be and node vuThere is link set between the subgraph of connection relation.
Then dummy node vuTotal mapping cost be
Const(vu→nk)+Cost(M)+Cost(D) (11)
The physical node of mapping cost minimization is chosen as mapping node in the domain of unicast dummy node according to formula (11).
(3), the cross-domain mapping of subgraph
According to the mapping process of step (2.2), the sum of mapping cost of all nodes in each subgraph is calculated separately, every In a subgraph, final mapping domain of the domain of mapping cost minimization as the subgraph, corresponding mapping node and mapping are chosen respectively Path is final mapping node and path, to complete the cross-domain mapping of all subgraphs.
What the goal of the invention of the present invention was realized in:
A kind of cross-domain mapping method of mixing virtual network, in a known bottom multiple-domain network and mixing virtual network (HVN) in the case of asking, by based on HVNMMD-D algorithms and based on the mapping method of pattern decomposition, distributing in bottom-layer network Node computing resource and link bandwidth resource give mixing virtual network requests, a kind of not only having met QoS to have found but also saved Map the mapping scheme of cost.Secondly, multicast portions in multiple domain bottom-layer network and hybrid network have been considered in the present invention Particularity, optimization has been done to common bandwidth resources and computing resource to meet the hair of current mixing virtual network cross-domain mapping Exhibition needs.
Meanwhile the cross-domain mapping method of present invention mixing virtual network also has the advantages that:
(1), applied widely
Traditional multiple domain mapping policy is in unicast virtual network requests or multicast virtual network request mostly What a kind of situation proposed, the case where mixing virtual network requests can not be suitable under normal conditions;And this method is applicable not only to Unicast virtual network requests and the simultaneous mixing virtual network requests of multicast virtual network request, and it is empty suitable for unicast The service request of quasi- network request or multicast virtual network request.Therefore compared with traditional mapping policy, this method is fitted It is wider with range.
(2), resource utilization is high
Since the present invention is when mapping HVN requests, the preferential multicast portions mapped in HVN, and reflecting in guiding root node During penetrating, while using the number of degrees of bottom-layer network node, the surplus resources of node and resource unit price as inducible factor, It can thus make bottom-layer network load more balanced, and receive more HVN requests, increase the utilization of bottom-layer network resource Rate increases income.
Description of the drawings
Fig. 1 is the fractionation schematic diagram of HVN requests;
Fig. 2 is the multicast network request mapping schematic diagram based on HVNMMD-D algorithms;
Fig. 3 is the unicast networks request mapping schematic diagram based on HVNMMD-D algorithms;
Fig. 4 is the HVN request mapping schematic diagrames based on pattern decomposition.
Specific implementation mode
The specific implementation mode of the present invention is described below in conjunction with the accompanying drawings, preferably so as to those skilled in the art Understand the present invention.Requiring particular attention is that in the following description, when known function and the detailed description of design perhaps When can desalinate the main contents of the present invention, these descriptions will be ignored herein.
Embodiment
For the convenience of description, first being illustrated to the relevant speciality term occurred in specific implementation mode:
HVNMMD-D(decomposition-based algorithm for hybrid virtual network Mapping across multiple domains) based on the mixing virtual network mapping algorithm for decomposing thought;
HVN (hybrid virtual network) mixes virtual network;
MVN (multicast virtual network) multicast virtual network;
UVN (unicast virtual network) unicast virtual network;
In the present invention, a kind of cross-domain mapping method of mixing virtual network includes mainly two kinds, i.e.,:Based on HVNMMD-D Algorithm and mixing virtual network cross-domain mapping method based on pattern decomposition, here, we are first to the virtual section in two kinds of algorithms Point is illustrated:Dummy node includes multicast virtual node and unicast dummy node, and multicast virtual node includes multicast root section again Point and multicast leaf node.Two methods are described in detail below.
One, the mixing virtual network cross-domain mapping method based on HVNMMD-D algorithms, the specific steps implemented include:
S1, HVN requests are split
According to the attribute of dummy node and virtual link in mixing virtual network, HVN requests are split into multicast network and are asked MVN and unicast networks is asked to ask UVN;
In the present embodiment, as shown in Fig. 1 (b), node a, b, c and inter-node link (a, b), (a, c) have simultaneously The characteristic of unicast networks and multicast network, as multicast network ask MVN;As shown in Fig. 1 (c), remaining part only has unicast Network characteristic, as unicast networks ask UVN;As shown in Fig. 1 (a), MVN and unicast networks is asked to ask UVN by multicast network Collectively constitute mixing virtual network.
The cross-domain mapping of S2, multicast network request MVN
S2.1, MVN is first subjected to preliminary mapping in each physical domain
S2.1.1, mapping cost in the domain of multicast root node is estimated
Cost(v→ns)=(cst (ns)+(MC-Con(ns)))*req(v) (1)
Wherein, Con (ns) it is physical node nsAdjacent node number, MC is maximum in all nodes of bottom-layer network in domain Con (ns), req (v) is resource quantity needed for MVN root nodes v, cst (ns) it is physical node nsResource unit price;
According to formula (1), candidate mappings node of the physical node of cost minimization as MVN root nodes v is chosen;
In the present embodiment, as shown in Fig. 2 (a), three physical domains is chosen and carry out preliminary mapping, estimated by the above method more Mapping costs of the root node a in three physical domains is broadcast, as shown in Fig. 2 (c), finds a time in domain 1, domain 2, domain 3 respectively Select mapping node;
S2.1.2, mapping cost in the domain of multicast leaf node is estimated
Cost(vL→nk)+Cost(p) (2)
Wherein
Cost(vL→nk)=cst (nk)*req(vL) (3)
Wherein, cst (nk) it is physical node nkResource unit price, req (vL) it is multicast leaf node vLRequired number of resources Amount, Cost (vL→nk) it is multicast leaf node vLIn physical node nkOn mapping spend;
lvTo connect multicast root node v and multicast leaf node vLVirtual link, p be virtual link lvIn physical network In mapping path, cst (ls) it is physical link lsResource unit price, req (lv) it is resource needed for virtual link, Cost (p) is Virtual link lvMapping cost;
According to formula (2), the physical node n of mapping least cost is chosensAs MVN leaf nodes vLCandidate mappings section Point, the path p between corresponding node are virtual link lvCandidate mappings path;
In the present embodiment, estimate multicast leaf node b, c in Fig. 2 (b) in three physics respectively according to the method described above Mapping cost in domain, as shown in Fig. 2 (c), respectively multicast leaf node b, c find a candidate in domain 1, domain 2, domain 3 Mapping node, the path between corresponding node is candidate mappings path;
The cross-domain mapping of S2.2, multicast network request MVN
S2.2.1, global view is built to multicast network
It is empty first to initialize global figure, then the both candidate nodes in each domain and path candidate are added in global figure, and root Increase cross-domain path candidate according to the virtual link connection relation in virtual network according to the both candidate nodes in not same area;
In this embodiment, according to the above method, first by Fig. 2 (c) both candidate nodes and link be added to global figure In, wherein both candidate nodes in 1~domain of domain 3 corresponding points in global figure are a1~a3, other both candidate nodes also do identical change It changes, then increases cross-domain path candidate according to the connection relation of 2 (b) interior joint, shown in result such as Fig. 2 (d);
S2.2.2, optimization modeling is carried out according to global view
Optimization aim is:
Constraints is:
Wherein, ldFor the path candidate in global figure, nd、mdFor ldTwo endpoints.x(nd) it is binary variable, work as time Select node ndIt is 1 to be selected then, is otherwise 0, y (ld) it is binary variable, as path candidate ldIt is 1 to be selected then, is otherwise 0, cad(nd) it is to be mapped in ndOn dummy node, cad (ld) it is to be mapped in ldOn virtual link, f (nd) it is reflecting for both candidate nodes Penetrate cost f (nd)=Cost (cad (nd)→nd), f (ld) be path candidate mapping cost f (ld)=Cost (ld),It is more Dummy node is broadcast,For multicast virtual link,For the both candidate nodes set in global figure,For the candidate road in global figure Diameter set,For multicast node set,For multicast link set;
S2.2.3, model solution
First to the variable y (l in modeld)、x(nd) integer relaxation is carried out, recycle cplex to be solved, in the knot of solution In fruit, all-multicast dummy node is traversed, x (n are selected for each multicast virtual noded) maximum both candidate nodes ndAs final Mapping node, select ndBetween path candidate be final mapping path, and with this carry out multicast network ask MVN across Domain mapping;
In the present embodiment, model solution, solving result Fig. 2 are carried out to the global figure in Fig. 2 (d) according to the above method (e) shown in, i.e., multicast network is asked in MVN cross-domain mappings to domain 2;
The cross-domain mapping of S3, unicast networks request UVN
Unmapped unicast dummy node v in S3.1, estimation UVNADomain in map cost
Unmapped unicast dummy node v is chosen from UVNA, then estimate unicast dummy node vADomain in map cost;
Cost(vA→nk)+Cost(D) (7)
Wherein
Cost(vA→nk) can be calculated according to formula (2);
Wherein, Cost (D) is unicast dummy node vAWith the dummy node being mapped in unicast networks in same physical domain Between virtual link set mapping spend, P is the mapping path set that virtual link is integrated into physical network, and p is one in P Path;
The physical node n of minimum mapping cost is chosen according to formula (7)kAs unicast dummy node vABoth candidate nodes, it is right It is the path candidate of virtual link to answer the path between node;
In the present embodiment, three physical domains are equally chosen according to the above method, are estimated in unicast virtual network graph 3 (b) Mapping cost in domain 1, domain 2, domain 3 of non-mapping node d, e, and find in domain 1, domain 2 the candidate mappings section of node d, e Mapping path between point and node, as shown in Fig. 3 (c);
The cross-domain mapping of S3.2, unicast networks request UVN
S3.2.1, global view is built to unicast networks
It is empty first to initialize global figure, then the both candidate nodes in each domain and path candidate are added in global figure, and root Increase cross-domain path candidate according to the virtual link connection relation in virtual network according to the both candidate nodes in not same area;
In the present embodiment, according to the above method, by Fig. 3 (c) both candidate nodes and alternative link be added to global figure In, wherein corresponding points is are respectively d1, e1 and d2, e2 to the both candidate nodes in domain 1 and domain 3 in global figure, then according to 3 (b) connection relation of interior joint increases cross-domain path candidate, shown in result such as Fig. 3 (d);
S3.2.2, optimization modeling is carried out according to global view
Optimization aim is:
Constraints is:
Wherein, ldFor the path candidate in global figure, nd、mdFor the both candidate nodes in global figure and it is ldTwo endpoints, x(nd) it is binary variable, as both candidate nodes ndIt is 1 to be selected then, is otherwise 0, y (ld) it is binary variable, work as path candidate ldIt is 1 to be selected then, is otherwise 0, cad (nd) it is to be mapped in ndOn dummy node, cad (ld) it is to be mapped in ldOn virtual chain Road, f (nd) be both candidate nodes mapping cost f (nd)=Cost (cad (nd)→nd), f (ld) be path candidate mapping cost f (ld)=Cost (ld),For unicast dummy node,For unicast virtual link,For the node set in global figure,For Set of paths in global figure.For unicast networks node set,For unicast networks link set;
S3.2.3, model solution
First to the variable y (l in modeld)、x(nd) integer relaxation is carried out, recycle cplex to be solved, in the knot of solution In fruit, all unicast dummy nodes are traversed, x (n are selected for each unicast dummy noded) maximum both candidate nodes ndAs final Mapping node, select ndBetween path candidate be final mapping path, and with this carry out unicast networks ask UVN across Domain mapping;
In the present embodiment, model solution, solving result Fig. 3 are carried out to the global figure in Fig. 3 (d) according to the above method (e) shown in, i.e., unicast networks are asked in UVN cross-domain mappings to domain 1.
Two, the mixing virtual network cross-domain mapping method based on pattern decomposition, the specific steps implemented include:
The spectrum segmentation of T1, HVN request
T1.1, adjacency matrix is established to HVN requests
If An×nFor the adjacency matrix of single HVN requests, wherein n is dummy node number in HVN requests, and each virtual Nodal community, i.e. multicast or unicast are carried in node, A (i, j) indicates the void between i-th of dummy node and j-th of dummy node Quasi- link bandwidth demand, wherein i, j ∈ [1, n];
T1.2, structure matrix Dn×n, and meet:
T1.3, structure matrix Bn×n, Bn×n=Dn×n-An×n
T1.4, calculating matrix Bn×nPreceding k characteristic value and corresponding feature vector, wherein the maximum occurrences of k be matrix Bn×nCharacteristic value total number;A feature space is constituted with feature vector again, each vector in feature space is by it in feature The position occurred in vector represents some dummy node, is clustered to the vectorial K-means in feature space, clusters acquired results It is exactly the dummy node set of each subgraph;
In the present embodiment, the multicast hybrid network in Fig. 1 is split into subgraph as shown in Figure 4 according to the method described above (a) and subgraph (b);
T2, the candidate domain for choosing subgraph
T2.1, subgraph is ranked up
The attribute for traversing all dummy nodes in subgraph, the subgraph containing multicast root node is made number one, containing more The subgraph for broadcasting leaf node comes thereafter successively, finally comes the subgraph for containing only unicast node finally successively;
T2.2, mapping cost in the domain of subgraph is estimated
If containing multicast root node in subgraph, its cost is estimated according to formula (1), then chooses mapping Least-cost Physical node is as mapping node in the domain of multicast root node;
If containing multicast leaf node in subgraph, estimate its domain in that mapping cost, link reflects in domain according to formula (2) Cost is penetrated, then chooses the physical node of mapping Least-cost as mapping node in the domain of multicast leaf node;
If containing unicast dummy node in subgraph, mapping cost in its domain first is calculated according to formula (7), recycles formula (10) current subgraph virtual unicast node v is calculateduWith mapped completion subgraph between link maps cost.
Wherein L be and node vuThere is link set between the subgraph of connection relation.
Then dummy node vuTotal mapping cost be
Const(vu→nk)+Cost(M)+Cost(D) (11)
The physical node of mapping cost minimization is chosen as mapping node in the domain of unicast dummy node according to formula (11);
In the present embodiment, such as Fig. 4 (a), three physical domains is chosen and carry out preliminary mapping, subgraph is estimated by the above method (a) the mapping cost of node a, b, c in three physical domains in, finds its candidate mappings section in domain 1, domain 2 and domain 3 respectively Candidate mappings path between point and node, as shown in Fig. 4 (d)
The cross-domain mapping of T3, subgraph
According to the mapping process of step T2.2, the sum of mapping cost of all nodes in each subgraph is calculated separately, every In a subgraph, final mapping domain of the domain of mapping cost minimization as the subgraph, corresponding mapping node and mapping are chosen respectively Path is final mapping node and path, to complete the cross-domain mapping of all subgraphs.
In the present embodiment, as shown in Fig. 4 (d), the minimum domain 2 of the sum of mapping cost is chosen as son according to the above method The mapping domain of (a) is schemed, shown in mapping result such as Fig. 4 (e);Repeat the mapping that step T2 and T3 complete subgraph 4 (b), such as Fig. 4 (f) shown in.
Although the illustrative specific implementation mode of the present invention is described above, in order to the technology of the art Personnel understand the present invention, it should be apparent that the present invention is not limited to the range of specific implementation mode, to the common skill of the art For art personnel, if various change the attached claims limit and determine the spirit and scope of the present invention in, these Variation is it will be apparent that all utilize the innovation and creation of present inventive concept in the row of protection.

Claims (3)

1. a kind of being based on the mixing virtual network mapping HVNMMD-D algorithms (decomposition-based for decomposing thought Algorithm for hybrid virtual network mapping across multiple domains) mixing it is empty Quasi- network cross-domain mapping method, which is characterized in that include the following steps:
(1), mixing virtual network HVN (hybrid virtual network) is split
According to the attribute of dummy node and virtual link in mixing virtual network, HVN requests are split into multicast virtual network MVN With unicast virtual network UVN;
(2), the cross-domain mapping of multicast virtual network MVN
(2.1), mapping costs of the estimation MVN in each physical domain
(2.1.1), mapping cost in the domain of multicast root node is estimated
Cost(v→ns)=(cst (ns)+(MC-Con(ns)))*req(v) (1)
Wherein, Con (ns) it is physical node nsAdjacent node number, MC is maximum Con in all nodes of bottom-layer network in domain (ns), req (v) is resource quantity needed for MVN root nodes v, cst (ns) it is physical node nsResource unit price;
According to formula (1), candidate mappings node of the physical node of cost minimization as MVN root nodes v is chosen;
(2.1.2), mapping cost in the domain of multicast leaf node is estimated
Cost(vL→nk)+Cost(p) (2)
Wherein
Cost(vL→nk)=cst (nk)*req(vL) (3)
Wherein, cst (nk) it is physical node nkResource unit price, req (vL) it is multicast leaf node vLRequired resource quantity, Cost (vL→nk) it is multicast leaf node vLIn physical node nkOn mapping spend;
lvTo connect multicast root node v and multicast leaf node vLVirtual link, p be virtual link lvIn physical network Mapping path, cst (ls) it is physical link lsResource unit price, req (lv) it is resource needed for virtual link, Cost (p) is virtual Link lvMapping cost;
According to formula (2), the physical node n of mapping least cost is chosensAs MVN leaf nodes vLCandidate mappings node, it is right It is virtual link l to answer the path p between nodevCandidate mappings path;
(2.2), the cross-domain mapping of multicast virtual network MVN
(2.2.1), to multicast virtual network struction global view
It is empty first to initialize global figure, then the both candidate nodes in each domain and path candidate are added in global figure, and according to not Both candidate nodes in same area increase cross-domain path candidate according to the virtual link connection relation in virtual network;
(2.2.2), optimization modeling is carried out according to global view
Optimization aim is:
Constraints is:
Wherein, ldFor the path candidate in global figure, nd、mdFor ldTwo endpoints, x (nd) it is binary variable, when candidate saves Point ndIt is 1 to be selected then, is otherwise 0, y (ld) it is binary variable, as path candidate ldIt is 1 to be selected then, is otherwise 0, cad (nd) it is to be mapped in ndOn dummy node, cad (ld) it is to be mapped in ldOn virtual link, f (nd) be both candidate nodes mapping Cost f (nd)=Cost (cad (nd)→nd), f (ld) be path candidate mapping cost f (ld)=Cost (ld),For multicast Dummy node,For multicast virtual link,For the both candidate nodes set in global figure,For the path candidate collection in global figure It closes,For multicast node set,For multicast link set;
(2.2.3), model solution
First to the variable y (l in modeld)、x(nd) integer relaxation is carried out, recycle cplex to be solved, in the result of solution In, all-multicast dummy node is traversed, x (n are selected for each multicast virtual noded) maximum both candidate nodes ndAs final Mapping node selects ndBetween path candidate be final mapping path, and the cross-domain of multicast virtual network MVN is carried out with this Mapping;
(3), the cross-domain mapping of unicast virtual network UVN
(3.1), unmapped unicast dummy node v in UVN is estimatedADomain in map cost
Unmapped unicast dummy node v is chosen from UVNA, then estimate unicast dummy node vADomain in map cost;
Cost(vA→nk)+Cost(D) (7)
Wherein
Cost(vA→nk) can be calculated according to formula (2);
Wherein, Cost (D) is unicast dummy node vAWith it is virtual between the dummy node that is mapped in unicast networks in same physical domain The mapping of link set is spent, and P is the mapping path set that virtual link is integrated into physical network, and p is the paths in P;
The physical node n of minimum mapping cost is chosen according to formula (7)kAs unicast dummy node vABoth candidate nodes, corresponding section Path between point is the path candidate of virtual link;
(3.2), the cross-domain mapping of unicast virtual network UVN
(3.2.1), global view is built to unicast virtual network
It is empty first to initialize global figure, then the both candidate nodes in each domain selected in step (3.1) and path candidate are added to In global figure, and increase cross-domain candidate according to the virtual link connection relation in virtual network according to the both candidate nodes in not same area Path;
(3.2.2), optimization modeling is carried out according to global view
Optimization aim is:
Constraints is:
Wherein, ldFor the path candidate in global figure, nd、mdFor the both candidate nodes in global figure and it is ldTwo endpoints, x (nd) For binary variable, as both candidate nodes ndIt is 1 to be selected then, is otherwise 0, y (ld) it is binary variable, as path candidate ldQuilt It is 1 to choose then, is otherwise 0, cad (nd) it is to be mapped in ndOn dummy node, cad (ld) it is to be mapped in ldOn virtual link, f(nd) be both candidate nodes mapping cost f (nd)=Cost (cad (nd)→nd), f (ld) be path candidate mapping cost f (ld)=Cost (ld),For unicast dummy node,For unicast virtual link,For the node set in global figure,It is complete Set of paths in office's figure.For unicast networks node set,For unicast networks link set;
(3.2.3), model solution
First to the variable y (l in modeld)、x(nd) integer relaxation is carried out, recycle cplex to be solved, in the result of solution In, all unicast dummy nodes are traversed, x (n are selected for each unicast dummy noded) maximum both candidate nodes ndAs final Mapping node selects ndBetween path candidate be final mapping path, and the cross-domain of unicast virtual network UVN is carried out with this Mapping.
2. a kind of mixing virtual network cross-domain mapping method based on pattern decomposition, which is characterized in that include the following steps:
(1), the spectrum segmentation of HVN
(1.1), adjacency matrix is established to HVN
If An×nFor the adjacency matrix of single HVN, wherein n is dummy node number in HVN, and from belt segment in each dummy node Point attribute, i.e. multicast or unicast, A (i, j) indicate that the virtual link bandwidth between i-th of dummy node and j-th of dummy node needs It asks, wherein i, j ∈ [1, n];
(1.2), matrix D is builtn×n, and meet:
(1.3), matrix B is builtn×n, Bn×n=Dn×n-An×n
(1.4), calculating matrix Bn×nPreceding k characteristic value and corresponding feature vector, wherein the maximum occurrences of k be matrix Bn×n Characteristic value total number;Constitute a feature space with feature vector again, each vector in feature space by its feature to The position occurred in amount represents some dummy node, is clustered to the vectorial K-means in feature space, and cluster acquired results are just It is the dummy node set of each subgraph;
(2), the candidate domain of subgraph is chosen
(2.1), subgraph is ranked up
The attribute for traversing all dummy nodes in subgraph, the subgraph containing multicast root node is made number one, multicast leaf is contained The subgraph of child node comes thereafter successively, finally comes the subgraph for containing only unicast node finally successively;
(2.2), cost is mapped in the domain of estimation subgraph
If containing multicast root node in subgraph, its cost is estimated according to formula (1), then chooses the physics of mapping Least-cost Node is as mapping node in the domain of multicast root node;
If containing multicast leaf node in subgraph, according to formula (2) estimate its domain in mapping cost, in domain link maps at This, then the physical node of mapping Least-cost is chosen as mapping node in the domain of multicast leaf node;
If containing unicast dummy node in subgraph, mapping cost in its domain first is calculated according to formula (7), recycles formula (10) Calculate current subgraph virtual unicast node vuWith mapped completion subgraph between link maps cost,
Wherein L be and node vuThere is link set between the subgraph of connection relation.
Then dummy node vuTotal mapping cost be
Const(vu→nk)+Cost(M)+Cost(D) (11)
The physical node of mapping cost minimization is chosen according to formula (11) as mapping node in the domain of unicast dummy node,
(3), the cross-domain mapping of subgraph
According to the mapping process of step (2.2), the sum of mapping cost of all nodes in each subgraph is calculated separately, in every height In figure, final mapping domain of the domain of mapping cost minimization as the subgraph, corresponding mapping node and mapping path are chosen respectively For final mapping node and path, to complete the cross-domain mapping of all subgraphs.
3. mixing virtual network cross-domain mapping method according to claim 1 or 2, which is characterized in that the virtual section Point includes multicast virtual node and unicast dummy node, and multicast virtual node includes multicast root node and multicast leaf node again.
CN201510510671.7A 2015-08-19 2015-08-19 A kind of cross-domain mapping method of mixing virtual network Expired - Fee Related CN105262663B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510510671.7A CN105262663B (en) 2015-08-19 2015-08-19 A kind of cross-domain mapping method of mixing virtual network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510510671.7A CN105262663B (en) 2015-08-19 2015-08-19 A kind of cross-domain mapping method of mixing virtual network

Publications (2)

Publication Number Publication Date
CN105262663A CN105262663A (en) 2016-01-20
CN105262663B true CN105262663B (en) 2018-10-16

Family

ID=55102177

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510510671.7A Expired - Fee Related CN105262663B (en) 2015-08-19 2015-08-19 A kind of cross-domain mapping method of mixing virtual network

Country Status (1)

Country Link
CN (1) CN105262663B (en)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107124306B (en) * 2017-04-24 2019-11-05 电子科技大学 Content delivery network server optimization dispositions method under network function virtualized environment
CN107395501B (en) * 2017-08-29 2020-04-14 电子科技大学 Cross-domain deployment method of network service function chain
CN107800649A (en) * 2017-10-29 2018-03-13 湖州师范学院 A kind of virtual optical network mapping method based on path of enhancing
CN108092805A (en) * 2017-12-11 2018-05-29 国网河南省电力公司信息通信公司 A kind of power telecom network cross-domain virtual network distribution mechanism based on group's optimization
CN108600101B (en) * 2018-03-21 2020-08-07 北京交通大学 Network service cross-domain arrangement method aiming at end-to-end time delay performance optimization

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102904794A (en) * 2012-09-27 2013-01-30 北京邮电大学 Method and device for mapping virtual network
CN104243263A (en) * 2014-08-27 2014-12-24 电子科技大学 Online hybrid virtual network mapping method

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102904794A (en) * 2012-09-27 2013-01-30 北京邮电大学 Method and device for mapping virtual network
CN104243263A (en) * 2014-08-27 2014-12-24 电子科技大学 Online hybrid virtual network mapping method

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
Exploring online virtual networks mapping with stochastic bandwidth demand in multi-datacenter;孙罡等;《Photon Network Communications》;20121231(第23期);第109-122页 *
The Framework and Algorithms for the Survivable Mapping of Virtual Network onto a Substrate Network;孙罡等;《IETE Technical Review》;20111031;第28卷(第5期);第381-391页 *
基于知识描述和遗传算法的跨域虚拟网络映射;肖蔼玲等;《软件学报》;20141031;第25卷(第10期);第2189-2205页 *
跨域逻辑承载网映射方法研究;张旻等;《通信学报》;20120831;第33卷(第8期);第200-207页 *

Also Published As

Publication number Publication date
CN105262663A (en) 2016-01-20

Similar Documents

Publication Publication Date Title
CN105262663B (en) A kind of cross-domain mapping method of mixing virtual network
Wang et al. Multi-agent imitation learning for pervasive edge computing: A decentralized computation offloading algorithm
Gong et al. Novel location-constrained virtual network embedding LC-VNE algorithms towards integrated node and link mapping
Houidi et al. Virtual network provisioning across multiple substrate networks
CN105103506B (en) For the method and system for the non-homogeneous bandwidth request allocation bandwidth in system for cloud computing
Kim et al. CometCloud: An autonomic cloud engine
Li et al. Resource allocation with multi-factor node ranking in data center networks
CN104461740A (en) Cross-domain colony computing resource gathering and distributing method
CN104769550B (en) The method and system of bandwidth allocation in system for cloud computing
Assis et al. Cloud federation: Characterisation and conceptual model
CN113169889B (en) Method and apparatus for mapping network slices onto a network infrastructure with SLA guarantees
CN109062687A (en) Computing resource lookup method and device for P2P network
CN107483286A (en) Merge the method with deployment services functional chain under a kind of environment based on cloud and mist
CN103384251A (en) Multi-class safety service combinable safety network construction method and device
Li et al. Cross-cloud mapreduce for big data
Miao et al. Multicast virtual network mapping for supporting multiple description coding-based video applications
Ghalehtaki et al. A bee colony-based algorithm for micro-cache placement close to end users in fog-based content delivery networks
Galvão A tabu search heuristic for the concentrator location problem
Sadegh et al. A two-phase virtual machine placement policy for data-intensive applications in cloud
CN103002067B (en) The acquisition methods of the IP address of virtual machine
CN103685285A (en) Method for limiting number of terminals under routing mode
Kang et al. A multiagent brokering protocol for supporting Grid resource discovery
CN106027396A (en) Route control method, device and system
Kumar et al. An efficient framework for resource allocation in cloud computing
CN106874108A (en) Thin cloud is minimized in mobile cloud computing use number technology

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20181016

Termination date: 20190819