CN102932479B - Virtual network mapping method for realizing topology awareness based on historical data - Google Patents

Abstract

The invention relates to a virtual network mapping method for realizing the topology awareness based on historical data. The method comprises the following operation steps: (1) according to a historical data set which is accumulated by a bottom layer physical network and is successfully mapped by a virtual network, calculating a dependency matrix M between nodes of the bottom layer physical network; (2) according to the size requirement of virtual nodes in the virtual network on central processing unit (CPU) resources, sorting all the virtual nodes of the virtual network from big to small; (3) according to the dependency matrix M, sequentially carrying out node mapping from the virtual nodes to the bottom layer physical nodes according to a certain sequence; and (4) after completing the node mapping, realizing mapping from a virtual link of the virtual network to a physical path of the bottom layer physical network according to a set link mapping method. The method disclosed by the invention realizes the scientific evaluation on the resource capacities of the bottom layer physical network and the optimization selection of the node mapping by sensing a topological structure of the virtual network, and improves the long-term average success rate of the virtual network mapping.

Description

A kind of mapping method of virtual network realizing topology ambiguity based on historical data

Technical field

The present invention relates to a kind of method realizing virtual network and map, belong to technical field of the computer network, particularly belong to network virtualization technical field.

Background technology

Network virtualization refers to and a shared physical network infrastructure is logically divided into multiple virtual network that is mutually isolated, that have heterogeneous networks topology.Virtual network generally comprises multiple dummy node and many virtual links, and each dummy node and every bar virtual link have different resource requirements, and if dummy node is to the resource requirement of central processing unit CPU, virtual link is to the demand of physical link bandwidth.Service provider SP is by renting the infrastructure section of bottom physical network, make full use of the access control power that bottom physical network infrastructure provides, can not need under the input prerequisite of carrying out the related physical network hardware, can the customed procotol of rapid deployment or framework and virtual network, provide diversified service to terminal use.

In the mapping process of virtual network to bottom physical network, owing to needing the resource requirement meeting node and link simultaneously, the mapping problems of network virtualization is a NP-hard problem.Current relative solution generally designs based on heuristic, but there are the following problems for current heuristic virtual network mapping scheme: (1) current resource standards of grading the CPU ability value of physical node are multiplied by this node adjacent link bandwidth sum as metric, but this resource standards of grading are inaccurate, cause the physical node that scheme likely selects CPU ability by force and link is weak to map, so that virtual network is mapped in the failure of link maps stage; (2) always use greedy algorithm to select the highest physical node of scoring to map, and the topology of virtual network is not considered in the position that have ignored the dummy node mapped namely.Therefore, in the process of carrying out virtual network mapping, how better bottom physical network resource ability to be evaluated, how basis maps complete dummy node and topological structure thereof, and the optimum choice realizing next step node mapping is the technical barrier that one, current Computer Network Project field urgent need will solve.

Summary of the invention

In view of this, the object of the invention is to invent a kind of method realizing virtual network and map, the historical data set that a large amount of virtual network that bottom-layer network can be utilized to accumulate successfully maps, realize the scientific evaluation to bottom physical network resource ability, and in conjunction with mapping complete dummy node and topological structure thereof, the optimum choice of next step node mapping can be realized.

In order to achieve the above object, the present invention proposes a kind of mapping method of virtual network realizing topology ambiguity based on historical data, described method comprises following operative step:

(1) according to the historical data set that the virtual network of bottom physical network accumulation successfully maps, calculate the dependence matrix M between bottom physical network nodes, specifically comprise following operating procedure: (11) all physical nodes to bottom physical network are numbered from 1, until numbering n, n is a natural number, equals the physical node number of bottom physical network; (12) from the historical data set that the virtual network of bottom physical network accumulation successfully maps, each map record is taken out; To each map record, all construct the empty matrix P of the capable n row of n, time initial, each element value of this matrix P is 0 value; In this map record, if i-th bottom physical node was at least successfully mapped by a dummy node in this map record, then allow matrix P i-th row i-th arrange element a _iivalue is 1; In this map record, if a physical pathway between i-th bottom physical node and a jth bottom physical node was at least successfully mapped by the virtual link of in this map record, then allow matrix P i-th row jth row element a _ijwith the element a that jth row i-th arranges _jiall value is the inverse of the jumping figure of this physical pathway, and wherein i and j is the natural number being more than or equal to 1, being less than or equal to n, i and j must be unequal; (13) all matrix P constructed in step 12 are carried out matrix and be added summation, obtain a new capable n column matrix S of n; (14) matrix S is normalized, obtains the dependence matrix M between bottom physical network nodes; The concrete mode of normalized is: for the element M of matrix M i-th row i-th row _iivalue is the average importance factors of this element representation bottom physical network i-th physical node; For the element M of matrix M i-th row jth row _ijvalue is the average degree of association factor between this element representation bottom physical network i-th physical node and a jth physical node; S in above-mentioned formula _iithe element of representing matrix S i-th row i-th row, S _ijthe element of representing matrix S i-th row jth row, i and j is the natural number being more than or equal to 1, being less than or equal to n, i and j must be unequal;

(2) needs are carried out to the virtual network mapped, according to the demand size of dummy node in this virtual network to cpu resource, from big to small all dummy nodes of this virtual network are sorted;

(3) according to described dependence matrix M, according to the node mapping method of setting, according to the order sequenced, the node mapping of dummy node to bottom physical node is carried out successively to the dummy node in described virtual network, specifically comprises following operating procedure: (31) take out the current dummy node also not carrying out node mapping coming foremost; (32) if this dummy node does not have father node, from described dependence matrix M, then find the current cpu resource that can meet this dummy node to require and the bottom physical node that on average importance factors is the highest, this dummy node is mapped on this physical node; If this dummy node has e father node, then first find out e the bottom physical node having correspondence mappings relation with all father nodes of this dummy node; Then from described dependence matrix M, find the bottom physical node that a current cpu resource that can meet this dummy node requires, and require this physical node to divide to be clipped to that the connection product of the average degree of association factor of described e bottom physical node is maximum, so this dummy node is mapped on this physical node; E is a natural number being more than or equal to 1; The father node of described dummy node refers to and to adjoin with this dummy node and to come the dummy node before this dummy node, and when this dummy node carries out node mapping, the father node of this dummy node completes node mapping; (33) step 31 is got back to, until all dummy nodes complete mapping;

(4) node mapping complete after, the mapping between the physical pathway of the virtual link realizing virtual network according to the link maps method of setting layer physical network on earth.

The link maps method of the setting described in described step 4 refers to k shortest path k-shortest path method.

Beneficial effect of the present invention is: mapping method of virtual network of the present invention, the historical data utilizing virtual network successfully to map achieves the scientific evaluation to bottom physical network resource ability, and can the topological structure of perception virtual network, realize the optimum choice of node mapping; Mapping method of virtual network of the present invention improves the long-term average success rate that virtual network maps effectively, brings more long-term average yield to bottom physical network infrastructure provider.

Accompanying drawing explanation

Fig. 1 is the flow chart of the inventive method.

Fig. 2 is the schematic diagram of a virtual network.

Fig. 3 is the schematic diagram of a bottom physical network.

Fig. 4 is the schematic diagram that virtual network shown in Fig. 2 is mapped to the physical network of bottom shown in Fig. 3.

Embodiment

For making the object, technical solutions and advantages of the present invention clearly, below in conjunction with accompanying drawing, the present invention is described in further detail.

See Fig. 1, introduce a kind of mapping method of virtual network realizing topology ambiguity based on historical data of the present invention, described method comprises following operative step:

(1) according to the historical data set that the virtual network of bottom physical network accumulation successfully maps, the dependence matrix M between bottom physical network nodes is calculated;

(2) needs are carried out to the virtual network mapped, according to the demand size of dummy node in this virtual network to cpu resource, from big to small all dummy nodes of this virtual network are sorted;

See Fig. 2, virtual network shown in Fig. 2 comprises 3 dummy nodes, i.e. a, b, c tri-dummy nodes, the cpu resource demand size of this dummy node of numeral in the square frame on node side, the bandwidth resources demand size of this virtual link of numeral between node on virtual link, the cpu resource demand of such as a dummy node is 10 units, and between a dummy node and b dummy node, the bandwidth resources demand of virtual link is 8 units.According to the demand size of dummy node to cpu resource, sort to a, b, c tri-dummy nodes from big to small, ranking results is: a, c, b.

(3) according to described dependence matrix M, according to the node mapping method of setting, according to the order sequenced, the node mapping of dummy node to bottom physical node is carried out successively to the dummy node in described virtual network;

(4) node mapping complete after, the mapping between the physical pathway of the virtual link realizing virtual network according to the link maps method of setting layer physical network on earth.

The particular content of described step 1 comprises following operating procedure:

(11) all physical nodes of bottom physical network are numbered from 1, until numbering n, n are natural numbers, equal the physical node number of bottom physical network;

See the bottom physical network of shown in Fig. 3, Fig. 3, comprise 6 physical nodes altogether, number successively.Figure interior joint circle represents, numeral in circle is exactly the numbering of this physical node, the cpu resource ability of this physical node of numeral in the square frame on node side, the bandwidth resources ability of this link of numeral between node on link, the cpu resource ability of such as No. 1 physical node is 40 units, and between No. 1 physical node and No. 2 physical nodes, the bandwidth resources ability of physical link is 20 units.

(12) from the historical data set that the virtual network of bottom physical network accumulation successfully maps, each map record is taken out; To each map record, all construct the empty matrix P of the capable n row of n, time initial, each element value of this matrix P is 0 value; In this map record, if i-th bottom physical node was at least successfully mapped by a dummy node in this map record, then allow matrix P i-th row i-th arrange element a _iivalue is 1; In this map record, if a physical pathway between i-th bottom physical node and a jth bottom physical node was at least successfully mapped by the virtual link of in this map record, then allow matrix P i-th row jth row element a _ijwith the element a that jth row i-th arranges _jiall value is the inverse of this physical pathway jumping figure, and wherein i and j is the natural number being more than or equal to 1, being less than or equal to n, i and j must be unequal;

Such as, according to Article 1 map record, we obtain the matrix P corresponding with the physical network of bottom shown in Fig. 3 ¹as follows:

$P^1=\left[\begin{array}{cccccc}1& 0& 1/2& 0& 0& 1/2\\ 0& 0& 0& 0& 0& 0\\ 1/2& 0& 1& 0& 0& 1\\ 0& 1& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0\\ 1/2& 0& 1& 0& 0& 1\end{array}\right]$

(13) all matrix P constructed in step 12 are carried out matrix and be added summation, obtain a new capable n column matrix S of n;

Such as, we are total k bar map record always, and by step 12, we obtain altogether k matrix P ¹, P ²..., P ^k, then this k matrix addition summation is obtained matrix S as follows:

$S=\underset{i=1}{\overset{k}{\mathrm{\Σ}}}{P}^i$

(14) matrix S is normalized, obtains the dependence matrix M between bottom physical network nodes; The concrete mode of normalized is: for the element M of matrix M i-th row i-th row _iivalue is the average importance factors of this element representation bottom physical network i-th physical node; For the element M of matrix M i-th row jth row _ijvalue is the average degree of association factor between this element representation bottom physical network i-th physical node and a jth physical node; S in above-mentioned formula _iithe element of representing matrix S i-th row i-th row, S _ijthe element of representing matrix S i-th row jth row, i and j is the natural number being more than or equal to 1, being less than or equal to n, i and j must be unequal.

The above-mentioned mode be normalized matrix S can be merged into shown in following formula:

The particular content of described step 3 comprises following operating procedure:

(31) the current dummy node also not carrying out node mapping coming foremost is taken out;

(32) if this dummy node does not have father node, from described dependence matrix M, then find the current cpu resource that can meet this dummy node to require and the bottom physical node that on average importance factors is the highest, this dummy node is mapped on this physical node;

If this dummy node has e father node, then first find out e the bottom physical node having correspondence mappings relation with all father nodes of this dummy node; Then from described dependence matrix M, find the bottom physical node that a current cpu resource that can meet this dummy node requires, and require this physical node to divide to be clipped to that the connection product of the average degree of association factor of described e bottom physical node is maximum, so this dummy node is mapped on this physical node; E is a natural number being more than or equal to 1;

Such as: to a dummy node v, it has e father node, the sequence number of the corresponding bottom physical node mapped of this e father node is: the bottom physical node that then this dummy node maps is one and currently can meets its cpu resource requirement, and meet following formula physical node (its sequence number is t):

$t=\arg \underset{j}{\max }\underset{i=1}{\overset{e}{\mathrm{\Π}}}{M}_{{\mathrm{jv}}_p^i}$

In above formula represent the jth row the of dependence matrix M column element.

The father node of described dummy node refers to and to adjoin with this dummy node and to come the dummy node before this dummy node, and when this dummy node carries out node mapping, the father node of this dummy node completes node mapping;

See Fig. 2, the virtual network shown in corresponding diagram 2, dummy node a does not have father node, and the father node of dummy node c is a, and the father node of dummy node b is a and c.In the process of node mapping, dummy node a maps at first, is then dummy node c, is finally dummy node b.

(33) step 31 is got back to, until all dummy nodes complete mapping.

The link maps method of the setting described in described step 4 refers to k shortest path k-shortest path method.

Be the final result that the virtual network shown in Fig. 2 is mapped to the bottom physical network shown in Fig. 3 see Fig. 4, Fig. 4, specifically: dummy node a is mapped to No. 6 physical nodes, dummy node c is mapped to No. 3 physical nodes, and dummy node b is mapped to No. 1 physical node.

Inventor has carried out a large amount of emulation experiment to method proposed by the invention, and in emulation experiment, we use topology generator GT-ITM software building one to have 100 nodes and have the physical network on 500 limits.All be connected with the probability of 0.5 between every two nodes.In physical network, the cpu resource ability of node and the bandwidth of link all obey being uniformly distributed of [50,100].The arrival rate of each virtual network mapping request obeys the Poisson distribution arriving 5 virtual network mapping request with every 100 chronomeres.Each virtual network comprises [10,20] individual dummy node, the connection probability between every two dummy nodes is 0.5 equally.In each emulation experiment, we allow bottom physical network accept 2500 virtual network mapping request, and using the arithmetic mean of ten simulation results as last the simulation experiment result.The results show method of the present invention is effective, can improve the Mean mapping success rate of virtual network.

Claims (2)

1. realize a mapping method of virtual network for topology ambiguity based on historical data, it is characterized in that: described method comprises following operative step:

(1) according to the historical data set that the virtual network of bottom physical network accumulation successfully maps, calculate the dependence matrix M between bottom physical network nodes, specifically comprise following operating procedure: (11) all physical nodes to bottom physical network are numbered from 1, until numbering n, n is a natural number, equals the physical node number of bottom physical network; (12) from the historical data set that the virtual network of bottom physical network accumulation successfully maps, each map record is taken out; To each map record, all construct the empty matrix P of the capable n row of n, time initial, each element value of this matrix P is 0 value; In this map record, if i-th bottom physical node was at least successfully mapped by a dummy node in this map record, then allow matrix P i-th row i-th arrange element a _iivalue is 1; In this map record, if a physical pathway between i-th bottom physical node and a jth bottom physical node was at least successfully mapped by the virtual link of in this map record, then allow matrix P i-th row jth row element a _ijwith the element a that jth row i-th arranges _jiall value is the inverse of the jumping figure of this physical pathway, and wherein i and j is the natural number being more than or equal to 1, being less than or equal to n, i and j must be unequal; (13) all matrix P constructed in step 12 are carried out matrix and be added summation, obtain a new capable n column matrix S of n; (14) matrix S is normalized, obtains the dependence matrix M between bottom physical network nodes; The concrete mode of normalized is: for the element M of matrix M i-th row i-th row _iivalue is the average importance factors of this element representation bottom physical network i-th physical node; For matrix M i-th. the element M of row jth row _ijvalue is the average degree of association factor between this element representation bottom physical network i-th physical node and a jth physical node; S in above-mentioned formula _iithe element of representing matrix S i-th row i-th row, S _ijthe element of representing matrix S i-th row jth row, i and j is the natural number being more than or equal to 1, being less than or equal to n, i and j must be unequal;

(2) needs are carried out to the virtual network mapped, according to the demand size of dummy node in this virtual network to CPU cpu resource, from big to small all dummy nodes of this virtual network are sorted;

(3) according to described dependence matrix M, according to the node mapping method of setting, according to the order sequenced, the node mapping of dummy node to bottom physical node is carried out successively to the dummy node in described virtual network, specifically comprises following operating procedure: (31) take out the current dummy node also not carrying out node mapping coming foremost; (32) if this dummy node does not have father node, from described dependence matrix M, then find the current cpu resource that can meet this dummy node to require and the bottom physical node that on average importance factors is the highest, this dummy node is mapped on this physical node; If this dummy node has e father node, then first find out e the bottom physical node having correspondence mappings relation with all father nodes of this dummy node; Then from fast dependence matrix M, find the bottom physical node that a current cpu resource that can meet this dummy node requires, and require this physical node to divide to be clipped to that the connection product of the average degree of association factor of described e bottom physical node is maximum, so this dummy node is mapped on this physical node; E is a natural number being more than or equal to 1; The father node of described dummy node refers to and to adjoin with this dummy node and to come the dummy node before this dummy node, and when this dummy node carries out node mapping, the father node of this dummy node completes node mapping; (33) step 31 is got back to, until all dummy nodes complete mapping;

(4) node mapping complete after, the mapping between the physical pathway of the virtual link realizing virtual network according to the link maps method of setting layer physical network on earth.

2. a kind of mapping method of virtual network realizing topology ambiguity based on historical data according to claim 1, is characterized in that: the link maps method of the setting described in described step 4 refers to k shortest path k-shortest path method.