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

Abstract

A virtual network mapping method based on historical data to realize topology awareness, comprising the following steps: (1) calculating the dependency matrix M between the underlying physical network nodes according to the historical data collection of the successful mapping of the virtual network accumulated by the underlying physical network; (2) sort all the virtual nodes in the virtual network from large to small according to the demand of the virtual nodes in the virtual network for CPU resources of the central processing unit; (3) according to the dependency matrix M, proceed from the virtual nodes to the bottom layer in sequence Node mapping of physical nodes; (4) After the node mapping is completed, the mapping between the virtual link of the virtual network and the physical path of the underlying physical network is realized according to the set link mapping method. The method of the invention realizes the scientific evaluation of the resource capability of the underlying physical network, realizes the optimal selection of node mapping by sensing the topology structure of the virtual network, and improves the long-term average success rate of virtual network mapping.

Description

A topology-aware virtual network mapping method based on historical data

technical field

The invention relates to a method for realizing virtual network mapping, which belongs to the technical field of computer networks, in particular to the technical field of network virtualization.

Background technique

Network virtualization refers to the logical division of a common physical network infrastructure into multiple isolated virtual networks with different network topologies. A virtual network generally includes multiple virtual nodes and multiple virtual links. Each virtual node and each virtual link has different resource requirements, such as the resource requirements of the virtual node for the central processing unit CPU, and the virtual link for the physical link. road bandwidth requirements. By renting the infrastructure slice of the underlying physical network, the service provider SP makes full use of the access control rights provided by the underlying physical network infrastructure, and can quickly deploy a customized network protocol without investing in related physical network hardware. Or the architecture is a virtual network, which provides diversified services to end users.

In the process of mapping the virtual network to the underlying physical network, since the resource requirements of nodes and links need to be satisfied at the same time, the mapping problem of network virtualization is an NP-hard problem. At present, related solutions are generally designed based on heuristic methods, but the current heuristic virtual network mapping scheme has the following problems: (1) The current resource scoring standard is to multiply the CPU capability value of the physical node by the adjacent chain of the node However, this resource scoring standard is not accurate, leading to the possibility that the scheme selects physical nodes with strong CPU capabilities but weak links for mapping, so that the virtual network mapping fails in the link mapping stage; (2 ) always uses the greedy algorithm to select the physical node with the highest score for mapping, ignoring the location of the already mapped virtual node, that is, not considering the topology of the virtual network. Therefore, in the process of virtual network mapping, how to better evaluate the underlying physical network resource capabilities, and how to optimize the next step of node mapping based on the mapped virtual nodes and their topological structures are the current computer network engineering It is a technical problem that urgently needs to be solved.

Contents of the invention

In view of this, the object of the present invention is to invent a method for realizing virtual network mapping, which can utilize a large number of historical data sets successfully mapped by the virtual network accumulated in the underlying network to realize a scientific evaluation of the resource capabilities of the underlying physical network, and can combine The virtual nodes and their topological structures that have been mapped realize the optimal selection of the next node mapping.

In order to achieve the above object, the present invention proposes a virtual network mapping method based on historical data to realize topology awareness, and the method includes the following steps:

(1) Calculate the dependency matrix M between the nodes of the underlying physical network according to the historical data set successfully mapped by the virtual network accumulated in the underlying physical network, which specifically includes the following steps: (11) For all physical nodes of the underlying physical network Carry out numbering until numbering n, n is a natural number, equal to the number of physical nodes of the underlying physical network; (12) from the historical data collection of the virtual network successfully mapped in the underlying physical network, take out each mapping record; for each mapping Records, construct an empty matrix P with n rows and n columns, each element value of the matrix P is 0 at the beginning; in the mapping record, if the i-th underlying physical node is at least one virtual node in the mapping record After successful mapping, let the element a ii of the i-th row and _i -column of the matrix P take the value of 1; in this mapping record, if a physical path between the i-th underlying physical node and the j-th underlying physical node At least one virtual link in the mapping record has been successfully mapped, then let the element a _ij in the i-th row and j-column of the matrix P and the element a _ji in the j-th row and i-column both take the value of the physical path The reciprocal of the number of jumps, wherein i and j are all natural numbers greater than or equal to 1 and less than or equal to n, i and j must be unequal; (13) adding and summing all matrices P constructed in step 12 to obtain a A new n-row n-column matrix S; (14) normalize the matrix S to obtain the dependency matrix M between the underlying physical network nodes; the specific way of normalization is: for the i-th row of the matrix M The value of element M _ii in column i is This element represents the average importance factor of the i-th physical node of the underlying physical network; for the element M ij in the i-th row and j-th column of the matrix M, _the value is This element represents the average correlation factor between the i-th physical node and the j-th physical node of the underlying physical network; in the above formula, S _ii represents the element in the i-th row and the i-th column of the matrix S, and S _ij represents the i-th row of the matrix S The elements in the jth column, i and j are both natural numbers greater than or equal to 1 and less than or equal to n, and i and j must not be equal;

(2) For a virtual network that needs to be mapped, sort all the virtual nodes of the virtual network from large to small according to the demand size of the virtual nodes in the virtual network for CPU resources;

(3) According to the dependency matrix M, according to the set node mapping method, the virtual nodes in the virtual network are mapped from the virtual nodes to the underlying physical nodes according to the arranged order, specifically including The following steps are as follows: (31) take out the virtual node that is currently at the forefront and has not yet been mapped to the node; (32) if the virtual node does not have a parent node, then find out from the dependency matrix M that currently satisfies the The CPU resource requirement of the virtual node and the underlying physical node with the highest average importance factor map the virtual node to the physical node; if the virtual node has e parent nodes, first find out the The e underlying physical nodes corresponding to the mapping relationship; then find an underlying physical node that can currently meet the CPU resource requirements of the virtual node from the dependency matrix M, and require the physical node to go to the e underlying physical nodes respectively The joint product of the average association degree factor of the physical node is the largest, so the virtual node is mapped to the physical node; e is a natural number greater than or equal to 1; the parent node of the virtual node refers to adjacent to the virtual node and sorted Arrange the virtual node before this virtual node, when this virtual node performs node mapping, the parent node of this virtual node has completed node mapping; (33) go back to step 31, until all virtual nodes complete mapping;

(4) After the node mapping is completed, the mapping between the virtual link of the virtual network and the physical path of the underlying physical network is realized according to the set link mapping method.

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

The beneficial effect of the present invention is that: the virtual network mapping method of the present invention realizes the scientific evaluation of the underlying physical network resource capabilities by using the historical data successfully mapped by the virtual network, and can perceive the topology structure of the virtual network to realize the optimal selection of node mapping ; The virtual network mapping method of the present invention effectively improves the long-term average success rate of virtual network mapping, and brings more long-term average benefits to the underlying physical network infrastructure provider.

Description of drawings

Figure 1 is a flow chart of the method of the present invention.

Figure 2 is a schematic diagram of a virtual network.

Figure 3 is a schematic diagram of an underlying physical network.

FIG. 4 is a schematic diagram of mapping the virtual network shown in FIG. 2 to the underlying physical network shown in FIG. 3 .

Detailed ways

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings.

Referring to Fig. 1, a kind of virtual network mapping method that realizes topology awareness based on historical data of the present invention is introduced, and the method includes the following steps:

(1) Calculate the dependency matrix M between the underlying physical network nodes according to the historical data set successfully mapped by the virtual network accumulated in the underlying physical network;

(2) For a virtual network that needs to be mapped, sort all the virtual nodes of the virtual network from large to small according to the demand size of the virtual nodes in the virtual network for CPU resources;

See Figure 2. The virtual network shown in Figure 2 includes three virtual nodes, namely a, b, and c. The numbers in the boxes next to the nodes indicate the CPU resource requirements of the virtual nodes, and the virtual links between nodes The number on the road indicates the bandwidth resource requirement of the virtual link. For example, the CPU resource requirement of virtual node a is 10 units, and the bandwidth resource requirement of the virtual link between virtual node a and virtual node b is 8 units. According to the demand of the virtual nodes for CPU resources, the three virtual nodes a, b, and c are sorted from large to small, and the sorting results are: a, c, b.

(3) according to the dependency matrix M, according to the set node mapping method, the virtual nodes in the virtual network are mapped from the virtual nodes to the underlying physical nodes in sequence according to the arranged order;

(4) After the node mapping is completed, the mapping between the virtual link of the virtual network and the physical path of the underlying physical network is realized according to the set link mapping method.

The specific content of said step 1 includes the following steps:

(11) all physical nodes of the underlying physical network are numbered from 1 to numbering n, where n is a natural number equal to the number of physical nodes of the underlying physical network;

Referring to FIG. 3 , an underlying physical network shown in FIG. 3 includes a total of 6 physical nodes, which are numbered sequentially. The nodes in the figure are represented by circles, the numbers in the circles are the numbers of the physical nodes, the numbers in the boxes next to the nodes indicate the CPU resource capabilities of the physical nodes, and the numbers on the links between nodes indicate the bandwidth of the links Resource capability, for example, the CPU resource capability of physical node 1 is 40 units, and the bandwidth resource capability of the physical link between physical node 1 and physical node 2 is 20 units.

(12) Take out each mapping record from the historical data set of successful mapping of the virtual network accumulated in the underlying physical network; for each mapping record, construct an empty matrix P with n rows and n columns, and initially each of the matrix P element value is 0; in the mapping record, if the i-th underlying physical node has been successfully mapped by at least one virtual node in the mapping record, let the element a ii in the i-th row and _i -column of the matrix P take the value is 1; in this mapping record, if a physical path between the i-th underlying physical node and the j-th underlying physical node is successfully mapped by at least one virtual link in the mapping record, then let the matrix P The element a _ij in the jth column of the i row and the element a _ji in the ith column of the jth row both take the value of the reciprocal of the hop count of the physical path, where i and j are natural numbers greater than or equal to 1 and less than or equal to n, i and j must not be equal;

For example, according to the first mapping record, we get the matrix ^P1 corresponding to the underlying physical network shown in Figure 3 as follows:

${\mathrm{<span\; class="notranslate">\; P</span>}}^{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>\left[\begin{array}{cccccc}\mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; 2</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; 2</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; 2</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; 2</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]$

(13) carry out matrix addition summation to all matrix P constructed in step 12, obtain a new n row n column matrix S;

For example, we have a total of k mapping records, through step 12, we get a total of k matrices P ¹ , P ² , ..., P ^k , then add and sum these k matrices to obtain a matrix S as follows:

$\mathrm{<span\; class="notranslate">\; S</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; P</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}$

(14) Perform normalization processing on the matrix S to obtain the dependency matrix M between the underlying physical network nodes; the specific method of normalization processing is: for the element M _ii of the i-th row and i-column of the matrix M, the value is This element represents the average importance factor of the i-th physical node of the underlying physical network; for the element M ij in the i-th row and j-th column of the matrix M, _the value is This element represents the average correlation factor between the i-th physical node and the j-th physical node of the underlying physical network; in the above formula, S _ii represents the element in the i-th row and the i-th column of the matrix S, and S _ij represents the i-th row of the matrix S For the elements in the jth column, both i and j are natural numbers greater than or equal to 1 and less than or equal to n, and i and j must not be equal.

The above methods of normalizing the matrix S can be combined into the following formula:

The specific content of said step 3 includes the following steps:

(31) Take out the virtual node that is currently at the forefront and has not yet been mapped to the node;

(32) If the virtual node does not have a parent node, find the underlying physical node that can currently meet the CPU resource requirements of the virtual node and have the highest average importance factor from the dependency matrix M, and map the virtual node to the on the physical node;

If the virtual node has e parent nodes, first find e underlying physical nodes that have a corresponding mapping relationship with all parent nodes of the virtual node; then find a current virtual node that can satisfy the The underlying physical node required by the CPU resources, and the joint product of the average correlation factor of the physical node to the e underlying physical nodes is required to be the largest, so the virtual node is mapped to the physical node; e is a value greater than a natural number equal to 1;

For example: for a virtual node v, it has e parent nodes, and the serial number of the underlying physical node mapped to the e parent nodes is: Then the underlying physical node mapped by the virtual node is a physical node (its sequence number is t) that can currently meet its CPU resource requirements and satisfy the following formula:

$\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; arg</span>}\underset{\mathrm{<span\; class="notranslate">\; j</span>}}{\mathrm{<span\; class="notranslate">\; max</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; e</span>}}{\mathrm{<span\; class="notranslate">\; \&Pi;</span>}}}{\mathrm{<span\; class="notranslate">\; m</span>}}_{{\mathrm{<span\; class="notranslate">\; jv</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}}$

In the above formula Indicates that the jth row of the dependency matrix M column elements.

The parent node of the virtual node refers to a virtual node adjacent to the virtual node and ranked before the virtual node. When the virtual node performs node mapping, the parent node of the virtual node has completed node mapping;

Referring to FIG. 2 , corresponding to the virtual network shown in FIG. 2 , virtual node a has no parent node, virtual node c has parent node a, and virtual node b has parent nodes a and c. In the process of node mapping, virtual node a is mapped first, then virtual node c, and finally virtual node b.

(33) Go back to step 31 until all virtual nodes complete the mapping.

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

Referring to Figure 4, Figure 4 is the final result of mapping the virtual network shown in Figure 2 to the underlying physical network shown in Figure 3, specifically: virtual node a is mapped to physical node No. 6, and virtual node c is mapped to physical node No. 3 , virtual node b is mapped to physical node 1.

The inventor has conducted a lot of simulation experiments on the method proposed by the present invention. In the simulation experiments, we used the topology generator GT-ITM software to construct a physical network with 100 nodes and 500 edges. Every two nodes are connected with a probability of 0.5. In the physical network, the CPU resource capability of the nodes and the bandwidth of the link are subject to the uniform distribution of [50, 100]. The arrival rate of each virtual network mapping request obeys the Poisson distribution of 5 virtual network mapping requests arriving every 100 time units. Each virtual network contains [10, 20] virtual nodes, and the connection probability between every two virtual nodes is also 0.5. In each simulation experiment, we let the underlying physical network accept 2500 virtual network mapping requests, and take the arithmetic mean of the ten simulation results as the final simulation experiment result. Experimental results prove that the method of the present invention is effective and can improve the average mapping success rate of the 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.