JP2015050594A - Arrangement device and method of virtual network to physical network - Google Patents

Abstract

PROBLEM TO BE SOLVED: To minimize the cost of a virtual link, in the arrangement of a virtual network having a Cycle structure.SOLUTION: A virtual node list is read, a virtual network is made to low order and high order induction subgraphs, and then an induction subgraph H consisting of a plurality of virtual nodes, and an induction subgraph K including the induction subgraph are arranged and calculated by an existing method. When determining a set Y giving the minimum cost, out of the arrangement to the physical node of induction subgraph where the induction subgraph H is removed from the induction subgraph K, a set Z of arrangement to physical node of the induction subgraph H replaces the induction subgraph K to induction subgraph H.

Description

  The present invention relates to an apparatus and method for placing a virtual network on a physical network, and in particular, when the performance conditions of the virtual network are given and there are a plurality of sets of physical resources that satisfy the requirements, the usage amount of the physical resource, Alternatively, the present invention relates to an apparatus and method for placing a virtual network on a physical network for selecting a set that minimizes the cost (cost) when using physical resources.

  Conventionally, as a method of generating a virtual network by selecting physical nodes and physical links that satisfy resource requirements, the method is a method of reducing the amount of calculation by dividing the virtual network into guided subgraphs and performing placement calculation. There is a simplified solution (see Non-Patent Document 1, for example).

  Below, the gradual reduction method is described.

  In the following, physical nodes and physical links have multiple types of resources, for example, physical nodes have processors, volatile memory, nonvolatile storage, peripheral devices, etc., and physical links have queues, signal frequencies, time slots, etc. possible. In addition, the types of resources of processors and the like can be subdivided by instruction sets and the like. A physical node or physical link that does not have the required type of resource is treated as if the resource amount of the resource is zero.

  Also, the amount of resources used by the virtual network (for example, the number of physical links) or the arrangement of the virtual network on the physical network is determined so as to reduce the necessary cost according to the amount (hereinafter, the objective function) To be unified by cost).

  In this solution, for example, a virtual network example a shown in FIG. 1 is given, and when the virtual network example a is arranged in the physical network example b shown in FIG. In calculating the set of, the virtual network example a is regarded as a tree structure with the virtual node v6 as a root (root node or root), and as shown in FIGS. The structure of the physical node arrangement that gives the minimum cost than the Tree structure, such as the structure, is decomposed into guided subgraphs that cost less, and the physical node placement that gives the minimum cost is calculated for each of the derived subgraphs. The physical node that gives the minimum cost of the Tree structure by repeating the integration of the physical node placement that gives the lowest cost calculated for each graph A set of physical node arrangements that gives a minimum cost is calculated with a smaller amount of calculation than calculating a set of node arrangements as they are. The meaning of each arrow in FIG. 2 means that for a certain virtual node (for example, virtual node v3 in FIG. 2A), the placement candidate physical node indicated by the tip of the arrow (for example, virtual node v3 in FIG. 2A) Is selected as the placement destination of the virtual node, the source of the arrow is the smallest candidate of the virtual node adjacent to the virtual node (for example, virtual node v2 in FIG. 2A) A set of physical node arrangements that provides a physical link with a cost (in the example, the minimum number of hops) is shown.

2A and 2B, a set of physical node arrangements that gives the minimum cost is calculated for each guidance subgraph, and a set of physical node arrangements that gives the minimum cost in v5 and v6 is further calculated. The root (v6) physical node that gives is calculated, and all sets of physical node arrangements of the virtual network a giving the minimum cost are thereby determined. The root physical node that gives the minimum cost is the path with the minimum cost. If | V _P | is the number of physical nodes and | E _L | is the number of virtual links, the computational amount of this solution is O (| V _P | ² | E _L |). However, this solution requires cost calculation between all physical nodes in advance, and the amount of calculation is O (| V _P | ³ ).

  In the following, the virtual network is exemplified by the virtual network example a in FIG. 1, and the physical network is exemplified by the physical network example b in FIG. 1, and the incremental reduction method for the virtual nodes v1, v2, v3, v4, and v5 in FIG. Will be described in detail.

  Note that the Tree structure is always the Tree structure at any point.

  Here, as a preliminary preparation, as shown in the table of costs between physical nodes in FIG. 1, a route that gives the minimum cost between all physical nodes is calculated in advance by the dijkstra method or the like. The solution consists of the following four procedures.

  (1) First, an arbitrary virtual node (for example, v6 in FIG. 1) is selected as a root from a virtual network (for example, the virtual network example a in FIG. 1).

  (2) In a derived subgraph composed of a set of virtual nodes for which the calculation of physical node allocation that gives the minimum cost has not been completed and a set of virtual nodes adjacent to those virtual nodes, only a virtual node of degree 1 and a virtual node of order 2 The physical node (and physical link) arrangement that gives the minimum cost is calculated for each of the induced subgraphs of the structure (Line structure) composed of the connections.

  (3) In a derived subgraph composed of a set of virtual nodes for which the calculation of physical node placement that gives the minimum cost is incomplete and a set of virtual nodes adjacent to those virtual nodes, a plurality of virtual nodes of degree 1 and virtuals of degree 3 or higher The physical node (and physical link) arrangement that gives the minimum cost is calculated for each of the derived subgraphs of the structure (Star structure) composed of one node connected.

  (4) The above steps (2) and (3) are repeated until the calculation of the physical node arrangement set that gives the minimum cost of all virtual nodes is completed.

  Supplementing the procedure (2), from the leaf-side virtual node to the root-side virtual node, for each virtual link arrangement, a set of physical node arrangements that gives the minimum cost is selected for each hop (equal cost ones are If there are multiple, select any one). The calculation result c ′ of the minimum cost of the previous hop is used as the calculation result of the minimum cost c of the next hop. For the cost, various coefficient values such as a weight value based on the number of hops and physical link capacity of the physical link and a weight value based on the processing speed of the physical node can be considered. In this specification, for the sake of clarity, the physical link Consider only the number of hops.

  Calculation example 1 of the line structure subgraph 1 and calculation example 2 of the line structure subgraph in FIG. 3 are combinations of physical node (and physical link) arrangements for the virtual link arrangement of v1 → v2 and the virtual link arrangement of v2 → v3. For example, when the physical node placement in the set of virtual nodes v1 and v2 is set to the physical nodes p6 and p4, respectively (in the calculation example 1 of the Line structure subgraph of FIG. 3). As the cost between physical nodes (physical links) in the hatch portion, 3 is input from the cost between physical nodes p6 and p4 in the table of costs between physical nodes in FIG. If the placement of the virtual node v2 is fixed to the physical node p2 and the one with the lowest cost is selected from the physical node placements of the virtual node v1, p6 is obtained. When the placement of the virtual node v2 is fixed to the physical node p3 and a plurality of candidate physical nodes have the lowest cost as in the case of selecting the lowest cost among the physical node placements of the virtual node v1 Is arbitrarily selected, for example, in this specification, the one with a small subscript of the physical node name is selected.

  Unless explicitly specified regarding the placement of the virtual network in the physical network, the virtual nodes of the previous hop and the next hop are assigned to different physical nodes. That is, in order to allocate a plurality of virtual nodes to physical nodes (for example, virtual machines) in the same physical enclosure (for example, virtual machine servers) for the algorithm for calculating the placement of the virtual network on the physical network and the mounting apparatus thereof. Unless the command is given, it is prohibited to assign a plurality of virtual nodes to physical nodes in the same physical enclosure, and the calculation result of the cost is, for example, that v2 and v3 in the calculation example 2 of the Line structure subgraph of FIG. NA (Not available) or the like is assigned as in the calculation result of the cost in the case of placement.

  The resource amount of the physical node and the physical link and the resource request amount of the virtual node and the virtual link are compared, and a group in which the resource request amount is larger than the resource amount is excluded from the allocation candidates.

  Supplementing step (3), select the set of physical node arrangements that gives the minimum cost for all virtual link arrangements of the virtual nodes on the root side that are directly connected to those of the order 3 or higher and the virtual nodes on the side of the order 1 (If there are multiple items of equal cost, select any one).

  In the calculation example of the Star structure subgraph of FIG. 3, when the physical node arrangement of the virtual node v5 is fixed to a certain physical node (for example, p5), the physical link arrangement of v3 → v5 and the virtual link arrangement of v4 → v5 FIG. 3 exemplifies the result of calculating the cost for a node (and physical link) arrangement set. For example, when a physical node arrangement for v3 is selected, a physical node arrangement that gives the minimum cost of v1 and v2 is shown in FIG. Since the calculation results 1 and 2 of the line structure induction subgraph of FIG. 1 are determined, these calculation results are used as c ′. For example, if v3 is arranged at p3 and v4 is arranged at p6, the cost becomes 9, respectively. For example, when v5 is arranged at p5, it is the minimum cost when v3 and v4 are arranged at p3 and p6. It is shown the door.

  When the physical node giving the minimum cost of the root virtual node (v6) is determined by repeating steps (2) and (3) for the virtual network example a in FIG. 1, the physical giving the minimum cost of the virtual network example a The node arrangement set can also be determined by following the calculation process.

There are two ways of arranging physical nodes of two adjacent virtual nodes: | V _P | × | V _P | = | V _P | ² and calculation for allocating virtual networks to physical networks in the gradual reduction method Is calculated by O (| V _P | ² | E _L |) because the arrangement of two adjacent virtual nodes (| V _P | ² ways) is repeated by the number of virtual links | E _L | In addition, the Dijkstra method requires the calculation amount O (| V _P | ³ ) to calculate the shortest path between all physical nodes and their distance and cost in advance. The quantity is O (| V _P | ² | E _L | + | V _P | ³ ).

Khanh-Toan Tran, N. Agoulmine, Y. Iraqi, "Cost-effective complex service mapping in cloud infrastructures", Network Operations and Management Symposium (NOMS), 2012.

  However, the conventional gradual reduction method can be applied to a calculation in which a virtual network having a topology that does not have a cycle structure such as a tree structure, a line structure, or a star structure is arranged in a physical network, but has a cycle structure. If the topology virtual network is applied to the calculation for allocating to the physical network, there is a problem that the virtual node to be calculated circulates and the calculation does not end, and therefore cannot be applied. Although it is possible to detect the circulation and finish the calculation, it is not guaranteed that the physical node arrangement with the minimum cost is obtained.

  FIG. 4 shows four examples of virtual networks having a Cycle structure. In general, a commercial network has a redundant configuration, and a Cycle structure as shown in FIG. 4 appears. Therefore, in an arrangement algorithm of a virtual network to a physical network, correspondence to the Cycle structure is an important issue.

  The present invention has been made in view of the above points, and provides an apparatus and method for arranging a virtual network on a physical network that can minimize the cost of a virtual link in the arrangement of a virtual network having a cycle structure on the physical network. The purpose is to provide.

According to one aspect, when performance conditions of a virtual network are given and there are a plurality of pairs of physical nodes and physical links that satisfy the requirements, a pair that minimizes the cost of the physical nodes and the physical links is selected. An arrangement device for a virtual network on a physical network,
A virtual node list storage means for storing a list of virtual nodes (virtual node list) of a virtual network having a Cycle structure;
A graph dividing means for reading out the virtual node list and dividing the virtual network into low-order and high-order guided subgraphs;
A reduction processing means for calculating the arrangement first from the low-order induction subgraph,
Have
The reduction processing means includes:
An induction subgraph H composed of a plurality of virtual nodes and an induction subgraph K including the induction subgraph H are arranged and calculated by an existing method, and a set Z of arrangements of the induction subgraph H to physical nodes When determining the set Y that gives the minimum cost among the arrangements of the induced subgraphs excluding the induced subgraph H from the induced subgraph K to the physical nodes, the induced subgraph K is replaced with the induced subgraph H. An apparatus for placing a virtual network on a physical network including means is provided.

According to one aspect, for a virtual network inductive subgraph having a cycle structure composed of virtual nodes having a minimum degree of 2, a physical node arrangement of one virtual node is selected and fixed, and By calculating a set of physical node arrangements that give a minimum cost so as to make one round toward one of the circulating routes, the maximum degree is 2 among the induced subgraphs of the Cycle structure that cannot be calculated by the conventional algorithm. It is possible to calculate the physical node arrangement that gives the minimum cost for the one consisting of the virtual nodes with the calculation amount O (| V _P | ³ | E _L |). In addition, for a virtual network derived subgraph having a cycle structure composed of virtual nodes having a minimum degree of D (3 or more), one virtual node having the lowest order is selected as a reduced virtual node, and the reduced virtual node is selected. A process of selecting one physical node arrangement group from the connection destination virtual nodes and selecting one physical node arrangement of the reduced virtual node that gives the minimum cost by the arrangement calculation from the physical node arrangement group. A physical node arrangement that gives the minimum cost is calculated O (| V _P | ^{D + 1} | E _L | by performing the physical node arrangement on the storage means as a D-dimensional array. ) Can be calculated. In addition, for a virtual network inductive subgraph that has a full mesh structure, a virtual node is selected, and it is regarded as a Star structure centered on the virtual node, and a set of physical node arrangements that gives the minimum cost. Is calculated for all virtual nodes, and the minimum cost of physical node placement in the Star structure centered on each virtual node is compared with each other so that the minimum cost is minimized. By selecting a set of physical node arrangements, the one that gives the minimum cost among the combinations of physical node arrangements in the guidance subgraph of the virtual network is calculated with the calculation amount O (| V _P | ² | E _L |) it can. A physical node that divides a virtual network into guided subgraphs composed of virtual nodes of low order, and sequentially calculates a set of physical node arrangements that give the minimum cost of the guided subgraph according to the minimum order, and gives the minimum cost By repeating the operation of removing the virtual node for which the calculation of the arrangement group has been completed from the virtual network, the calculation amount of the calculation of the physical node arrangement group that gives the minimum cost for the higher-order virtual node having a large calculation amount can be reduced. Can do.

It is an example of a virtual network. It is a conventional gradual reduction method. It is a calculation example of the induced subgraphs of the Line structure and Star structure by the conventional gradual reduction method. It is an example of a virtual network having a Cycle structure. It is a structural example of the virtual network arrangement | positioning apparatus in one embodiment of this invention. It is a flowchart of the whole operation | movement of the virtual network arrangement | positioning apparatus in one embodiment of this invention. It is a flowchart of the process outline | summary of the reduction process part of the order 1 induction | guidance | derivation subgraph in one embodiment of this invention. It is an example (the 1) of order 1 induction | guidance | derivation subgraph reduction in one embodiment of this invention. It is the example (the 2) of the reduction of the degree 1 induction | guidance | derivation subgraph in one embodiment of this invention. It is a flowchart of the Tree-root determination process of the simplification process of the degree 1 induction | guidance | derivation subgraph in one embodiment of this invention. It is a flowchart of the reduction process of a degree 2 induction | guidance | derivation subgraph in one embodiment of this invention. It is an example of the reduction of the degree 2 induction | subdivision subgraph (reduction of Cycle structure where one virtual node of order 3 or more exists or does not exist) in one embodiment of this invention. It is a detailed flowchart of step 160 of FIG. 11 in one embodiment of this invention. It is an example of the reduction of the degree 2 induction | guidance | derivation subgraph (reduction of Cycle structure in which two or more virtual nodes of degree 3 or more exist) in one embodiment of this invention. It is an example of the process of step 1604 of FIG. 13 in one embodiment of this invention. It is a detailed flowchart of step 166 of FIG. 11 in one embodiment of this invention. FIG. 7 is a detailed process of step 170 in FIG. 6 according to an embodiment of the present invention (a reduction process of a guidance subgraph of a non-full mesh structure having an order D of 3 or higher (high order)). FIG. 7 is a detailed process (full mesh reduction process) in step 180 of FIG. 6 in an embodiment of the present invention. FIG.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  The present invention selects and fixes one physical node arrangement of one virtual node of a guidance subgraph of the Cycle structure, and gives a minimum cost so as to make one turn toward one of the routes that circulate from the virtual node. By calculating the node arrangement set, it is possible to calculate the physical node arrangement that gives the minimum cost of the derivation subgraph of the Cycle structure that is not completed by the conventional algorithm.

  First, terms used below will be described.

  P2 (3): means that the value of the accumulated cost when passing through the physical node 2 is 3.

  Cost: The amount of resources used by the link or node, the cost required to use it, the number of hops (illustrated as the number of hops, but may be other than the number of hops), etc.

  -Resources: Resources required for computation and communication such as CPU, memory and bandwidth.

  Resource usage: Amount of resource usage.

  Guided subgraph: A graph in which a subset of a node (vertex) set of a graph is taken out, and all links (edges) belonging to the node set from which both end points are taken are link sets.

  Order: The number of links connected to the node.

  Cycle (Cycle structure): A graph having a structure that allows the user to reach the original node by following the link connected to the node without turning back.

  -Line (Line structure): A graph in which nodes and links are connected linearly. More precisely, a graph formed by concatenating two nodes of degree 1 and zero or more nodes of degree 2.

  Star (Star structure): A Star structure graph in which three or more nodes are directly connected to a certain node by a link. More precisely, a graph formed by concatenating three or more degree 1 nodes and one or more third order nodes.

Tree (Tree structure): A graph consisting of two or more nodes that do not have a Cycle. Tree structure includes Line structure and Star structure.
Leaf: A leaf (leaf node) in a tree structure graph.

  -Root: A root (root node) in a tree structure graph.

  Full mesh (full mesh structure or complete graph): A graph with edges between any two vertices of the graph.

  -Multiple sides: those sides where there are multiple sides between two vertices. In the present invention, the virtual links of multiple sides are handled as a single side by superimposing the costs of the respective virtual links.

  First, the outline of the present invention will be described.

  The present invention takes a higher-order virtual node (v1 in FIG. 12) from the Cycle structure as a reference point, and a route (for example, (v1, v2, v3, v1)) that circulates in an arbitrary direction from the reference point. Calculate the minimum cost. In calculating the minimum cost, the algorithm of Non-Patent Document 1 is applied to the direction of the route for each candidate physical node of the reference point. By this operation, a set (v1, v2, v3) = {(p1, p3, p2), (p6, p4, p3)} that gives the minimum cost when selecting each candidate physical node of the reference point is obtained. This is stored as the calculation result v ′ of v1. As shown in FIG. 12 to be described later, when another cycle structure (v1, v2, v5, v1) is connected to v1, the remaining Cycle connected to v1 using the calculation result v ′ of v1. The structure is sequentially calculated by the method of FIG. 12, and a set of physical node arrangements that gives the minimum cost can be calculated.

In the above processing, since the algorithm of Non-Patent Document 1 is repeated O (| V _P |) times, the amount of calculation by the processing of the present invention is O (| V _P | ³ | E _L |)).

  FIG. 5 shows an example of the configuration of a virtual network placement apparatus according to an embodiment of the present invention.

  The virtual network arrangement apparatus shown in FIG. 1 includes a virtual node list creation unit 10, a virtual node list storage unit 20, a virtual node order determination unit 30, an order 1 virtual node reduction processing unit 40, an order 2 virtual node reduction processing unit 50, and a high order. A virtual node reduction processing unit 60, a full mesh processing unit 70, and a memory 80 are included. The memory 80 includes a one-dimensional array storage area, a two-dimensional array storage area, a D-dimensional array storage area, and a full mesh physical node storage area.

  FIG. 6 is a flowchart of the overall operation of the virtual network placement apparatus according to the embodiment of the present invention.

  Before performing the following processing, create a graph of the virtual network. Specifically, some form of data [{end virtual node v1 of virtual link e1, end point virtual node v2 of virtual link e2},...] Representing a virtual link with two virtual nodes at both ends of the virtual link is stored. The data is stored in a means (for example, the virtual node list storage unit 20 or the memory 80). Further, the data of the induced subgraph and the data of the duplicate graph included in the steps of each processing unit are in the same list format as the above data string and are stored in the memory 80.

  Step 110) First, the virtual node list creation unit 10 generates a virtual node list based on the input virtual node information and stores it in the virtual node list storage unit 20. The virtual node list has, for example, a configuration of list format data [{virtual node name v1, virtual node degree 3, virtual node specification}, {virtual node name v2, virtual node order 1},...]. It should be noted that the variables (v, v ′, v ″, etc.) are generated by each simplified process by referring to each simplified processing unit 40, 50, 60.

  Step 120) The virtual node degree determination unit 30 reads the virtual node list in the virtual node list storage unit 20, and determines whether or not the minimum value of the virtual node order and the maximum value of the virtual node degree in the virtual node list are the same. To do. If they are the same, the process proceeds to step 180, and if they are different, the process proceeds to step 130.

  Step 130) Next, the virtual node degree determination unit 30 determines whether the minimum value of the virtual node degree in the virtual node list is “1”, “2”, or any other high order, and the order = If it is 1, the process proceeds to step 140. If the order = 2, the process proceeds to step 150. If the order = the high order, the process proceeds to step 160.

  Step 140) The degree-1 virtual node reduction processing unit 40 performs the processing shown in FIG.

  Step 150) The degree-2 virtual node reduction processing unit 50 performs the processing shown in FIG.

  Step 170) The high-order virtual node reduction processing unit 60 performs the processing shown in FIG.

  Step 180) The full mesh processing unit 70 performs the processing shown in FIG.

<Order 1 virtual node reduction processing unit 40>
The degree 1 virtual node reduction processing unit 40 reduces the induced subgraphs to be Tree, Line, and Star.

  FIG. 7 is an overall flowchart of the degree 1 induction subgraph reduction processing unit according to the embodiment of the present invention.

  Step 141) First, the degree-1 virtual node reduction processing unit 40 sets the root of the reduction subgraph to be reduced shown in FIG. 8 (hereinafter, Tree-root, T in the figure).

  1) Remove the virtual node of degree 1 and take out the part (cycle structure part of the virtual network) that becomes degree 2 or higher.

  2) If a derived subgraph of Tree, Line, or Star is connected to the extracted virtual node, that virtual node becomes Tree-root.

  Step 142) The degree 1 virtual node reduction processing unit 140 calculates a set of physical node arrangements that give the minimum cost in each of the induced subgraphs to be reduced by the technique of Non-Patent Document 1. Details will be described later.

  Step 143) The induced subgraph to be reduced is replaced by Tree-root.

  In the example of FIG. 9, a set of physical node arrangements that gives a minimum cost with v5 as Tree-root and Tree-root (v5) as a root is calculated. The minimum cost (value of v5 in FIG. 7) for the physical node that is a candidate for assignment to the tree-root is stored as a one-dimensional array (v5 ′).

  The process of step 141 will be described in detail.

  FIG. 10 is a flowchart of the tree-root determination process in the reduction of the degree 1 induction subgraph according to the embodiment of the present invention.

  Step 1411) The degree 1 virtual node reduction processing unit 140 duplicates the graph (duplicated graph) and stores it in some storage means (for example, the memory 80).

  Step 1412) The virtual node list in the virtual node list storage unit 20 is read and duplicated and stored in a memory (not shown).

  Step 1413) It is determined whether or not the minimum value of the degree of the virtual node in the replica of the virtual node list (replication list) is 0, 1, or 2 or more. If it is 1, the process proceeds to step 1414. If it is 0, the process proceeds to step 1418. If it is 2, the process proceeds to step 1419.

  Step 1414) When the minimum value of the degree of the virtual node in the replication list = 1, one virtual node of degree 1 is extracted from the replication list and set as v.

  Step 1415) Mark the virtual node adjacent to v in the replication list.

  Step 1416) Delete v from the duplicate graph.

  Step 1517) Delete v from the replication list and return to the processing of Step 1413.

  Step 1418) When the minimum value of the degree of the virtual node in the replication list = 1, the virtual node included in the replication list is determined to be Tree-root, and the process proceeds to FIG.

  Step 1419) If the minimum value of the degree of the virtual node in the replication list is ≧ 2, the virtual node remaining in the replication list is determined to be Tree-root, and the process proceeds to FIG.

  Next, the degree 2 virtual node reduction processing unit 50 and the degree 2 induction subgraph reduction process in step 150 of FIG. 6 will be described.

  FIG. 11 is a flowchart of the reduction process of the degree 2 induction subgraph according to the embodiment of the present invention.

  Step 151) The degree-2 virtual node reduction processing unit 50 reads the virtual node list from the virtual node list storage unit 20, duplicates it, and stores it in a memory (not shown).

  Step 152) One virtual node of degree 2 is arbitrarily selected from the replication list and set as v.

  Step 153) One of the two virtual nodes adjacent to v is v ′ and the other is v ″.

  Step 154) Mark v, v ', v "in the replication list.

  Step 155) For the order of v ′, it is determined whether the order of v ′ is 2 or the order of v ′ is 3 or more. When v ′ = 2, the process proceeds to step 156, and when v ′ ≧ 3, the process proceeds to step 161.

  Step 156) When v ′ = 2, the adjacent node of v ′ is newly set as v ′.

  Step 157) Mark v ′ in the duplicate list.

  Step 158) It is determined whether the orders of v ′ and v ″ are equal (v ′ = v ″). If v ′ = v ″, the process proceeds to step 159. If v ′ ≠ v ″, the process returns to step 155.

  Step 159) Let v ′ be u.

  Step 160) If there is one virtual node of degree 3 or higher, or reduction processing of the existing Cycle structure is performed, and the process proceeds to Step 167. Details will be described with reference to FIG.

  Step 161) In step 155, if the order of v ′ is 3 or more (v ′ ≧ 3), is the order of v ″ 2 (v ″ = 2) or 3 or more (v ″ ≧ 3)? If v ″ = 2, the process proceeds to step 162; otherwise, the process proceeds to step 165.

  Step 162) When v ″ = 2, the adjacent virtual node of v ″ is newly set as v ″.

  Step 163) Mark v ″ in the replication list.

  Step 164) It is determined whether the order of v ′ and the order of v ″ are equal (v ′ = v ″). If v ′ = v ″, the process proceeds to Step 159. If v ′ ≠ v ″, Return to step 161.

  Step 165) When the order of v ″ is 3 or more in Step 161, v ′ and v ″ are set to u1 and u2, respectively, and the process proceeds to Step 166.

  Step 166) Perform reduction processing of the Cycle structure in which there are two or more virtual nodes having an order of 3 or more, and proceed to Step 167.

  Step 167) The calculated virtual node of degree 2 is deleted from the list of virtual nodes on the memory.

  In performing the reduction of the order 2 part of the above step 160, the reduction of the cycle structure in which one virtual node of order 3 or higher exists or does not exist will be described with reference to FIG.

  1) If there is one virtual node of degree 3 or higher, or there is one virtual node of degree 3 or higher in a non-existing Cycle structure, that virtual node must be the starting virtual node and must not exist An arbitrary virtual node is set as a starting virtual node. In the example of FIG. 12, v1 is the starting virtual node.

  2) Select one physical node to be assigned to the origin virtual node. In the example of FIG. 12, p1 or p6 is selected.

  3) An arbitrary route is selected from two routes from the starting virtual node, and a set of physical node arrangements that gives the minimum cost is calculated by the algorithm of Non-Patent Document 1 for the peripheral circuit starting from the starting virtual node.

  4) Perform 2) and 3) above for all physical nodes that are candidates for assignment of the starting virtual node.

  5) The minimum cost for a physical node that is a candidate for assignment to the origin virtual node is stored as a one-dimensional array.

  FIG. 13 is a detailed flowchart of step 160 in FIG. 11 according to the embodiment of the present invention. In the simplification process of the degree 2 induction subgraph, there is one or more virtual nodes of degree 3 or higher. The reduction processing of Cycle structure is shown.

  Step 1601) The degree-2 virtual node reduction processing unit 50 sets H as a guidance subgraph composed of virtual nodes marked in the replication list.

  Step 1602) From the physical nodes that are the placement candidates for u in Step 159, select one that has not been subjected to placement calculation.

  Step 1603) An arbitrary route is selected from the two routes from u.

  Step 1604) For the peripheral circuit of the induction subgraph H starting from u, a set of physical node arrangements that gives the minimum cost is calculated by the algorithm of Non-Patent Document 1.

  Step 1605) It is confirmed whether or not all the physical nodes that are the placement candidates for u have been calculated. If the calculation has not been completed, the process returns to Step 1602, and if it has been calculated, the process proceeds to Step 1606.

  Step 1606) The minimum cost for the set of physical node placements on u is stored in a memory (not shown) as a two-dimensional array.

  FIG. 14 shows an example of the reduction of the Cycle structure in which there are two or more virtual nodes of degree 3 or more in the reduction process of the degree 2 induction subgraph. In the figure, “*” indicates that the cost has not been determined.

  1) In a Cycle structure in which there are two or more virtual nodes of degree 3 or more, two virtual nodes of degree 3 or more that are the two end points of the line of the virtual node of degree 2 are the starting virtual node 1 and the starting virtual node 2, respectively. To do.

  2) Perform for all physical nodes that are candidates for placement in the origin virtual node 1.

  3) Perform 2) for all physical nodes that are candidates for placement in the origin virtual node 1.

  4) The minimum cost for the set of physical nodes placed at the origin virtual node 1 and the origin virtual node 2 is stored in the two-dimensional array storage area of the memory 80 as a two-dimensional array.

  Next, in the simplification of the Cycle structure in which there are two or more virtual nodes of degree 3 or more in the simplification of the degree 2 induction subgraph of step 1604 in FIG. 13, the two-dimensional array stored in the two-dimensional array region is reduced. As shown in FIG. 15, the used layout calculation uses 2 stored in the two-dimensional array storage unit 55 instead of the cost when the physical nodes are directly connected when calculating the cost for the physical node layout set. Use a dimensional array.

  FIG. 16 is a detailed flowchart of step 166 of FIG. 11 according to the embodiment of the present invention. In the reduction process of degree 2 induction subgraph, reduction of the Cycle structure in which two or more virtual nodes of degree 3 or more exist is performed. Is.

  Step 1661) Let H be a guidance subgraph consisting of virtual nodes marked in the replication list.

  Step 1662) From the physical nodes that are the placement candidates for u1, one that has not been subjected to placement calculation is selected.

  Step 1663) For the route from u1 to u2 in the guidance subgraph H, a set of physical node arrangements that gives the minimum cost is calculated by the algorithm of Non-Patent Document 1.

  Step 1664) The minimum cost for the set of physical nodes that are candidates for placement in u1 is stored in the two-dimensional array storage area of the memory 80 as a two-dimensional array.

  Next, in step 120 of FIG. 6, the minimum value of the virtual node order in the virtual node list is different from the maximum value of the virtual node order and the minimum value of the virtual node order in the virtual node list is = 3. The process of step 170 executed by the point-in-time virtual node reduction processing unit 60 will be described.

  FIG. 17 shows detailed processing (reduction processing of a guidance subgraph of a non-full mesh structure having an order D of 3 or higher (high order)) in step 170 of FIG. 6 in one embodiment of the present invention.

  In step 120, when the minimum value of the virtual node order in the virtual node list is equal to the maximum value of the virtual node order and the degree (D) is 3 or more, the reduction process of the induced subgraph is performed as follows.

  1) One virtual node of the lowest order D (for example, v1 in the example of FIG. 17) is selected as a reduced virtual node.

  2) Select one physical node layout set from the virtual nodes (v2, v3, v4) to which the reduced virtual node is connected, and then select one reduced virtual node physical node that gives the minimum cost.

  3) 2) is performed for all sets of physical node arrangements, and the cost for each set is stored in the D-dimensional array storage area of the memory 80. In the example of FIG. 17, when the physical node arrangement of v2, v3, and v4 is determined, the physical node arrangement of v1 that gives the minimum cost is determined, and the cost at that time is set as a three-dimensional array, for example, C [v2 = p1; v3 = p2; v4 = p3] = 3 is stored in the D time point array storage unit 65.

  Next, the full mesh process executed when the minimum value of the virtual order node in the virtual node list is equal to the maximum value of the virtual node order in step 120 of FIG. 6 will be described.

  FIG. 18 is a detailed process of step 180 of FIG. 6 according to the embodiment of the present invention, and shows an example of a reduction process (full mesh process) of a guidance subgraph having a full mesh structure of degree D of 3 or more.

  This case is a reduction method in the case where the virtual network has a falling mesh structure, or in the case where a guidance subgraph that appears by repeatedly reducing the virtual network has a full mesh structure.

  The full mesh processing unit 70 selects one virtual node of the virtual network to be placed, regards it as a Star structure centered on the virtual node, and applies the placement algorithm of Non-Patent Document 1 to give the minimum cost A set of placement is calculated. Similarly, for a Star structure centered on all other virtual nodes, the placement algorithm of Non-Patent Document 1 is applied to calculate a set of physical node placements that gives the minimum cost, and Stars centered on each virtual node The minimum costs given by the set of physical node arrangements that give the minimum cost to the structure are respectively compared, and the smallest set among the minimum costs becomes the set of physical node arrangements that gives the minimum cost of the virtual network.

In addition, the minimum cost of the full mesh is guaranteed without calculating the cost of the spanning portion (dotted line portion in FIG. 18) when the virtual network having the full mesh structure is regarded as the Star structure by repeating the simplification in this way. Can do. The amount of calculation in this case is O (| V _P | ² | E _L |). In other words, assuming that the number of virtual nodes is | V _L |, the computation amount of each Star structure is O ((| V _L | −1) | V _P | ² ), and it is repeated | V _L |
O (| V _L | (| V _L | −1) | V _P | ² ) = O (2 ((| V _L | (| V _L | −1)) / 2) | V _P | ² ) = O (2 | E _L || V _P | ² ) = O (| E _L || V _P | ² )
It becomes. However, it was used that | V _L | (| V _L | −1)) / 2 = | E _L | in the full mesh.

  In FIG. 18, when obtaining a set of physical node arrangements that gives the minimum cost of the full mesh structure, the following processing is performed. In the following, the virtual network shown in FIG. 18A and the physical node shown in FIG. 18B are used.

  1) Considering the Star structure centered on each virtual node in FIG. 18A, the placement algorithm of Non-Patent Document 1 is applied to calculate a physical node placement set that gives the minimum cost.

  2) A set of physical node arrangements that gives the smallest among the minimum costs calculated in each Star structure of FIG. 18B (in the example of FIG. 18B, the physical node whose cost is 12 in the Star structure 3) The arrangement set) is a physical node arrangement set that gives the minimum cost of the full mesh structure, and is stored in the full mesh storage area of the memory 80 as a calculation result.

  Note that the operations of the components of the virtual network placement apparatus shown in FIG. 5 can be constructed as a program, installed in a computer used as the virtual network placement apparatus, executed, or distributed via a network. .

  The present invention is not limited to the above-described embodiments, and various modifications and applications are possible within the scope of the claims.

DESCRIPTION OF SYMBOLS 10 Virtual node list preparation part 20 Virtual node list memory | storage part 30 Virtual node degree determination part 40 Order 1 virtual node reduction process part 50 Order 2 virtual node reduction process part 60 High order virtual node reduction process part 70 Full mesh process part 80 Memory

The present invention relates to a deployment device and method for virtual networks physical network, in particular, given the performance conditions of the virtual network, if the set of physical resources to meet the performance requirements are more present, the amount of physical resource The present invention also relates to an apparatus and a method for placing a virtual network on a physical network for selecting a set that minimizes the cost (cost) of using physical resources.

According to one aspect, given performance requirements of the virtual network, if the set of physical nodes and physical links satisfying the performance requirements there are a plurality, selecting a set of the physical node and the cost of the physical link is a minimum An arrangement device for a virtual network to a physical network,
A virtual node list storage means for storing a list of virtual nodes (virtual node list) of a virtual network having a Cycle structure;
A graph dividing means for reading out the virtual node list and dividing the virtual network into low-order and high-order guided subgraphs;
A reduction processing means for calculating the arrangement first from the low-order induction subgraph,
Have
The reduction processing means includes:
An induction subgraph H composed of a plurality of virtual nodes and an induction subgraph K including the induction subgraph H are arranged and calculated by an existing method, and a set Z of arrangements of the induction subgraph H to physical nodes When determining the set Y that gives the minimum cost among the arrangements of the induced subgraphs excluding the induced subgraph H from the induced subgraph K to the physical nodes, the induced subgraph K is replaced with the induced subgraph H. An apparatus for placing a virtual network on a physical network including means is provided.

Claims (8)

Given the performance conditions of a virtual network and when there are multiple sets of physical node arrangements and physical link arrangements that meet the requirements, select the set that minimizes the cost of at least one of the physical nodes or physical links An arrangement device for a virtual network to a physical network,
A virtual node list storage means for storing a list of virtual nodes (virtual node list) of a virtual network having a Cycle structure;
A graph dividing unit that reads the virtual node list and divides the virtual network into guided subgraphs according to a minimum degree;
A simplification processing means for calculating the arrangement first from a low-order (small) guidance subgraph,
Have
The reduction processing means includes:
An induction subgraph H composed of a plurality of virtual nodes and an induction subgraph K including the induction subgraph H are arranged and calculated by an existing method, and a set Z of arrangements of the induction subgraph H to physical nodes When determining the set Y that gives the minimum cost among the arrangements of the induced subgraphs excluding the induced subgraph H from the induced subgraph K to the physical nodes, the induced subgraph K is replaced with the induced subgraph H. An apparatus for placing a virtual network on a physical network, characterized by comprising: The reduction processing means includes:
When the minimum value of the order of the virtual node in the virtual node list is different from the maximum value of the order of the virtual node, and the minimum value of the order of the virtual node list is 1, a guided subgraph having a Tree structure, a node 1st reduction processing to perform reduction processing of induction subgraphs having a Star structure in which three or more nodes are directly connected to one node by links. Means,
When the minimum value of the degree of the virtual node in the virtual node list is different from the maximum value of the order of the virtual node and the minimum value of the virtual node order in the virtual node list is 2, a derived subgraph having degree 3 or more A second reduction processing means for performing a reduction process of a Cycle structure in which one or more of them exist and a reduction process of a Cycle structure in which two or more virtual nodes of degree 3 or more exist;
When the minimum value of the virtual node order in the virtual node list is different from the maximum value of the virtual node order, and the minimum value of the virtual node order in the virtual node list is 3, a high-order guided subgraph A third reduction processing means for performing the reduction processing of
Fourth reduction processing means for performing a fourth reduction process for reducing a full-mesh structure derived subgraph when the minimum value of the virtual node degree in the virtual node list is equal to the maximum value of the virtual node order; ,
The apparatus for arranging a virtual network on a physical network according to claim 1. The first reduction processing means includes:
Repeat the operation of removing all virtual nodes of order 1 and their connected virtual links, and the remaining virtual nodes of order 2 and above, and the tree is connected before performing the operation of removing all virtual nodes of order 1 and their connected virtual links A tree having a virtual node as a root is selected, a minimum cost of a path to all the candidate physical nodes of the root is obtained by the placement calculation, and the physical node having the minimum cost is stored as a one-dimensional array The apparatus for placing a virtual network on a physical network according to claim 2, further comprising means for storing in the physical network. The second reduction processing means includes:
In a Cycle structure where there is one or more virtual nodes of order 3 or higher,
If there is one virtual node of degree 3 or higher, that virtual node is set as the starting virtual node. If there is no virtual node, any virtual node is set as the starting virtual node, and one physical node assigned to the starting virtual node is selected. Means,
A process of selecting an arbitrary path from the two paths from the starting virtual node and obtaining a set of physical nodes having a minimum cost by the placement calculation for a peripheral circuit starting from the starting virtual node. Means for repeating for a physical node that is a candidate for allocation of a starting virtual node;
3. The apparatus for arranging a virtual network in a physical network according to claim 2, further comprising means for storing in a storage means a minimum cost for a physical node that is an allocation candidate of the starting virtual node as a one-dimensional array. The second reduction processing means includes:
In the Cycle structure where there are two or more virtual nodes of degree 3 or more,
Means for setting two virtual nodes of degree 3 or more that are both ends of the line of the virtual node of degree 2 as the starting virtual node A and the starting virtual node B, respectively;
One physical node that is a placement candidate for the origin virtual node A is selected, and a process for calculating the minimum cost by the placement calculation for a route toward the origin virtual node B is determined as a placement candidate for the origin virtual node A. Means for all the physical nodes
3. The arrangement device for a virtual network in a physical network according to claim 2, further comprising means for storing in a storage means a minimum cost for a set of physical node arrangements on the origin virtual node A and the origin virtual node B as a two-dimensional array. The third reduction means includes:
One virtual node of the lowest order is selected as the reduced virtual node, one set of physical node arrangement is selected from the connection destination virtual nodes of the reduced virtual node, and the minimum cost is calculated from the set of physical node arrangement by the arrangement calculation. 3. The processing for selecting one physical node arrangement of the reduced virtual nodes to be given is performed for all sets of physical node arrangements, and the cost for each group is stored in a storage means as a D-dimensional array. A device that places a virtual network on a physical network. The fourth reduction means includes:
Select one virtual node, regard it as a Star structure centered on the virtual node, perform a process of calculating a set of physical nodes that give the minimum cost by the placement calculation, and reduce the minimum cost. 3. The apparatus for placing a virtual network on a physical network according to claim 2, further comprising means for storing | V _L | as a one-dimensional array in the storage means. When a performance condition of a virtual network is given and there are a plurality of sets of physical nodes that satisfy the requirement, a method of arranging a virtual network on a physical network that selects a set that minimizes the cost of the physical node,
A virtual node list storage means for storing a list of virtual nodes (virtual node list) of a virtual network having a Cycle structure;
In an apparatus having a graph dividing means and a reduction processing means,
The graph dividing means reads the virtual node list and divides the virtual network into low-order and high-order guided subgraphs; and
The reduction processing unit calculates and calculates a guidance subgraph H including a plurality of virtual nodes and a guidance subgraph K including the guidance subgraph H by an existing method, and places the guidance subgraph H on physical nodes. When determining the set Y that gives the minimum cost among the arrangements of the induced subgraphs obtained by subtracting the induced subgraph H from the induced subgraph K to the physical nodes of the induced subgraph K, the set Z A reduction processing step to replace with the guidance subgraph H;
A method for arranging a virtual network on a physical network, characterized by:

Images