JP2012050025A - Route control device, route control method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide technology enabling scale out of a tunnel termination device without interrupting communication.SOLUTION: A center side tunnel termination system 10a comprises: a first processing unit 11 transmitting/receiving a packet including a call control signal for establishing a tunnel by using IPsec to/from a terminal 20A through a third processing unit 13, and generating tunnel termination information for terminating the tunnel; a second processing unit 12 performing termination processing of the packet received from the terminal 20A through the third processing unit 13, based on the tunnel termination information received from the first processing unit 11; the third processing unit 13 transferring the packet including the call control signal to the first processing unit 11 based on header information of the packet received from the terminal 20A, and transferring the packet passing the tunnel to the second processing unit 12, by Open Flow control.

Description

  The present invention relates to a packet routing technique in a communication network.

  The communication network includes nodes and links, and a terminal such as a PC (Personal Computer) is connected to a node such as a router arranged in the network. Each node has a relay function for accumulating packets generated in the node and packets transferred from other nodes and transferring them to nodes connected by a link. Note that “occurred” means that a packet is received from a terminal connected to the node. A packet generated in an arbitrary node has a destination node, and is transmitted to the destination node through a link. When a node has a plurality of adjacent nodes (links), the node determines to which adjacent node the packet is transmitted. This is called path control.

  In the Internet, normally, a node existing in the middle of a packet transmission path is called a relay node, and the relay node performs path control using hop-by-hop routing that determines a path to the next node. Specifically, each relay node has a route table in which the destination node that is the destination of the packet and the route to the destination node are written, and should be routed next based on the destination node of the received packet Determine an adjacent node. Each node exchanges information regarding the network status and updates the routing table as necessary. As a method for exchanging information regarding the network status, OSPF (Open Shortest Path First) in which each node exchanges node connection information and node connection information is generally used. For example, the connection information of each node is a network topology and a link cost. The link cost is a cost assigned to the link. Generally, the smaller the cost, the easier the link is selected. A representative method for creating a route table that minimizes the total cost indicating the total link cost of the links constituting the route from the exchanged information is the Dijkstra's algorithm (Non-Patent Document 1). Non-Patent Document 2).

"Computer Network", Hideo Miyahara, Yuji Oie, Kyoritsu Shuppansha (1999) "Data Networks (2nd Edition)", D. Bertsekas, R. Gallager, Prentice-Hall (1992)

  In the Dijkstra method, a route having the minimum total cost (hereinafter referred to as a minimum cost route) is calculated, and a packet is delivered based on the minimum cost route. When the number of generated packets is as close to 0 as possible, this Dijkstra method is optimal. However, if the number of generated packets increases and congestion occurs in a node on the minimum cost route, the route avoiding the node where the congestion has occurred is recalculated, and the packet is calculated based on the recalculated route. Must be delivered. When congestion occurs in this way, in the conventional technology, it may be possible to avoid congestion by distributing some packets using a route different from the minimum cost route so as to bypass the node where the congestion has occurred. I came. However, it solves the problem of how many packets can be diverted to avoid congestion and how to prevent the packet from congesting at the detour node to avoid congestion Therefore, the route control of the entire network has not been studied.

  Therefore, an object of the present invention is to provide a technique for performing route control of the entire network to avoid congestion.

  The present invention relates to a path control apparatus that determines to which node a received packet is to be delivered next in order to avoid congestion in a node constituting the network, comprising the network topology of the network and the network A storage unit that stores a link cost indicating a cost allocated to a link between nodes to be read, a destination node of a received packet, and a destination node reading unit that determines whether the destination node is an adjacent node; When the destination node reading unit determines that the destination node is not an adjacent node, the minimum cost route that minimizes the total link cost of the links on the route to the destination node using the network topology and the link cost To the total number of the extracted minimum cost routes, and then The packet delivery probability indicating the ratio of the minimum cost route number of the route from the node delivering the packet to the destination node, and the parameter indicating the weight of the packet delivery probability at the boundary point between the congestion state and the non-congestion state A packet delivery probability matrix calculation for calculating the value of the parameter when the evaluation function has the extreme value using the variable of the evaluation function having the extreme value, and correcting the packet delivery probability using the parameter value as a weight And the destination node readout unit determine that the destination node is an adjacent node, the next node to be delivered is determined as the destination node, and the destination node readout unit determines that the destination node is not an adjacent node. If determined, random numbers are generated in the range of 0 to 1, and packet delivery probabilities of adjacent nodes are added in order. When it becomes larger than the random number generated in the adjacent node, the destination determination unit that determines the adjacent node as the next node to be delivered, and the packet that is sent to the next node that is determined by the delivery destination determination unit as the destination And a packet delivery unit for delivery.

  The present invention also relates to a route control method used in a route control apparatus for determining to which node a received packet is to be delivered next in order to avoid congestion in nodes constituting the network, The control device includes a storage unit that stores a network topology of the network and a link cost that indicates a cost assigned to a link between nodes configuring the network, and a processing unit, and the processing unit receives the received packet A destination node reading step for determining whether or not the destination node is an adjacent node, and when the destination node reading step determines that the destination node is not an adjacent node, the network topology and the link cost Is used to link the link on the route to the destination node. A packet delivery probability indicating a ratio of a minimum cost route number of a route from a node to which the next packet is delivered to the destination node to a total number of the extracted minimum cost routes. And a parameter indicating the weight of the packet delivery probability is used as a variable of an evaluation function having an extreme value at a boundary point between a congestion state and a non-congestion state, and when the evaluation function has the extreme value When it is determined that the destination node is an adjacent node by a packet delivery probability matrix calculation step for calculating the value of the parameter and correcting the packet delivery probability using the parameter value as a weight, and the destination node reading step, Next, a node to be delivered is determined as the destination node, and the destination node is adjacent to the destination node by the destination node reading step. If it is determined that the node is not a node, a random number is generated in the range from 0 to 1, and the packet delivery probability of the adjacent node is added in order. A delivery destination determination step for determining as a node to be delivered next, and a packet delivery step for delivering a packet with the next delivery node determined in the delivery destination determination step as a destination are executed.

  According to such a configuration, the path control device uses a parameter representing the weight of the packet delivery probability for the packet delivery probability calculated using the number of minimum cost routes to the destination node, and indicates the congestion state and non-congestion. The parameter value when the evaluation function has the extreme value is calculated as a variable of the evaluation function having the extreme value at the boundary point with the state, and the packet delivery probability is corrected using the parameter value as a weight. In this way, the route of the entire network for avoiding congestion at the boundary point between the congestion state and the non-congestion state by using the packet delivery probability corrected by the weight (hereinafter referred to as the weighted packet delivery probability). Control can be performed.

  In the present invention, the route control method is a program for causing the route control device, which is a computer, to execute. A computer in which such a program is installed can realize functions based on this program.

  ADVANTAGE OF THE INVENTION According to this invention, the technique which performs the route control of the whole network for avoiding congestion can be provided.

It is a figure which shows the network structural example in this embodiment. It is a figure which shows the structural example of a route control apparatus. It is a figure which shows the example of the processing flow of a route control apparatus. It is a figure which shows the example of the processing flow which calculates the parameter r from which the evaluation function in a packet delivery probability matrix calculating part becomes an extreme value. It is a figure which shows the example of the processing flow which calculates the packet delivery probability matrix based on the shortest cost path | route control in a packet delivery probability matrix calculating part. It is a figure which shows the network structural example in the modification 2. FIG. It is a figure which shows the structural example of the route control apparatus in the modification 2. FIG. It is a figure which shows the structural example of the delivery destination acquisition apparatus in the modification 2. FIG.

  A mode for carrying out the present invention (hereinafter referred to as “the present embodiment”) will be described in detail with reference to the drawings as appropriate.

  A network configuration example according to this embodiment will be described with reference to FIG. As illustrated in FIG. 1, the network 20 includes nodes 10 (10A, 10B, 10C, 10D, 10E, 10F) such as routers and links between the nodes 10. A terminal 30 such as a PC is connected to one of the nodes 10. In FIG. 1, the terminal 30A is connected to the node 10B, and the terminal 30B is connected to the node 10E. Each node 10 includes a path control device 11. The route control device 11 determines to which adjacent node the packet is to be delivered based on a predetermined calculation. Then, the node 10 delivers a packet received from the terminal 30 (hereinafter referred to as a generated packet) or a packet that has arrived from another node 10 to an adjacent node determined by the path control device 11.

(Configuration of path control device)
Next, a configuration example of the path control device 11 will be described with reference to FIG. As illustrated in FIG. 2, the route control device 11 includes a processing unit 12, a storage unit 13, and a communication unit 14. The processing unit 12 is configured by a CPU (Central Processing Unit) and a main memory (not shown). The application program stored in the storage unit 13 is expanded in the main memory, and a packet acquisition unit 121 and a destination node reading unit 122 to be described later. The packet delivery probability matrix calculation unit 123, the delivery destination determination unit 124, and the packet delivery unit 125 are implemented.

The packet acquisition unit 121 acquires the packet that has arrived the oldest from among the packets temporarily stored in the buffer 131 of the storage unit 13 to be described later.
The destination node reading unit 122 reads the destination node of the packet acquired by the packet acquisition unit 121.
The packet delivery probability matrix calculation unit 123 calculates a packet delivery probability, which will be described later, to generate a packet delivery probability matrix in order to determine which adjacent node to deliver.
The delivery destination determination unit 124 generates a random number (0 to 1) in a random number generation unit (not shown) for each adjacent node, compares the random number with the packet delivery probability of the packet delivery probability matrix, and delivers the packet to the adjacent node Is determined probabilistically. For example, a random number is generated in the range of 0 to 1, and the packet delivery probabilities of adjacent nodes are added in order. When the random number is larger than a random number generated in a certain adjacent node, the packet is delivered to the next adjacent node. It is determined as a node to perform.
The packet delivery unit 125 delivers a packet with the adjacent node determined by the delivery destination determination unit 124 as a destination.

The storage unit 13 stores a network topology 132, a queue length 133, a link cost 134, and a delivery probability matrix 135, and includes a buffer 131.
The buffer 131 temporarily accumulates packets that have arrived at the destination node and packets that have occurred. The newly accumulated packet is added to the end of the already accumulated packet. Then, the packet acquisition unit 121 reads out the oldest accumulated packet.
The network topology 132 is information regarding the connection form of each node 10. For example, in FIG. 1, the information is that the node 10A is connected to the nodes 10B, 10D, 10E, and 10F.
The queue length 133 refers to a packet waiting for delivery in the buffer 131 as a queue, and represents the number of queues as the queue length.
The link cost 134 is a cost assigned to the link between the nodes 10. The delivery probability matrix 135 stores packet delivery probabilities calculated in other nodes 10. This is because the packet delivery probability can only be calculated related to its own node and the adjacent node, and the packet delivery probability of the other node 10 needs to be received from the other node 10.

  The communication unit 14 is an interface for communicating with an adjacent node, and performs packet transmission / reception control.

(Processing flow of the route control device)
Next, an example of the processing flow of the route control device 11 will be described with reference to FIG. 3 (see FIG. 2 as appropriate).
In step S <b> 301, the packet acquisition unit 121 acquires from the buffer 131 the packet at the head (the oldest accumulated) of the queue.
In step S <b> 302, the destination node reading unit 122 reads the destination node of the packet acquired by the packet acquisition unit 121.
In step S303, the destination node reading unit 122 determines whether the destination node is an adjacent node. If the destination node is not an adjacent node (No in step S303), the process proceeds to step S304. If the destination node is an adjacent node (Yes in step S303), the process proceeds to step S306.

In step S304, the packet delivery probability matrix computing unit 123 computes the packet delivery probability and generates a packet delivery probability matrix. Details of the calculation of the packet delivery probability will be described later.
In step S305, the delivery destination determination unit 124 generates a random number, and probabilistically determines a delivery destination (adjacent node to deliver) based on the random number and the packet delivery probability calculated from the packet delivery probability matrix. Then, the process proceeds to step S307.
In step S306, the delivery destination determination unit 124 sets an adjacent node as a delivery destination.
In step S307, the packet delivery unit 125 delivers the packet to the delivery destination determined in step S305 or step S306. Then, the process proceeds to step S301.

(Calculation of packet delivery probability)
Next, an example of calculating the packet delivery probability will be described below.
First, preconditions (a) to (e) for generating a packet delivery probability matrix are shown.
(A) A packet is generated at an arbitrary time in each node 10.
(B) The destination node of a packet is determined according to its purpose at the node where the packet is generated.
(C) The destination node of the packet is described inside the packet, such as the header part of the packet.
(D) The destination node is a node other than the node where the packet always occurs.
(E) A plurality of packets may be generated simultaneously in one node.

(Packet delivery probability matrix and packet arrival rate)
Next, the packet delivery probability matrix and the packet arrival rate will be described. As shown in step S305 in FIG. 3, the node to which the packet is to be delivered next is determined stochastically based on the packet delivery probability and the random number. Here, a packet delivery probability that is a matrix element of the packet delivery probability matrix P is expressed as p _ij . The packet delivery probability p _ij represents the probability that the packet at the current node i is next delivered to the adjacent node j. The packet delivery probability matrix P is obtained by arranging packet delivery probabilities p _ij in a two-dimensional array for all i and j pairs. Such route control based on the packet delivery probability matrix P is based on Markov property regardless of the storage of the past route traced so far. Here, in the present embodiment, a case is considered where the packet delivery probability p _ij is parameterized by a relatively small number of variables r. That is, the parameter r is introduced so that the value of the packet delivery probability p _ij can be determined based on a certain policy. The policy is expressed by an evaluation function using the parameter r as a variable. For example, the policy is to avoid congestion by minimizing the number of packets in the network or to maximize the number of packets that fall into a congestion state. A specific example of the evaluation function is a function for minimizing the number of packets in the network or a function for maximizing the number of occurrences of packets falling into a congestion state. These evaluation functions are functions that represent boundary points between the congestion state and the non-congestion state when the value is an extreme value. In the evaluation function described above, a packet arrival rate (to be described later) indicating a congestion state at a node is used as a variable. This packet arrival rate is expressed using the parameter r as a variable, and is important for determining the congestion state. For example, when the packet arrival rate is large, a congestion state is likely to occur. Therefore, the packet delivery probability matrix and the packet arrival rate determined by the parameter r will be described below.

The packet delivery probability p _ij is expressed as shown in Equation (1). In the following, the character with an arrow (→) above the character in the formula represents a vector. However, in the text, the → is omitted.

Here, the average packet arrival rate b _j to the node _j includes the number of packets generated per unit time (packet generation rate) λ at the node j and the number b of packets transferred from the adjacent node i per unit time. It becomes the sum with _i p _ij and is expressed by equation (2).

Equation (2) is changed into e = (1,1,..., 1) ^t (where the number of matrix elements 1 is n), b = (b ₁ , b ₂ ,..., B _n ) ^t , r = ( r ₁ , r ₂ ,..., r _n ) ^t , [P (r)] _ij = p _ij (r), and expressed as a matrix, can be expressed by equation (3) The superscript t represents transposition.

ただし、式（４）中のＩは単位行列を示す。 Therefore, by solving the equation (3), the average packet arrival rate b (a function of the parameter r) becomes the equation (4).

However, I in Formula (4) shows a unit matrix.

のように表す。 Here, the packet delivery probability when the packet delivery probability differs depending on the destination node k of each packet as in the shortest path control, etc.

It expresses like this.

ただし、ｅ^kは、(1,…,1,0,1,…,1)^ｔ、つまり、ｋ番目の要素だけ０であとはすべて１のベクトルのことである。また、Ｎは、ノード数である。 In this case, the average packet arrival rate b ^k of the destination node ^k is expressed as Expression (5) if the destination node is randomly selected other than its own node.

However, e ^k is (1,..., 1,0, 1,..., 1) ^t , that is, if only the k-th element is 0, it is a vector of all 1s. N is the number of nodes.

Therefore, the average packet arrival rate b (r) for all destination nodes is obtained as equation (6).

(Determination of parameter r)
Here, a method for determining the parameter r will be described. First, regarding the packet delivery, the following two cases will be described as policy examples.
(Case A) Minimize the number of packets in the network (Case B) Maximize the number of packets that fall into a congestion state

((Case A) Case of minimizing the number of packets in the network)
According to Little's law, the number of packets in the network is proportional to the average packet delivery time (the average time taken to deliver a packet from the node where the packet occurred to the destination node). Therefore, if the number of packets in the network is reduced, the time required to reach the destination destination node can be reduced.
The number of packets q _i accumulated in the buffer 131 at a certain node i can be expressed by Equation (7) from the queue theory.

However, μ is the average number of packets that the node i transfers per unit time, the transfer rate, or the service rate in terms of queue theory.
Therefore, the total number of packets N (r) is the sum of q _i and can be expressed by Expression (8). That is, Expression (8) is the evaluation function described above.

つまり、Ｎ（ｒ）が最小となるパラメータｒが、ネットワーク内のパケット数を最小化するパケット配送確率のパラメータとなる。 Therefore, the number of packets in the network N to (r) in case that minimizes the value of the parameter r _m when N (r) is an extreme value is formulated as equation (9).

That is, the parameter r that minimizes N (r) is a packet delivery probability parameter that minimizes the number of packets in the network.

((Case B) Case of maximizing the number of packets that fall into a congestion state)
Next, Case B will be described. Packets start to be generated at time t = 0, and if the total number of packets generated up to time t is G (t), <G (t)> = λNt. Here, <> represents an expected value, λ represents a packet generation rate, and N represents the number of nodes. Further, assuming that the number of packets currently staying in the network is M (t), M (t) is expressed by Expression (10) when considering a sufficiently large time t in which the influence of the transition period can be ignored. .

When the packet generation rate λ is smaller than a certain threshold value λc, the total number of packets in the network becomes a constant value f (λ) determined by the packet generation rate λ. In this case, according to Little's law, the packet arrives at the destination node in a finite time. This state is that the network is in a non-congested state. On the other hand, when λ is greater than or equal to the threshold λc, the total number of packets in the network increases in proportion to time t at a constant rate h (λ). In this case, the packet arrival time will increase as much as possible, and this state is said that the network is in a congested state.
Here, η shown in Expression (11) is used as an index for determining the congestion state of the network.

In Expression (11), η = 0 in the non-congested state and η = b (λ) / (λN)> 0 in the congested state, so it is determined whether the value of η is 0 or positive. By doing so, it is possible to determine whether it is a non-congested state or a congested state. Here, λc that shifts from a non-congested state to a congested state is called a critical packet generation rate. When all of the nodes i satisfy Expression (12), the network is obtained from Expressions (7) and (8). Since the number of packets is kept almost constant, η = 0 and it can be said that the packet is in a non-congested state.

Note that when λ is gradually increased, λ satisfying equation (13) is λc for the first time, and of course, λc is also a function of r. That is, Expression (13) is the evaluation function described above.

In this case, in the case to maximize the number of packets generated falling into a congestion state, the values of the parameters r _m that maximizes λc is formulated as equation (14).

  That is, r that maximizes λc (r) is a parameter of the packet delivery probability when maximizing the packet occurrence rate of transition from the non-congested state to the congested state.

The processing flow described in the above cases (A) and (B) is summarized, and the processing flow for calculating the parameter r at which the evaluation function has an extreme value in the packet delivery probability matrix calculation unit 123 is shown in FIG. Will be described.
In step S401, the packet delivery probability is expressed as a function of the parameter r as shown in equation (1).
In step S402, the average packet arrival rate at each node 10 is calculated. The average packet arrival rate is expressed by Equation (6).
In step S403, a parameter r at which the evaluation function (variable r) is an extreme value is calculated. For example, the parameter r is calculated by the equation (9) for the case (A) and the equation (14) for the case (B). When the parameter r is calculated, the packet delivery probability p _ij (r) defined by the equation (1) can be calculated.

As a method for calculating the parameter r, as shown in the equation (15), for the case (A), the extreme value is calculated by differentiating N (r) in the equation (9). For case (B), the extremum is calculated by differentiating λc (r) in equation (14). Note that linear programming, quadratic programming, steepest descent, or the like may be used.

(Processing flow of packet delivery probability matrix calculator)
Next, the processing flow of the packet delivery probability calculation unit will be described below. In the processing flow, to which node a packet is delivered next is stochastically determined by a packet delivery probability and a random number. The packet delivery probability p _ij ^k represents the probability that the packet at node i is next delivered to the adjacent node j when the destination node of the packet is k. By introducing such a packet delivery probability matrix having the packet delivery probability p _ij ^k as a matrix element, it is possible to realize route control that does not depend on the storage of past routes that have been traced so far (No. 1). 1 treatment). Further, after calculating the packet delivery probability p _ij ^k that enables the shortest path control in the first processing, and adding the delivery weight for each node to the packet delivery probability p _ij ^k , the packet of a specific node Concentration can be avoided (second process).
First, a processing flow for calculating a packet delivery probability matrix that enables shortest path control will be described as the first processing. Next, as a second process, a processing flow for calculating a packet delivery probability matrix taking delivery weights into account will be described.

(First process)
Here, as the shortest path, the minimum cost path that minimizes the total cost is used. When delivering a packet by a minimum cost route, a route of a node series that uses the minimum link cost from the node S where the packet is generated to the destination node is used. When there are a plurality of minimum cost routes to the destination node, each minimum cost route is selected at random with equal probability. That is, the packet delivery probability p _ij ^k is expressed by dividing the number of minimum cost paths passing through the adjacent node j by the total number of minimum cost paths to the destination node k. A specific packet distribution probability calculation algorithm will be described below.

First, a certain node k is fixed. When the node k is the destination node, the packet delivery probability p _ij ^k at each node i is ^obtained . Since the packet delivery probability p _ij ^k is fixed, k is abbreviated and the superscript k is omitted. The symbols used in the algorithm are as follows.

V: All node set P: Node set with minimum cost E _j : Node set to be delivered from node j N _j : Number of shortest cost paths from node j d _j : Cost of node j to node k Total; that is, a value obtained by adding all link costs c between nodes on the path from node j to node k c _ij : link cost from node i to node j

The processing flow of the first process in the packet delivery probability matrix calculation unit 123 will be described with reference to FIG. 5 (see FIG. 2 as appropriate).
In step S <b> 501, the packet delivery probability matrix calculation unit 123 acquires the network topology 132 and the link cost 134 from the storage unit 13.
In step S502, the packet delivery probability matrix calculation unit 123 initializes calculation variables as shown in Expression (16). In addition, the backslash in Formula (16) means to exclude. {S} represents a node set including the node S that is the starting point of the route.

In step S503, the packet delivery probability matrix calculation unit 123 extracts a node for which the minimum cost is determined, and calculates the packet delivery probability of that node.
Specifically, a node set Q that satisfies Expression (17) is calculated.

The right side of Expression (17) is a function that returns a node k that minimizes _dj . That is, the node set Q is a node set for which the minimum cost is determined this time. Then, the minimum cost path number N _j of all the nodes j contained in the node set Q, the minimum cost path number N _l node l to be delivered from the node j, as shown in equation (18), the packet Define delivery probability p _ij .

Then, as shown in Expression (19), a new node set P is obtained by adding the node set Q to the current node set P (the node set for which the minimum cost is determined).

Next, in step S504, the packet delivery probability matrix calculation unit 123 updates the provisional minimum cost of a node whose minimum cost is not yet determined, extracts a delivery destination node that realizes the provisional minimum cost, and the minimum cost route at that time Calculate the number. Further, the packet delivery probability matrix calculation unit 123 applies the calculated minimum cost path number to the equation (18) to calculate the packet delivery probability.
Specifically, as shown in Expression (20), the node set E _j to be delivered from the node _j , the cost d _j from the node j to the node k, the minimum cost path number N _j from the node _j , Update.

  In step S505, the packet delivery probability matrix calculation unit 123 determines whether there is a node whose minimum cost is undetermined. Specifically, it is determined whether P = V (all minimum costs are determined). If there is a node whose minimum cost is not yet determined (Yes in step S505), the process returns to step S503. If there is no node for which the minimum cost has not yet been determined (No in step S505), in step S506, the packet distribution unit 125 distributes the calculated packet distribution probability to generate a packet distribution probability matrix in another node. To do.

(Second process)
In the second process, the packet delivery probability matrix calculation unit 123 calculates a weighted packet delivery probability considering the delivery weight for each node with respect to the packet delivery probability of the minimum cost path calculated in the first process. The delivery weight means that the packet delivery probability of the minimum cost route calculated in the first process is adjusted according to the above policy. Here, the packet delivery probability of the minimum cost route calculated in the first process is expressed as follows. Note that (S) means a minimum cost route.

Here, assuming that the delivery weight assigned to the node i is r _i , the packet delivery probability can be expressed by Expression (21).

A value obtained by normalizing Expression (21) is re-expressed as Expression (22) as a weighted packet delivery probability.

That is, in the second process, the average packet arrival rate b _j is expressed as Expression (23) by transforming Expression (2). However, c _{i on} the right side of Expression (23) can be expressed by Expression (24). That is, p _ij (r) on the right side of Equation (2) can be replaced with p _ij r _j (r).

When Expression (23) is expressed as a matrix, it is expressed as Expression (25).

However, diag (r) is a diagonal component, r _1, r _2, · ·, represents a diagonal matrix is r _n, 1. / C, each component representing a vector of 1 / c _j. Therefore, the average packet arrival rate b is expressed as in Equation (26).

  Then, b (r) of the equation (26) is applied to the equation (6) to calculate the extreme values of the equations (9) and (14) as shown in the above cases (A) and (B). By doing so, it is possible to calculate the parameter r at which the evaluation function becomes an extreme value. Then, the packet delivery probability calculated in the first process is corrected by applying the calculated parameter r to Equation (21).

(Simulation example)
The evaluation results when the weighted minimum cost path control shown in the second process described above is performed in the random network by computer simulation will be described below.
In the computer simulation, the optimum weight for minimizing the number of packets in the network in the case (A) in which the packet generation rate of each node is 0.1 per unit time and the number of packets in the network is minimized. r was calculated.
As the packet delivery model, steps S301 to S307 in FIG. 3 are executed based on the preconditions (a) to (e) when the packet delivery probability matrix is generated. The delivery weight r was calculated according to the second process described above. The extreme values were calculated using an optimization tool box of matlab (registered trademark). In addition, the random network is artificially configured, and the number of nodes is 10.

Table 1 shows the results of the computer simulation. The case of “no weight” in Table 1 is a case where the weights r are all equal (1.00), and the average value of the number of packets in the network is 7.35. In the case of “optimum weight”, the average value of the number of packets in the network is 6.78, which is clearly smaller than that in the case of “no weight”.

  Therefore, it has been found that calculating the weight assigned to each node by the calculation method described in the present embodiment has the effect of reducing the number of packets in the network and alleviating the congestion state.

(Modification 1)
The routing control method described above can be processed on hardware and software for a general network, but in the first modification, the routing control method described above is also applied to a virtually generated overlay network. Explain what you can do.
An overlay network is a network constructed by a connection established between terminals. The terminal may be software or a hardware device such as a server. By applying the route control method described above to route control of an overlay network that requires low delay and high throughput, route delay and throughput that avoid a congestion state are realized. Furthermore, in the present embodiment, the average packet arrival rate at each node is calculated. However, when the terminal is considered as a node in the overlay network, the terminal is inferior in packet processing capability as compared with the router device. Therefore, for the terminal, congestion from various viewpoints can be achieved by devising the evaluation function, such as reducing the processing load on the terminal by adding a small weight to the calculated average packet arrival rate and incorporating it into the evaluation function. It is possible to avoid the situation.

(Modification 2)
In the above-described embodiment, as illustrated in FIG. 1, the configuration in which each node 10 includes the path control device 11 has been described. In the second modification, the processing of the packet delivery probability matrix calculation unit 123 and the delivery destination determination unit 124 of the route control device 11 is executed in the route control device 41 shown in FIG. That is, the path control device 41 calculates the packet delivery probability for all the nodes 10 constituting the network 20 and determines the delivery destination. The path control device 41 is communicably connected to any one of the nodes 10 constituting the network 20. Then, the delivery destination acquisition device 51 of the node 10 receives the delivery destination information determined by the route control device 41 and delivers the packet to the delivery destination. Below, the structural example of the route control apparatus 41 and the delivery destination acquisition apparatus 51 is demonstrated. In FIG. 6, only one path control device 41 is shown, but the node 10 in charge of determining the delivery destination is divided into two or more groups, and the path control device 41 is provided for each group. It doesn't matter.

First, a configuration example of the route control device 41 will be described with reference to FIG. In addition, the same code | symbol is attached | subjected to the same structure as FIG. 2, and description is abbreviate | omitted.
As illustrated in FIG. 7, the route control device 41 includes a processing unit 42, a storage unit 43, and a communication unit 14. The processing unit 42 includes a CPU and a main memory (not shown). The application program stored in the storage unit 43 is expanded in the main memory, and a destination node information acquisition unit 421, a packet delivery probability matrix calculation unit 123, a delivery destination The determination unit 124 and the delivery destination information transmission unit 422 are implemented.

The destination node information acquisition unit 421 acquires information regarding the destination node from the delivery destination acquisition device 51 of the node 10 via the buffer 131.
The delivery destination information transmission unit 422 transmits information on the delivery destination determined by the packet delivery probability matrix calculation unit 123 and the delivery destination determination unit 124 to the delivery destination acquisition device 51 of the node 10 that has transmitted the information on the destination node. .
The storage unit 43 stores at least a buffer 131, a network topology 132, and a link cost 134.

  Next, a configuration example of the delivery destination acquisition apparatus 51 will be described with reference to FIG. As illustrated in FIG. 8, the delivery destination acquisition device 51 includes a processing unit 52, a storage unit 53, and a communication unit 14. The processing unit 52 includes a CPU and a main memory (not shown). The application unit stored in the storage unit 53 is expanded in the main memory, and the packet acquisition unit 121, the destination node reading unit 122, and the destination node information transmission unit 523 are expanded. The delivery destination information acquisition unit 524 and the packet delivery unit 125 are implemented.

The destination node information transmission unit 523 transmits information regarding the destination node to the route control device 41.
The delivery destination information acquisition unit 524 acquires information about the delivery destination from the route control device 41.

  As shown in FIG. 6, the advantage that the route control device 41 is provided independently of the node 10 is that there is no need to deliver the packet delivery probability to the other nodes 10 in step S506 shown in FIG. The processing load for calculating the packet delivery probability in the node 10 can be reduced.

As described above, the packet delivery probability matrix calculation unit 123 of the route control device 11 described in the present embodiment determines to which node the packet is to be delivered next based on the packet delivery probability and the random number. Specifically, a parameter r used for calculating the packet delivery probability is introduced, and an average packet arrival rate that is a variable of an evaluation function for calculating a boundary point between the congestion state and the non-congestion state is calculated. That is, the average packet arrival rate is expressed using the parameter r as a variable. Next, in the evaluation function, the value of the parameter r serving as the boundary point is calculated, and the packet delivery probability is calculated using the value of the parameter r. By using the packet delivery probability calculated in this way, it is possible to perform route control of the entire network to avoid congestion.
In the first process, the minimum cost route is used as the shortest route, but instead of the link cost, the route with the smallest number of link hops may be used as the shortest route.

  Further, the path control device 11 described in the present embodiment can be realized by causing a general computer to execute the above-described processing programs. The program can be provided via a communication line, or can be written and distributed on a recording medium such as a CD-ROM (Compact Disc-Read Only Memory).

10 node 11, 41 route control device 12, 42 processing unit 13, 43 storage unit 121 packet acquisition unit 122 destination node reading unit 123 packet delivery probability matrix calculation unit 124 delivery destination determination unit 125 packet delivery unit 131 buffer 134 link cost 135 delivery Probability matrix

Claims (3)

A path control device that determines to which node a received packet is to be delivered next in order to avoid congestion in a node constituting the network,
A storage unit storing a link cost indicating a network topology of the network and a cost assigned to a link between nodes constituting the network;
A destination node reading unit that reads a destination node of the received packet and determines whether the destination node is an adjacent node;
When the destination node reading unit determines that the destination node is not an adjacent node, the minimum cost that minimizes the total link cost of the links on the route to the destination node using the network topology and the link cost A route is extracted, and a packet delivery probability indicating a ratio of a minimum cost route number of a route from a node to which a packet is delivered next to the destination node to a total number of the extracted minimum cost routes is calculated, and the packet delivery probability Using the parameter indicating the weight of the evaluation function as a variable of the evaluation function having the extreme value at the boundary point between the congestion state and the non-congestion state, and calculating the value of the parameter when the evaluation function has the extreme value, A packet delivery probability matrix calculation unit for correcting the packet delivery probability using a parameter value as a weight;
When the destination node reading unit determines that the destination node is an adjacent node, the next node to be delivered is determined as the destination node, and the destination node reading unit determines that the destination node is not an adjacent node In this case, random numbers are generated in the range of 0 to 1, and the packet delivery probabilities of the adjacent nodes are added in order. A delivery destination determination unit to be determined;
A route control apparatus comprising: a packet delivery unit that delivers a packet with the next delivery node determined by the delivery destination determination unit as a destination. A path control method used in a path control device that determines to which node a received packet is to be delivered next in order to avoid congestion in nodes constituting the network,
The route control device
A storage unit storing a link cost indicating a network topology of the network and a cost allocated to a link between nodes constituting the network, and a processing unit,
The processor is
A destination node reading step of reading a destination node of the received packet and determining whether the destination node is an adjacent node;
If the destination node read step determines that the destination node is not an adjacent node, the minimum cost that minimizes the total link cost of the links on the route to the destination node using the network topology and the link cost A route is extracted, and a packet delivery probability indicating a ratio of a minimum cost route number of a route from a node to which a packet is delivered next to the destination node to a total number of the extracted minimum cost routes is calculated, and the packet delivery probability Using the parameter indicating the weight of the evaluation function as a variable of the evaluation function having the extreme value at the boundary point between the congestion state and the non-congestion state, and calculating the value of the parameter when the evaluation function has the extreme value, Packet delivery probability matrix calculation step for correcting the packet delivery probability using a parameter value as a weight ,
When the destination node reading step determines that the destination node is an adjacent node, the next node to be delivered is determined as the destination node, and the destination node reading step determines that the destination node is not an adjacent node In this case, random numbers are generated in the range of 0 to 1, and the packet delivery probabilities of the adjacent nodes are added in order. A delivery destination determination step to be determined;
And a packet delivery step of delivering a packet to the next delivery node determined in the delivery destination determination step as a destination.   The program for making the said route control apparatus which is a computer perform the route control method of Claim 2.

Images