JP2012227707A - Packet relay system, packet relay method, and program for packet relay - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a packet relay system capable of more effectively dispersing a transfer load for relaying packets.SOLUTION: Transfer probability calculation means 81 calculates transfer probability that is probability of dispersing the packets to respective nodes in a network according to topology information. Shortest route determination means 82 determines the shortest route to the destination node of the packets according to the topology information. Transfer destination determination means 83 stochastically selects a node selected according to transfer probability and a node in the shortest route and determines them as transfer destination nodes.

Description

  The present invention relates to a packet relay system, a packet relay method, and a packet relay program for distributing a transfer load when relaying a packet.

  In recent years, it is known that many networks including the Internet have a scale-free property. The scale-free property refers to a feature in which the number of links connected to each node (hereinafter also referred to as an order) forms a wrinkle distribution. In a network having scale-free properties, there are only a small number of nodes having a large degree. This node is called a hub node.

  It is also known that a packet often passes through a hub node in the shortest path connecting two arbitrary points in the network. For this reason, when the packets are always transferred so as to pass through the shortest path to the destination node, the packets to the hub node are concentrated, resulting in congestion.

  In a communication network, avoiding communication traffic congestion and minimizing packet loss is one of the most important issues. One of the simplest congestion avoidance measures is to monitor the communication traffic of the entire network in real time and bypass the packet before congestion occurs. However, when the size of the network increases, it is difficult to monitor communication of the entire network, and an effective detour cannot be set.

  A packet route control method described in Patent Document 1 is known as a packet relay method for avoiding the concentration of packets to a hub node. In the packet route control method described in Patent Document 1, the shortest route control in which an adjacent node in the shortest route direction is a transfer destination, and a random walk in which one of the adjacent nodes is uniformly and randomly selected as a transfer destination. The packet transfer destination is determined in combination with the control.

  Specifically, in the packet route control method described in Patent Document 1, the probability that the node i forwards the packet whose destination node is the node k to the adjacent node j is expressed by the following Expression 1.

Here, p ^{(S) k} _ij is a probability that the node i forwards the packet whose destination node is k to the node j by the shortest path control. Further, p ^{(R) k} _ij is a probability that the node i forwards the packet whose destination node is k to the node j by random walk control.

  This equation 1 means that the packet transfer process by the shortest path control and the packet transfer process by the random walk control are selected at random with probabilities δ and 1-δ, respectively. As described above, in the packet route control method described in Patent Document 1, the distribution of packets to nodes where packets are concentrated is reduced by distributing the route of packets to other than the shortest route.

Patent Document 1 describes a method for calculating the probability of transferring a packet to an adjacent node. In the calculation method described in Patent Document 1, the queue length q _l of each adjacent node ₁ is monitored as the network load status. Then, a value obtained by weighting the probability calculated by Expression 1 with the queue length of the adjacent node is set as the packet transfer probability. The packet transfer probability in this case is expressed by the following Equation 2.

  Here, f (q) is an arbitrary decreasing function with respect to q (ie, queue length). In the packet path control method described in Patent Document 1, the load is distributed by transferring the packet to an adjacent node having a lower load by using the packet transfer probability.

  Note that Non-Patent Document 1 describes a method for calculating a packet transfer probability in which, when a packet is transferred by a hop-by-hop method, a trajectory through which the packet passes is random.

JP 2010-178062 A

Z. Burda, J .; Duda, J. et al. M.M. Luck, and B.L. Waclaw, “Localization of maximum entropy random walk”, Phys. Rev. Lett. 102, 160602, 2008

  By using the packet path control method described in Patent Document 1, it is possible to avoid the concentration of packets to the hub node to some extent without monitoring the communication of the entire network. However, even if the packet route control method described in Patent Document 1 is used, there is a problem that the effect of packet transfer load distribution is small. In the packet route control method described in Patent Document 1, random walk control is used to uniformly and randomly transfer one of adjacent nodes in order to distribute the packet transfer load. In this random walk control, This is because the packets are concentrated on the hub node. Hereinafter, this reason will be described.

Consider a model in which a packet is transferred in the network by random walk control. That is, it is assumed that the packet moves at random every one time to one of the adjacent nodes of the node where the packet exists. When node j receives a packet, the probability T _ij of transferring the packet to node i is expressed by the following Equation 3.

Here, A _ij is a variable whose value becomes 1 when a link is established between the node i and the node j and becomes 0 when a link is not established. K _j is the number of nodes adjacent to node j. Further, the probability p _i ^(t) that the packet exists in the node i at time t is expressed by the following Expression 4.

Further, using a transfer probability matrix T having an i-th row and j-th column component T _ij and a column vector p (t) having an i-th component p _i , the above equation 4 can be expressed by the following equation 5. . T is a matrix of n rows and n columns, and n is the number of nodes in the network.

Assuming that sufficient time has elapsed and p ^(t) has converged to the steady state p ^* , the above equation 5 can be expressed by the following equation 6.

That is, in random walk control, the eigenvector corresponding to eigenvalue 1 of the transfer probability matrix is equal to the probability that a packet exists in each node in a steady state. Each component in p ^{* at} this time is called node centrality, and represents the ease of arrival of a packet when random walk control is performed.

  In scale-free networks, it is generally known that the centrality of hub nodes tends to be high. For this reason, it can be said that the packet transfer by the random walk control has the property of easily concentrating the packets to the hub node. Therefore, the packet route control method for transferring packets by random walk control described in Patent Document 1 is considered to have little effect on packet transfer load distribution.

  Therefore, an object of the present invention is to provide a packet relay system, a packet relay method, and a packet relay program that can more effectively distribute a transfer load when relaying a packet.

  The packet relay system according to the present invention includes a transfer probability calculation means for calculating a transfer probability that is a probability of distributing a packet to each node in the network based on the topology information, and a shortest distance to the destination node of the packet based on the topology information. The shortest path determining means for determining a path, and the transfer destination determining means for selecting the node selected according to the transfer probability and the node in the shortest path stochastically and determining the transfer destination node. To do.

  The packet relay method according to the present invention calculates a transfer probability that is a probability of distributing a packet to each node in the network based on the topology information, determines the shortest path to the destination node of the packet based on the topology information, A node selected according to a transfer probability and a node in the shortest path are selected probabilistically and determined as a transfer destination node.

  A packet relay program according to the present invention includes a transfer probability calculation process for calculating a transfer probability, which is a probability of distributing a packet to each node in a network, based on topology information, and a packet destination node based on topology information. The shortest path determination process for determining the shortest path to and the transfer destination determination process for selecting the node selected according to the transfer probability and the node in the shortest path stochastically to determine the transfer destination node It is characterized by.

  According to the present invention, it is possible to more effectively distribute the transfer load when relaying a packet.

It is a block diagram which shows the structural example of 1st Embodiment of the packet relay system by this invention. It is a flowchart which shows the operation example of the packet relay system of 1st Embodiment. It is a block diagram which shows the structural example of 2nd Embodiment of the packet relay system by this invention. It is a flowchart which shows the operation example of the packet relay system of 2nd Embodiment. It is a block diagram which shows the minimum structural example of the packet relay system by this invention.

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

Embodiment 1. FIG.
FIG. 1 is a block diagram showing a configuration example of a first embodiment of a packet relay system according to the present invention. In the following description, the packet relay system according to the present invention is referred to as a network node 100 or simply a node. The network node 100 according to the present embodiment includes a transfer destination determination unit 110, a packet distribution control unit 120, a shortest path control unit 130, a communication unit 140, and a topology information management unit 150.

  At least one network node 100 exists on the communication network, and each network node 100 operates similarly. Here, the operation of the node j will be described.

  The communication unit 140 communicates with adjacent nodes to exchange packets. When receiving the packet, the communication unit 140 inquires the transfer destination determination unit 110 about the transfer destination of the packet, and transfers the received packet to the transfer destination indicated by the transfer destination determination unit 110. The communication unit 140 may use any protocol as a communication protocol.

  The topology information management unit 150 collects network topology information and holds the collected topology information. The topology information management unit 150 also provides network topology information in response to a request from the packet distribution control unit 120 or the shortest path control unit 130. Here, the topology information includes information indicating the presence / absence of a link between the nodes in the network. However, the topology information may include information other than information indicating the presence / absence of a link between nodes.

  The transfer destination determination unit 110 determines the next transfer destination of the packet received by the communication unit 140. The transfer destination determination unit 110 reads the destination of the packet, and determines that the destination node is the transfer destination node if the destination is its own adjacent node. On the other hand, when the destination node of the packet is not its own adjacent node, the transfer destination determination unit 110 determines the transfer determined according to the transfer destination determined by the shortest path control unit 130 described later and the probability calculated by the packet distribution control unit 120. The destination of the packet is determined by selecting the destination stochastically.

  Here, the selection of the transfer destination stochastically means specifically, for example, the transfer probability calculated by the packet distribution control unit 120 based on a probability δ (0 ≦ δ ≦ 1) predetermined by an administrator or the like. It is also possible to select the node selected according to the probability δ and select the transfer destination determined by the shortest path control unit 130 with the probability 1−δ. In other words, the probability of transfer to each adjacent node calculated by the packet distribution control unit 120 and the transfer probability to the adjacent node determined by the shortest path control unit 130 (here, 100%) are weighted using the probability δ. It can also be said that an adjacent node is selected according to the transfer probability.

  Alternatively, the transfer destination determination unit 110 may monitor the load status of each node in advance, such as the queue length of an adjacent node. Then, the transfer destination determination unit 110 may select a packet transfer destination using a probability obtained by further weighting the transfer probability weighted using the probability δ according to the load status. Specifically, the transfer destination determination unit 110 may set the weighting more heavily as the load is smaller with respect to the transfer probability of the adjacent node.

  In the following, for description, a node selected according to the transfer probability calculated by the packet distribution control unit 120 is selected with a probability δ, and a transfer destination determined by the shortest path control unit 130 is selected with a probability 1−δ. Is described.

  In response to a request from the transfer destination determination unit 110, the packet distribution control unit 120 refers to the topology information held by the topology information management unit 150 and distributes the packets as uniformly as possible to each node in the network. The transfer probability to the node is calculated. That is, the packet distribution control unit 120 calculates a transfer probability that distributes the packet uniformly to each node.

  First, the packet distribution control unit 120 calculates a transfer probability matrix such that a packet exists with equal probability at each node in the network. Then, the packet distribution control unit 120 sets each element in the jth column in the calculated transfer probability matrix as the transfer probability to each adjacent node. Hereinafter, a method for calculating the transfer probability will be described in detail.

  The packet distribution control unit 120 creates a transfer probability matrix T ′ in which the eigenvector corresponding to the eigenvalue 1 is q based on the topology information, as shown in Equation 7 below. Here, when the number of nodes in the network is n, T ′ is an n-by-n matrix, and the vector q is an n-dimensional column vector in which all components are 1 / n.

  Here, the meaning represented by Expression 7 will be described. Consider a case where each node forwards a packet according to only the transfer probability specified by a certain transfer probability matrix (that is, the packet moves around in the network) ignoring the packet destination and the shortest path. In this case, it is known that after a sufficient amount of time has passed, the probability that a certain packet exists at the node i is equal to the i-th component of the eigenvector corresponding to the eigenvalue 1 of the transfer probability matrix. Therefore, after a sufficient time has passed, the transfer probability for a packet to exist at node i with probability 1 / n (that is, the probability that a packet exists at each node is equal) is the eigenvector corresponding to eigenvalue 1. It can be calculated by calculating a transfer probability matrix whose i-th component is 1 / n. That is, Expression 7 is an expression that approximately uses this property, and T ′ in Expression 7 represents a transfer probability matrix that uniformly distributes packets to each node.

Further, the transfer probability matrix T ′ needs to satisfy the following expressions 8, 9, and 10. Here, the element T ′ _ij of the transfer probability matrix T ′ represents the probability of transferring this packet to the node i when the node j receives the packet.

Expression 8 means that the probability that the node j receiving the packet transfers to the node i is 0 or more and 1 or less. Equation 9 means that the total probability that each column (ie, each node j) transfers to the adjacent node i is 1. Further, Expression 10 means that the transfer probability value becomes 0 when there is no link between the node i and the node j. Here, A _ij is a variable whose value becomes 1 when a link is established between the node i and the node j and becomes 0 when a link is not established. The presence / absence of a link between nodes is determined from the topology information. That is, Expression 10 represents a constraint condition of the transfer probability matrix T ′ determined by the presence / absence of a link between nodes.

  In general, there may be no T ′ that satisfies Expression 7 while satisfying the restrictions of Expressions 8 to 10. In this case, the packet distribution control unit 120 obtains T ′ so that the eigenvector r corresponding to the eigenvalue 1 is as close to q as possible while satisfying the constraints of Expressions 8 to 10. Then, the packet distribution control unit 120 notifies the transfer destination determination unit 110 of the element in the j-th column of T ′ as the transfer probability to each adjacent node. For example, the packet distribution control unit 120 may obtain T ′ that minimizes the function shown in the following Expression 11.

  The packet distribution control unit 120 may obtain T ′ that minimizes the function exemplified in Expression 11 using various methods known in the field of optimization, such as a hill-climbing method and a genetic algorithm.

  The shortest path control unit 130 searches the shortest path to the destination node of the packet with reference to the topology information management unit 150 in response to the request from the transfer destination determination unit 110. Then, the shortest path control unit 130 returns, to the transfer destination determination unit 110, transfer destination information for the packet to reach the destination node in the shortest time. The transfer destination information includes information indicating an adjacent node to which the communication unit 140 transfers a packet. The shortest path control unit 130 searches for the shortest path using an arbitrary method such as a Dijkstra method.

  The transfer destination determination unit 110, the packet distribution control unit 120, and the shortest path control unit 130 are realized by a CPU of a computer that operates according to a program (packet relay program). For example, the program is stored in a storage unit (not shown) of the network node 100, and the CPU reads the program and operates as the transfer destination determination unit 110, the packet distribution control unit 120, and the shortest path control unit 130 according to the program. May be. In addition, each of the transfer destination determination unit 110, the packet distribution control unit 120, and the shortest path control unit 130 may be realized by dedicated hardware.

  Next, the operation of the network node 100 in this embodiment will be described. First, when receiving a packet (step S201), the communication unit 140 inquires the transfer destination determination unit 110 about the transfer destination of the packet. The transfer destination determination unit 110 reads the destination node of the packet (step S202). If the destination of the packet is its own adjacent node (Yes in step S203), the transfer destination determination unit 110 replies to the inquiry from the communication unit 140 with the transfer destination of the packet as the destination node (step S207).

  On the other hand, when the destination of the packet is not its own adjacent node (No in step S203), the transfer destination determination unit 110 inquires the packet distribution control unit 120 about the transfer probability for the packet to be uniformly distributed in the network. Further, the transfer destination determination unit 110 inquires of the shortest path control unit 130 about a transfer destination for the packet to reach the destination node through the shortest path. The packet distribution control unit 120 calculates the transfer probability based on the topology information (step S204), and notifies the transfer destination determination unit 110 of the calculation result. Further, the shortest path control unit 130 determines a transfer destination based on the shortest path based on the topology information (step S205), and notifies the transfer destination determination unit 110 of the determined transfer destination.

  Next, the transfer destination determination unit 110 probabilistically selects the transfer destination according to the transfer probability provided by the packet distribution control unit 120 and the transfer destination determined by the shortest path control unit 130 (step S206). For example, the transfer destination determination unit 110 selects a transfer destination according to the transfer probability provided by the packet distribution control unit 120 with the probability δ, and selects the transfer destination determined by the shortest path control unit 130 with the probability 1−δ. Finally, the communication unit 140 transfers the packet to the transfer destination designated by the transfer destination determination unit 110 (step S208).

  As described above, according to the present embodiment, the packet distribution control unit 120 calculates the probability (transfer probability) of distributing a packet to each node in the network based on the topology information, and the shortest path control unit 130 The shortest path to the packet destination node is determined based on the topology information. Then, the transfer destination determination unit 110 probabilistically selects the node selected according to the transfer probability and the node in the shortest path as the transfer destination node. Therefore, the transfer load when relaying the packet can be more effectively distributed.

  In other words, in this embodiment, the packets are not always transferred to the shortest path, but are transferred so as to be uniformly distributed to the nodes in the network with a probability δ. As a result, packets can be prevented from concentrating on the hub node, and network congestion can be avoided.

  When the destination node is an adjacent node, the transfer destination determination unit 110 determines the adjacent node as the transfer node. Thus, when the transfer destination is clear, the transfer load can be effectively suppressed by suppressing useless transfer processing.

Embodiment 2. FIG.
FIG. 3 is a block diagram showing a configuration example of the second embodiment of the packet relay system according to the present invention. In addition, about the structure similar to 1st Embodiment, the code | symbol same as FIG. 1 is attached | subjected and description is abbreviate | omitted. The network node 300 in this embodiment includes a transfer destination determination unit 110, a packet distribution control unit 320, a shortest path control unit 130, a communication unit 140, a topology information management unit 150, and a transfer capacity information management unit 360. I have.

  That is, the network node 300 of this embodiment includes a transfer capacity information management unit 360, and the network node in the first embodiment is different from the operation of the packet distribution control unit 120 in the operation of the packet distribution control unit 320. Different from 100.

  The transfer capacity information management unit 360 collects information (hereinafter referred to as transfer capacity information) indicating the packet transfer capacity of each node in the network, and holds the collected transfer capacity information. Further, the transfer capacity information management unit 360 provides transfer capacity information of each node in response to a request from the packet distribution control unit 320.

  The packet distribution control unit 320 refers to the topology information held by the topology information management unit 150 and the transfer capacity information held by the transfer capacity information management unit 360, and distributes the packet to an amount proportional to the transfer capacity of each node. Calculate the transfer probability. That is, the packet distribution control unit 320 calculates a transfer probability for distributing packets according to the transfer capacity of each node.

Specifically, each component q ′ _j of the vector q ′ is defined as in the following Expression 12. Here, the vector q ′ is an n-dimensional column vector corresponding to the eigenvector q in the first embodiment. U _j represents the transfer capacity of each node j. That is, a larger value is set for q ′ _j corresponding to a node having a large transfer capacity.

  Then, the packet distribution control unit 320 obtains a transfer probability matrix T ′ in which the eigenvector corresponding to the eigenvalue 1 is a vector q ′ as shown in the following Expression 13. The packet distribution control unit 320 notifies the transfer destination determination unit 110 of the element in the jth column of T ′ as the transfer probability to each adjacent node.

  However, this T ′ also needs to satisfy the above formulas 8 to 10. However, there may be no T ′ that satisfies Expression 13. In this case, the packet distribution control unit 320 obtains T ′ where the eigenvector r corresponding to the eigenvalue 1 is as close to q ′ as possible while satisfying the constraints of Expressions 8 to 10. Then, the packet distribution control unit 320 notifies the transfer destination determination unit 110 of the element in the jth column of T ′ as the transfer probability to each adjacent node. For example, the packet distribution control unit 320 may obtain T ′ that minimizes the function shown in the following Expression 14.

  The packet distribution control unit 320 may obtain T ′ that minimizes the function exemplified in Expression 14 using various methods known in the field of optimization, such as a hill-climbing method and a genetic algorithm.

  The transfer destination determination unit 110, the packet distribution control unit 320, and the shortest path control unit 130 are realized by a CPU of a computer that operates according to a program (packet relay program). In addition, each of the transfer destination determination unit 110, the packet distribution control unit 320, and the shortest path control unit 130 may be realized by dedicated hardware.

  Next, the operation of the network node 300 in this embodiment will be described. First, when receiving a packet (step S401), the communication unit 140 inquires the transfer destination determination unit 110 about the transfer destination of the packet. The transfer destination determination unit 110 reads the destination node of the packet (step S402). If the destination of the packet is its own adjacent node (Yes in step S403), the transfer destination determination unit 110 replies to the inquiry from the communication unit 140 with the transfer destination of the packet as the destination node (step S407).

  On the other hand, when the destination of the packet is not its own adjacent node (No in step S403), the transfer destination determining unit 110 sets the transfer probability for distributing the packet to an amount proportional to the transfer capacity of each node in the network. Contact 320. Further, the transfer destination determination unit 110 inquires of the shortest path control unit 130 about a transfer destination for the packet to reach the destination node through the shortest path. The packet distribution control unit 320 calculates the transfer probability based on the topology information and the transfer capacity information (step S404), and notifies the transfer destination determination unit 110 of the calculation result. Further, the shortest path control unit 130 determines a transfer destination based on the shortest path based on the topology information (step S405), and notifies the transfer destination determination unit 110 of the determined transfer destination.

  Next, the transfer destination determination unit 110 probabilistically selects the transfer destination according to the transfer probability provided by the packet distribution control unit 320 and the transfer destination determined by the shortest path control unit 130 (step S406). For example, the transfer destination determination unit 110 selects a transfer destination according to the transfer probability provided by the packet distribution control unit 120 with the probability δ, and selects the transfer destination determined by the shortest path control unit 130 with the probability 1−δ. Finally, the communication unit 140 transfers the packet to the transfer destination designated by the transfer destination determination unit 110 (step S408).

  As described above, according to this embodiment, the packet distribution control unit 320 distributes an amount of packets proportional to the transfer capacity of each node to each node in the network based on the topology information and the transfer capacity information. Calculate the transfer probability. That is, in this embodiment, packets are not always transferred to the shortest path, but are transferred so as to be distributed in an amount proportional to the transfer capacity of the nodes in the network with probability δ. As a result, it is possible to prevent the packets from concentrating on the hub node and to reduce the transfer load of the node having a small transfer capacity, thereby avoiding network congestion.

  For example, in the packet route control method described in Patent Document 1, the transfer capacity of each node in the network is not considered. That is, a node having a small transfer capacity and a node having a large transfer capacity are handled in the same row. For this reason, in the packet path control method described in Patent Document 1, a node having a small transfer capacity becomes a bottleneck, which may cause congestion. However, according to the present embodiment, the probability of selection increases as the transfer capacity of the node increases, so that the packet transfer load can be more efficiently distributed.

  Next, a minimum configuration example of the present invention will be described. FIG. 5 is a block diagram showing a minimum configuration example of the packet relay system according to the present invention. The packet relay system according to the present invention has a transfer probability calculation means 81 (for example, a packet) that calculates a transfer probability (for example, a transfer probability matrix T ′) that is a probability of distributing a packet to each node in the network based on the topology information. The distribution control unit 120), the shortest route determination means 82 (for example, the shortest route control unit 130) that determines the shortest route to the destination node of the packet based on the topology information, the node selected according to the transfer probability, and the shortest A transfer destination determination unit 83 (for example, a transfer destination determination unit 110) that probabilistically selects a node in the route and determines it as a transfer destination node is provided.

  With such a configuration, it is possible to more effectively distribute the transfer load when relaying a packet.

  A part or all of the above-described embodiment can be described as in the following supplementary notes, but is not limited thereto.

(Supplementary note 1) Based on the topology information, a transfer probability calculating means for calculating a transfer probability that is a probability of distributing the packet to each node in the network, and a shortest path to the destination node of the packet is determined based on the topology information. A packet comprising: a shortest path determining unit; a node selected according to the transfer probability; and a transfer destination determining unit that probabilistically selects a node in the shortest path and determines a transfer destination node. Relay system.

(Supplementary note 2) The packet relay system according to supplementary note 1, wherein the transfer probability calculation means calculates a transfer probability that distributes packets substantially uniformly to each node in the network.

(Supplementary Note 3) Transfer capacity information storage means for storing transfer capacity information indicating the packet transfer capacity of each node in the network is provided, and the transfer probability calculation means is based on the topology information and the transfer capacity information. The packet relay system according to supplementary note 1, wherein a packet transfer probability that distributes an amount of packets proportional to a transfer capacity of the node to a node is calculated.

(Supplementary Note 4) The transfer probability calculation means calculates a transfer probability matrix having an eigenvalue of 1 with the probability that a packet exists at each node on the network as each component of the eigenvector, and transfers each element in the calculated transfer probability matrix The packet relay system according to any one of Supplementary Note 1 to Supplementary Note 3 that is a probability.

(Supplementary note 5) The packet relay system according to any one of supplementary notes 1 to 4, wherein the forwarding destination determining unit determines that the destination node is a forwarding node when the destination node is an adjacent node.

(Supplementary note 6) Based on the topology information, a transfer probability that is a probability of distributing the packet to each node in the network is calculated, a shortest path to the destination node of the packet is determined based on the topology information, and the transfer probability A packet relay method characterized by probabilistically selecting a node selected accordingly and a node on the shortest path to determine a forwarding destination node.

(Additional remark 7) The packet relay method of additional remark 6 which calculates the transfer probability which distributes a packet substantially uniformly to each node in a network.

(Supplementary note 8) Supplementary note 6 for calculating, based on the packet transfer capacity and topology information of each node in the network, a transfer probability for distributing an amount of packets proportional to the packet transfer capacity of the node to each node in the network Packet relay method.

(Supplementary note 9) Transfer probability calculation processing for calculating a transfer probability that is a probability of distributing a packet to each node in the network based on the topology information, and a shortest path to the destination node of the packet based on the topology information Packet relay for executing the shortest path determination process to be determined, and the transfer destination determination process in which a node selected according to the transfer probability and a node in the shortest path are stochastically selected and determined as a transfer destination node Program.

(Additional remark 10) The packet relay program of Additional remark 9 which makes a computer calculate the transfer probability which distributes a packet substantially uniformly to each node in a network by a transfer probability calculation process.

(Supplementary note 11) In a transfer probability calculation process, an amount of packets proportional to the packet transfer capacity of the node is distributed to each node in the network based on the packet transfer capacity and topology information of each node in the network. The packet relay system according to appendix 9, wherein a transfer probability is calculated.

  The present invention is preferably applied to a packet relay system that realizes packet transfer load distribution. For example, a communication network can be constructed using the packet relay system according to the present invention. The present invention can be realized by, for example, a CPU of a computer that operates according to a packet relay program.

100, 300 Network node 110 Transfer destination determination unit 120, 320 Packet distribution control unit 130 Shortest path control unit 140 Communication unit 150 Topology information management unit 360 Transfer capacity information management unit

Claims (10)

Transfer probability calculation means for calculating a transfer probability, which is a probability of distributing packets to each node in the network, based on topology information;
Shortest path determination means for determining the shortest path to the destination node of the packet based on the topology information;
A packet relay system comprising: a destination selected according to the forwarding probability and a forwarding destination determining means for stochastically selecting a node on the shortest path and determining the forwarding destination node. The packet relay system according to claim 1, wherein the transfer probability calculation means calculates a transfer probability for distributing the packet substantially uniformly to each node in the network. Transfer capacity information storage means for storing transfer capacity information indicating the packet transfer capacity of each node in the network;
The packet relay according to claim 1, wherein the transfer probability calculation means calculates a transfer probability for distributing an amount of packets proportional to the transfer capacity of each node to each node in the network based on the topology information and the transfer capacity information. system. The transfer probability calculation means calculates a transfer probability matrix having an eigenvalue of 1 with the probability that a packet exists at each node on the network as each component of the eigenvector, and uses each element in the calculated transfer probability matrix as a transfer probability. The packet relay system according to any one of claims 1 to 3. The packet relay system according to any one of claims 1 to 4, wherein, when the destination node is an adjacent node, the transfer destination determination unit determines the node as a transfer node. Based on the topology information, calculate the transfer probability, which is the probability of distributing the packet to each node in the network,
Determine the shortest path to the packet destination node based on the topology information,
A packet relay method characterized by probabilistically selecting a node selected according to the transfer probability and a node in the shortest path as a transfer destination node. The packet relay method according to claim 6, wherein a transfer probability that distributes packets substantially uniformly to each node in the network is calculated. 7. The packet relay according to claim 6, wherein a transfer probability for distributing an amount of packets proportional to the packet transfer capacity of the node to each node in the network is calculated based on the packet transfer capacity and topology information of each node in the network. Method. On the computer,
A transfer probability calculation process for calculating a transfer probability, which is a probability of distributing packets to each node in the network, based on the topology information;
A shortest route determination process for determining the shortest route to the destination node of the packet based on the topology information, and
A packet relaying program for executing a forwarding destination determination process for stochastically selecting a node selected according to the forwarding probability and a node on the shortest path as a forwarding destination node. On the computer,
The packet relay program according to claim 9, wherein the transfer probability calculation processing calculates a transfer probability that distributes the packets substantially uniformly to each node in the network.

Images