JP2012200014A - N-tree constructing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To minimize traffic congestion by reducing end-to-end delay, and maximize the bandwidth available in shared N-tree ALM nodes.SOLUTION: Compared to an MST-based method of constructing source-based trees and applying bias towards the shortest delay path, the N-tree constructing method of the present invention makes it possible to construct N source trees at the same time and allocate bandwidth in a fair manner for each stream. That is to say, in the N-tree constructing method of the present invention, each node (MY, SG, JP, VN) constructs N-trees at the same time based on fair bandwidth and allowed delay requirement.

Description

  The present invention relates to an N-tree construction method for performing fair bandwidth sharing in an N-tree.

  Multipoint communication, particularly multipoint communication on the Internet, is widely performed by a multicast method in which the same content is delivered simultaneously to all participating nodes during a session.

  The IP multicast has a problem that an intermediate node configuration, an end node configuration, and an Internet service provider participation requirement are complicated to enable IP layer multicasting. On the other hand, application layer multicasting (ALM) is popular because it is simple because it implements the application layer on an existing unicast network architecture.

  Since ALM does not require a multicast transfer function in the IP layer, AV (audio video) packet transfer by members themselves between participating nodes and AV packet transfer in the application layer can be flexibly performed.

  ALM allows all participants to exchange AV content or streams between participants without any prior configuration during a particular session. Thus, ALM is extremely useful for AV-based applications such as video conferencing between multiple parties.

  The number of AV streams exchanged between participating nodes increases in the order of N × (N−1), where N is the number of members. It is difficult for a communication network to maintain a fair communication cost in terms of bandwidth and waiting time, particularly in the Internet environment between end nodes. This is because the Internet as an open network does not guarantee the fairness of bandwidth between nodes during the exchange of AV streams by participating nodes.

  Since a plurality of ISPs (Internet Services Providers) maintain an Internet path that is inherently dynamic, it is difficult to construct and maintain a minimum stream delivery delay tree for a plurality of trees. This leads to the loss of specific members due to, for example, the underlying bandwidth-independent source base tree (SBT) routing mechanism with the unicast packet routing architecture. It also results in unfair resource distribution, such as bandwidth sharing between participating nodes in an AV session.

  Since packet routing is performed in the application layer for ALM, participating nodes are organized into groups. Packet routing then needs to inform all participating nodes of the packet forwarding mechanism.

  The network layer routing mechanism does not provide a bandwidth fair distribution tree solution because it is a short-sited closest path construction method. Known network layer routing mechanisms include Dijkstra, Bellman-Ford for short path generation, SBT based on the “greedy” concept of the Prim algorithm, and Minimum Spanning Tree (MST). And unicast packet routing on the Internet is based on these algorithms and does not reflect the requirements of ALM routing.

  These solutions are not aware of the end node network bandwidth and the overall network topology, thus causing traffic congestion in the shared path of the nodes. These solutions also result in long network latency and AV quality degradation for ALM applications.

  Hereinafter, the prior art, that is, the Prim-MST (minimum spanning tree) algorithm mechanism will be described using a simple example (FIG. 17). FIG. 17 shows four different locations with bandwidth B (X) capacity of each node's line and delay D (A, B) between the two nodes: Vietnam (VN), Japan (JP), Malaysia (MY), Singapore (SG). In this example, a transmission rate set / slot of 8 Mbps, 4 Mbps, 3 Mbps, or 2 Mbps is used.

  In this case, when the conventional Prime-MST algorithm based on the greedy algorithm principle is executed, each of the 1-N distributed trees for N nodes (N = 4 in this example) is calculated as follows.

(1) First, for MY having the largest allowable bandwidth, a route for transmitting a packet to another node is selected. Specifically, first, it is examined whether MY can secure a bandwidth of 8 Mbps for all other nodes. In this case, if MY selects a route for directly transmitting a packet to VN and SG, a bandwidth of 8 Mbps can be secured. However, since the delay D = 300 ms between MY and JP exceeds the maximum latency of 250 ms, MY cannot select a route for directly transmitting a packet to JP. Therefore, for the packet transmission from MY to JP, a route for transmitting the packet via SG is selected. As a result, MY can secure a bandwidth of 8 Mbps with respect to JP (FIG. 18A). As a result, the remaining bandwidth is 20 Mbps for MY, 12 Mbps for SG, 16 Mbps for JP, and 6 Mbps for VN.

(2) Next, a route for transmitting a packet to another node is selected for the SG having the second largest allowable bandwidth. Specifically, first, it is examined whether the SG can secure a bandwidth of 8 Mbps for all other nodes. However, if the SG secures a bandwidth of 8 Mbps for MY and VN (8 Mbps × 2), the remaining bandwidth of SG exceeds 12 Mbps. Therefore, SG cannot secure a bandwidth of 8 Mbps. Therefore, next, it is examined whether the SG can secure a bandwidth of 4 Mbps for all other nodes. In this case, if SG selects a route for directly transmitting packets to MY and VN, a bandwidth of 4 Mbps can be secured. Further, if a route for transmitting a packet from the SG to the JP via the VN is selected, the SG can secure a bandwidth of 4 Mbps for the JP (FIG. 18B). As a result, the remaining bandwidth is 20 Mbps for MY, 4 Mbps for SG, 16 Mbps for JP, and 2 Mbps for VN.

(3) Next, for the JP having the third largest allowable bandwidth, a route for transmitting a packet to another node is selected. Specifically, first, it is examined whether JP can secure a bandwidth of 4 Mbps for all other nodes. In this case, if JP selects a route for directly transmitting packets to SG and VN, a bandwidth of 4 Mbps can be secured. Further, if a route for transmitting packets from JP to MY via SG is selected, JP can secure a bandwidth of 4 Mbps for MY (FIG. 18C). As a result, the remaining bandwidth is 20 Mbps for MY, 0 Mbps for SG, 8 Mbps for JP, and 2 Mbps for VN.

(4) Finally, the VN with the smallest initial bandwidth selects a path for transmitting packets to other nodes. Here, since the remaining bandwidth of the VN is 2 Mbps, the VN selects a route for directly transmitting a packet to the MY and secures a bandwidth of 2 Mbps. Further, if a route for transmitting packets from MN to JP is selected from VN to JP, VN can secure a bandwidth of 2 Mbps for JP. Further, if a route for transmitting a packet from VN to SG via MY and JP is selected, VN can secure a bandwidth of 2 Mbps for SG (FIG. 18D).

  As described above, the N-to-N video conference session between the participating nodes can be established by constructing N one-to-N source-based trees also by the conventional N-tree algorithm (Prim-MST). This method can satisfy latency requirements based on SBT.

US Patent Application Publication No. 2006/0251062 US Patent Application Publication No. 2006/0153100 US Patent Application Publication No. 2005/0080894 US Patent Application Publication No. 2005/0243722

Min Sik Kim et al., “Optimal Distribution Tree for Internet Streaming Media”, 23rd IEEE ICDCS Conference Proceedings, May 2003

  However, the conventional N-tree algorithm does not satisfy the fairness of the bandwidth between N pairs of N-trees because the N-tree algorithm is calculated individually without harmonizing each tree in the N-tree construction stage.

  In the case of FIG. 18, the conventional N-tree algorithm can only secure a bandwidth of 2 Mbps for VN.

  Further, if an N-tree having a fair bandwidth is to be constructed using a conventional N-tree algorithm, calculation must be performed for all combinations. The calculation amount will be enormous.

  It is an object of the present invention to minimize traffic congestion by reducing end-to-end node delay, further maximizing the available bandwidth on shared N-tree nodes, and fair bandwidth among all ALM nodes. It is an object to provide an N-tree construction method for constructing an N-tree that can share the same.

  The N-tree construction method of the present invention comprises: a. Ordering ALM members in the initial list with reference to metric collector information and group membership information for all nodes; b. Ordering the members of the initial list into an ordered list based on a first empirical rule for every node in every N-tree; c. Selecting from the members of the ordered list those satisfying the link connection status for other nodes in the N-tree based on a second empirical rule, and ordering into a rule-based ordered list for each N-tree And d. Two members having a logical connection between two nodes satisfying a third empirical rule are selected as a pair node from the members of the rule-based ordered list and the members of the initial list, and the rule-base pair Ordering in a list; e. Selecting a node satisfying the fourth empirical rule from among the node pairs in the rule base pair list; f. Forming an N-tree of a given root node selected from the initial list for all receiving nodes in the group based on a fifth empirical rule; g. Ensuring that all links between nodes of a particular N-tree are connected; h. Assuring all nodes in a particular N-tree based on the nodes of the initial list included in the N-tree structure; i. Guaranteeing all N-trees based on the nodes of the initial list included in a plurality of N-tree structures.

  According to the present invention, the end-to-end node delay can be reduced by utilizing a combination of heuristic approach, metric information, and construction and delivery of ALM forwarding table by continuous path information maintenance. The traffic congestion can be minimized and the available bandwidth at the shared N-tree node can be maximized.

The figure which shows the communication system which concerns on one embodiment of this invention The block diagram which shows the internal structure of the client entity and server entity which concern on one embodiment of this invention The figure which shows an example of the optimal stream path | route information table The figure which shows the data structure of the N-tree root node information in OSPIT The figure which shows the data structure of the N-tree member node information in OSPIT Diagram showing bandwidth fairness forwarding table Diagram showing forwarding table (FT) message structure Diagram showing forwarding table entry (FTE) message structure The figure which shows the whole process of the N-tree algorithm by ALM_FRC mechanism Diagram showing FT message exchange sequence The figure which shows the execution method of N-tree algorithm The figure which shows the reception and update method of the bandwidth fairness forwarding table in a client node The figure which shows the detailed process of N-tree algorithm execution The figure which shows the detailed process of N-tree algorithm execution The figure which shows the link rearrangement process of an N-tree algorithm The figure which shows the link movement process of N-tree algorithm Diagram showing packet transfer method by client node The figure which shows the specific example of the N-tree algorithm of this invention The figure which shows the precondition at the time of performing N-tree algorithm concretely The figure which shows the specific example of the conventional N-tree algorithm

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

  The inventor named the N-tree algorithm of the present invention a “water-spread” policy. The N-tree algorithm of the present invention builds N trees for N sources based on a “water diffusion” policy. The N-tree algorithm of the present invention also applies a “water spread” policy to each of the N sources simultaneously to satisfy the bandwidth fairness tree.

  Here, according to the “Water Diffusion” policy, a container containing N water source tunnels (holes) is based on the size of the tunnel and the height of the tunnel (pressure: equivalent to the experience database used when selecting branches) And spread water from the place to different distances. Large tunnels diffuse more water than small tunnels. A tunnel close to the bottom of the vessel will diffuse water farther than a higher tunnel. Each tunnel may have a different capacity (hole size: equivalent to bandwidth). This is interpreted as a node with a large bandwidth carrying more streams to many nodes in the N-tree compared to a node with a narrow bandwidth. As a specific example of the “tunnel height”, in the case of FIG. 17, the height is calculated by bandwidth × first coefficient + reciprocal number of delay × second coefficient, and the like.

The algorithm has three main phases:
i) Connection establishment phase: Connections between all nodes are established, and network metrics (node connection status) are collected.
ii) N-tree construction phase: The status of an N-tree having mesh node connectivity is updated using a connection table called an Optimal Stream Path Information Table. Node join and leave status is updated during the ALM session.
iii) Control or N-tree optimization phase: Perform an improvement procedure to improve the fair bandwidth allocation of the entire N-tree in steps with minimal delay. Then, node movement or link movement is performed between nodes or links for N-tree optimization, resulting in an ALM with FRC (Fast Route Convergence). At this time, the metric information is used to construct an AV stream delivery path for the node. Metric information includes bandwidth between client nodes, latency / delay between client nodes, client node IP address and port number, client node function including route in ALM session, child-parent-grandparent relationship, used and This includes the number of unused upload links and the number of used and unused download links.

  When communication is performed between ABC nodes, the connection relationship between the nodes ABC can be calculated from these metric information. For example, when there is connectivity between A and B and there is no connectivity between A and C, the measured metric information indicates that the bandwidth between A and C is 0, and the latency between R and A (RTT: Round Trip) Time) is infinite.

  Note that the metric information in the present application is not indicated by a discrete amount (0 or 1) indicating the presence or absence of connection, but is a measurement value of a full mesh (bidirectional of AB, AC, BC). Shown in reflected continuous quantity. That is, the metric between the nodes A and C may be 1 Mbps, 0.1 Mbps, or 0 Mbps.

  The N-tree algorithm is an AV conference application in which a small group of participants exchange AV streams and the like in order to have a high level of interactive function and good AV quality through a session between all participants on the network. Is preferred.

  In order to guarantee bandwidth fairness to all nodes, the N-tree algorithm of the present invention considers the capacity of the upload slot and download slot of each node. The upload slot and the download slot refer to the bandwidth capacity of the uplink and downlink of the node. The bandwidth is divided into a plurality of discrete ranges, for example, 64 kb and 128 kb.

  The “Water Spread” policy is adopted along with the three main phases of the above N-tree construction algorithm for bandwidth fairness distribution tree construction. This solution implements ALM_FRC that applies to all nodes participating in the network for a given AV session. The N-tree algorithm may be executed in a centralized or distributed manner between nodes.

  FIG. 1 is a diagram showing a communication system according to an embodiment of the present invention. As shown in FIG. 1, in this embodiment, a plurality of client nodes (104, 106, 108, 110) and a server node (102) are each connected to a network (112). A multimedia application such as an AV conference application existing in each client can simultaneously exchange AV streams between a plurality of client nodes during an ALM session. The server node (102) can communicate with all other client nodes and simultaneously exchange network information between them.

  In the present invention, nodes and entities, which are terms indicating physical and logical network entities, can be used interchangeably. Both client nodes or server nodes and client entities or server entities may be replaced by the general meaning of network communication. The server node and the client entity are both terminal devices in the wireless communication system.

  The server node (102) collects metric information from each client node at a predetermined timing (dynamically) or periodically and executes an N-tree algorithm (water diffusion policy). Next, the server node (102) stores the bandwidth fairness ALM path information, which is the result of the N-tree algorithm, in the forwarding table. Further, the server node (102) is responsible for dynamically transmitting the bandwidth fairness N-tree routing information generated from the bandwidth fairness ALM path information stored in the forwarding table to all participating client nodes. . Note that the predetermined timing includes, for example, a timing when the stream from the top is interrupted.

  The bandwidth fairness N-tree routing information received by the client node is used at each client node to forward the received ALM packet to the next client node until the final destination is reached.

  FIG. 2 is a block diagram showing an internal configuration of the client entity (256) and the server entity (210) according to the embodiment of the present invention.

  An ALM session is defined as a logical session and is connected over a network in multiple collaborative applications that simultaneously exchange data packets between two or more client entities (256). ALM sessions require a server entity (210) for session coordination between participating members.

  The client entity (256) is an ALM function at a node that participates in an ALM session for network packet streaming between all nodes. The server entity (210) is an ALM function that is responsible for collecting metric information and executing an N-tree algorithm that results in an ALM_FRC distribution tree. Both the client entity (256) and the server entity (210) are logical entities that may exist in any node.

  Only one server entity is required for a centralized N-tree session. In the distributed case, each node has one server entity and one client entity. Hereinafter, a plurality of client entity functions in an ALM session and a case of one server entity function (centralized N-tree session technique) will be described.

  In FIG. 2, the server entity (210) includes a metric collector (220), a group membership (222), an ALM forwarding table constructor (218), and an ALM forwarding table distributor (ALM Forwarding Table Distributor). (278). The metric collector (220), group membership (222) and ALM forwarding table constructor (218) are connected by a link (208) to exchange metric information therebetween.

  The server entity (210) metric collector (220) and group membership (222) are responsible for dynamically collecting metric information from all client nodes (104, 106, 108, 110) in the ALM session. Metric information includes bandwidth between client nodes, latency / delay between client nodes, client node IP address and port number, client node function including route in ALM session, child-parent-grandparent relationship, used and unused This includes the number of upload links and the number of used and unused download links. These metric information is stored in a metric database (226) and a membership database (230) that can be accessed via links (224, 228), respectively. 4 and 5 illustrate the information / content data structure in detail. The membership database (230) stores the IP address of each client node.

  The root node refers to the source node of the N-tree from which the network packet originated. If all of the nodes transmit network packets for N members in the N-tree, there are N root nodes. Child-parent-grandparent refers to the relationship between nodes in the N-tree. The child node is a node that directly receives a network packet from the parent node via the download slot of the parent node. The parent node transmits network packets directly to the child node via the child node's upload slot. On the other hand, the grandparent node is a node connected to a given child node via the parent node, and the network packet flows from the grandparent node to the parent node before reaching the child node. A grandchild node is a direct relationship between a grandparent node and a child node. The node upload slot and download slot refer to the output capacity and input capacity of the node. A link slot refers to the bandwidth capacity of the link between two nodes associated with an upload slot and a download slot for a particular ALM session.

  The N-tree constructed for ALM_FRC shall be subject to upload / download slot constraints, link slot constraints, and latency constraints for a given ALM session.

  The optimum stream path information table (204) stores N-tree session information (information on the state of the tree) for all nodes in the connected state shown in FIGS. The optimum stream path information table (204) includes N-tree connection information. On the other hand, the Heuristic Rules Database (206) stores a set of empirical rules that are used to build a bandwidth fair N-tree. The basic rules and augmentation rules that exist in the empirical rule database (206) are:

  The basic rule of the N-tree algorithm is to build N trees with one tree per node for network packet delivery. This means that each node must receive N-1 streams with N-1 download slots. Each tree adds N-1 upload links to all links used, so for N trees there are at least a total of N x (N-1) upload links from all participants. To do. In addition, each tree requires several additional slots and links for control message data transmission.

  Additional empirical rules include upload / download slot constraints, link slot constraints, latency constraints for a given session. It is NP-hard to construct an optimal N-tree that satisfies all the above three constraints. It is therefore permissible to violate some of the above constraints in order to build the best and optimal N-tree. For example, violations of latency constraints cause delays in receiving the stream, while violations of upload constraints or link constraints invalidate part of the stream delivery. The empirical rule satisfies upload or download constraints and link constraints first, and then satisfies link latency constraints. The N-tree is constructed with an acceptable latency by applying the above rules.

Also, the empirical rule database (206) stores a plurality of empirical rules applicable for N-tree construction as follows.
Rule 1: A node should first use its bandwidth to upload its own stream, rather than carrying another node's stream. This reduces the waiting time of the tree.
Rule 2: If a selection is made between nodes, the node with the largest remaining slot capacity is selected.
Rule 3: If a selection is made between links, the link with the largest remaining slot capacity is selected.
Rule 4: A link that gives the least end-to-end latency to the tree is selected between links having the same remaining slot capacity in Rule 3.

  As shown in FIG. 6, the server forwarding table (202) stores bandwidth fairness ALM path information which is packet forwarding information for all nodes in a specific N-tree session.

  The ALM forwarding table constructor (218) accesses the metric information collected by the metric collector (220) and group membership (222). The ALM forwarding table constructor (218) applies the empirical rules stored in the empirical rule database (206) to move the N-tree session information stored in the optimum stream path information table (204). And update continuously and continuously. The ALM transfer table constructor (218) is a main module in the server entity (210).

  Next, the ALM forwarding table constructor (218) executes the N-tree algorithm (water diffusion policy), and transmits the bandwidth fairness ALM path information, which is the result of the N-tree algorithm, to the server via the link (282). Store in the transfer table (202). The bandwidth fairness ALM route information is in the form of a forwarding table.

  The ALM forwarding table constructor (218) is connected via the links (216, 214, 208) through the empirical rule database (206), the optimum stream path information table (204), the server forwarding table (202), the metric collector (220), Access group membership (222) and empirical rules database (206), respectively. The ALM forwarding table constructor (218) performs "Initial List", "Ordered List", "Rule-Based Ordered List" in the List Database (260) during execution of the N-tree algorithm. “Rule Based Ordered List” and “Rule Based Pair List” are used for bandwidth fairness N-tree construction.

  The “initial list” contains information about all members in the ALM session that is collected by the group membership (222) and metric collector (220). An “ordered list” provides a sorted list of members based on specific rules, eg, upload and / or download slot capabilities. A “rule-based ordered list” provides an ordered list of members based on specific rules, eg, link latency between two nodes. A “rule-based pair list” provides a list of members that form a link between two nodes based on specific rules, eg, selected empirical rules. Use of these lists is illustrated in FIGS. 13 and 14 as part of the N-tree algorithm description.

  The ALM forwarding table distributor (278) inputs the bandwidth fairness ALM path information stored in the server forwarding table (202) via the link (212), and from this information, each client entity (256 in the N-tree). ) For bandwidth fairness N-tree routing information. The bandwidth fairness N-tree routing information includes a pair of an input port number, a destination address, and a destination port number. The ALM forwarding table distributor (278) then transmits bandwidth fairness N-tree routing information to each client entity (256) via the network link (234).

  The client entity (256), which is two or more nodes in the ALM network, is connected to the server entity (210) via the network links (234, 276) and can communicate dynamically. The client entity (256) has four main modules: ALM Forwarding Table Receiver (280), Metric Provider (244), Group Membership (246), ALM Forwarding Table Executor. (240).

  The ALM forwarding table receiver (280) receives bandwidth fair N-tree routing information from the ALM forwarding table distributor (278) via the network link (234) and client forwarding for packet routing via the link (242). It stores in a table (Client Forwarding Table) (236).

  Metric provider (244) and group membership (246) are responsible for collecting data dynamically from all client nodes (104, 106, 108, 110) in the ALM session. Metric information includes bandwidth between client nodes, latency / delay between client nodes, client node IP address and port number, client node function within ALM session including route, child-parent-grandparent relationship, used and not used Includes the number of used upload links, the number of used and unused download links, etc.

  These metric information is stored in both the metric database (250) and the membership database (254), which are accessible via links (248, 252), respectively.

  Metric information is dynamically and periodically delivered to the metric collector (220) and group membership (222) of the server entity (210) by the metric provider (244) and group membership (246), respectively. The The information delivery is appropriately performed via network links (232, 258).

  The metric information is then updated to the server entity (210) metric database (226) and membership database (230). ALM forwarding table executor (240) accesses metric provider (244) and group membership (246) via link (238) for local metric data exchange. Link (248) connects metric provider (244) and metric database (250), and link (252) connects group membership (246) and membership database (254).

  The ALM forwarding table receiver (280) is responsible for dynamically and periodically receiving bandwidth fairness routing information from the server entity (210) and updating the received content to the local client forwarding table (236). . In addition, the ALM forwarding table receiver (280) can independently request each routing information from the ALM forwarding table distributor (278) via the link (276).

  Bandwidth fairness N-tree routing information in the client forwarding table (236) is used in the ALM forwarding table executor (240) to process ALM packets received from other client entities (106, 108, 110). Used dynamically. Thereafter, the bandwidth fair N-tree routing information in the client forwarding table (236) is forwarded to the next destination identified by the routing engine (262). The ALM transfer table executor (240) retrieves the next destination information from the client transfer table (236) via the link (274) and provides it to the routing engine (262) via the link (272). The routing engine (262) performs a search function for every received packet (264) in the application layer for each network packet in the ALM in each client entity. The received packet (264) is searched in the routing engine (262) based on the source address (606), the source port number (654), and the transport layer protocol, and then modified, and designated as a transmission packet (266). Forwarded to the node. FIG. 6 and the associated description explain this procedure in detail.

  Communication between the ALM forwarding table distributor (278) and the ALM forwarding table receiver (280) via the links (276, 234) is performed using a specific message structure. The transfer table content (808) including the bandwidth fairness transfer table information shown in FIG. 8 is included in the transfer table entry (FTE) information shown in FIG. One or more forwarding table entry (FTE) information is combined and sent as a forwarding table (FT) message from the server entity (210) to each client entity (256) via link 234.

  Next, the structure of the optimum stream path information table (204) will be described in detail with reference to FIG.

  In FIG. 3, the root (302) entry in row (326) indicates the number of root nodes in the N-tree. The root node is the source node from which the ALM application packet is transmitted in the N-tree. These packets are addressed to all ALM recipients in the ALM session. The route (302) may include a plurality of route nodes indicated by reference numerals (304, 306, 308, 310, 312). Rows (328, 330, 332, 334, 336) are entries that store participating node information for a given N-tree rooted at a particular root node (304, 306, 308, 310, 312). .

  The node (314) entry in column (348) stores the identifier of each node (316, 318, 320, 322, 324) in the N-tree session. Thus, if there are N nodes in a particular N-tree, there are N entries in row 314. Similarly, if all N nodes are root nodes that generate source packets, row (302) has N columns in the table.

  The table entries with reference numerals (338, 340, 342, 344, 346) store route node related information as illustrated in FIG. On the other hand, all other table entries except reference code (338, 340, 342, 344, 346) entries store node specific information for a specific N-tree as illustrated in FIG. The optimal stream path information table (204) may be a static or dynamic table that can be expanded or contracted to accommodate the expansion of participating nodes within an N-tree session.

  FIG. 4 shows the N-tree root node information specified in the optimal stream path information table (204), metric database (226, 250), and membership database (230, 254) for all trees in the ALM session. The data structure of is shown. Field (402) specifies the number of used input link slots and download slots for a particular node for the entire N-tree session.

  On the other hand, field (404) specifies the number of unused input link slots and download slots for a particular node for the entire N-tree session. Similarly, field (406) specifies the number of used output link slots and upload slots for a particular node for the entire N-tree session. On the other hand, the field (408) specifies the number of unused output link slots and upload slots for a particular node. Field (410) is reserved.

  FIG. 5 shows the data structure of N-tree member node information for nodes other than the root node of a particular tree within an ALM session. A field (502) specifies the origination node of the stream based on each root tree. Field (504) represents a discrete form of delay or latency between the route and each node. Field (506) represents a child or parent relationship between two different nodes and indicates that the function exists. This is followed by a port number information for connection and a field (508) representing next node information or identifier such as an IP address for packet transfer. This field also indicates that there is a next node.

  That is, this field is a leaf node for packet transfer (a terminal node that does not transfer). Similarly, the following fields (510, 512, 514, 516) represent grandparents (before the second generation) -child and great grandparents (before the third generation) -child relationship (after the first generation). Although only three levels are explicitly shown in the present embodiment, the present invention is not limited to this, and more levels may be defined for further utilization. Finally, field (518) represents a reserved field.

  FIG. 6 shows a general view of the bandwidth fairness forwarding table (602) of both the server forwarding table (202) and the client forwarding table (236) used in the server entity (210) and the client entity (256). Indicates the format. The first level of the table (602) includes the following three columns: source node / source address (606), source port number (654), and tri-table pointer (604). The source node (source address) (606) stores the IP addresses of the nodes (608, 610, 612, 614). The source port number (654) stores the source port number (654) of all outgoing nodes (646, 648, 650, 652). The tri table pointer (604) stores pointers (616, 618, 620, 622) for the transfer table (624) of the next level.

  Multiple entries of the originating node (606) can coexist with their respective next level forwarding table (624). The next level transfer table includes a destination address (626) and destination port number information (636). The next level forwarding table (624) specifies multiple IP addresses (628, 630, 632, 634) and port numbers (638, 640) that specify where in the node the received packet needs to be forwarded next. 642, 644). Both tables (602, 624) also have transport protocol fields (656, 658) that indicate the type of transport protocol for the particular stream.

  The forwarding table information is exchanged between the server entity (210) and the client entity (256) by using the forwarding table message (FT) shown in FIG. The search for network packet forwarding in the N-tree at each node is performed by using an IP address, a port number, and a transport protocol (eg, UDP or TCP). The routing engine (262) performs a search function on every input packet (264) in the application layer for each network packet in the ALM in each client entity.

  The forwarding table (602) is used in the client entity to perform a search for a destination address (626) or the like for each input packet, and then forwarded to a specified next node in the N-tree. . The routing engine (262) also changes the source IP address and port number of the network packet during the forwarding phase before forwarding the outgoing packet (266). This function is important to ensure adherence to the streaming path constructed by the N-tree algorithm.

  In a normal unicast application, the source IP address and port number of the network packet are not changed. Packet forwarding is then performed along the overlay tree without fair bandwidth allocation. However, one of the features of the present invention is that the source IP address and destination port are changed by the forwarding node to follow the bandwidth fair N-tree path to realize ALM_FRC.

  FIG. 7 shows a forwarding table (FT) message structure with a total of six fields. The type (702) field indicates a request message type or a response message type. The content (704) field specifies the content indicating whether it is a full forwarding table entry (FTE) update or a partial forwarding table entry (FTE) update from the server entity. The version (706) field specifies the version of the ALM and N-tree algorithm. The unused (708) field indicates that the field is not currently used. The FTE (710) field contains single or multiple forwarding table entry (FTE) messages. The FT message ends with the reserved (712) field.

  FIG. 8 shows the structure of a forwarding table entry (FTE) message 710 having five fields. The source node IP field (802) indicates the IP address of the node that transmitted the FT message. The timestamp field (804) indicates the time when the message was generated. An unused field (806) and a reserved field (810) (those to be used in the future) indicate that they are not currently used. The timestamp field (804) is used to specify the time when the FT message was constructed.

  The timestamp field (804) ensures that only the latest message is used at the client entity to update the routing table of the N-tree. If the transport protocol has its own timestamp field, the timestamp field (804) is not used. However, if the transport protocol does not have a time stamp field, a time stamp field (804) is required.

  Subsequent forwarding table content field (808) contains bandwidth fairness forwarding table information from the server entity to all client entities. The forwarding table entry information selected from the server entity (210) is combined to form an FT message that is sent to each client entity (256). FT messages containing single or multiple FTE messages are dynamically sent by AFTC to affected client entities when there is an update in the server forwarding table (202).

  Next, an overall mechanism for realizing ALM_FRC by executing fair bandwidth sharing in the N-tree will be described with reference to FIG.

  First, the metric provider (244) and the group membership (246) of the client entity (256) perform regular delivery of metric information (ST902). The periodically transmitted metric information is collected by the metric collector (220) and the group membership (222), respectively (ST904).

  Next, when new metric information arrives, ALM forwarding table constructor (218) accesses metric database (226) via metric collector (220) and membership database (230) via group membership (222). ). The new metric information is used by the ALM forwarding table constructor (218) to update the optimum stream path information table (204) (ST906).

  Next, the ALM forwarding table constructor (218) executes an N-tree algorithm (ST908). The ALM forwarding table constructor (218) updates the optimum stream path information table (204) and the server forwarding table (202) by periodically executing this N-tree algorithm (ST910). The execution steps of the N-tree algorithm are illustrated in FIG. 11 and are described in the following section.

  The updated server forwarding table (202) content is then transferred from the ALM forwarding table distributor (278) of the server entity (210) to each client entity (256) using a forwarding table (FT) message. It is transmitted (ST912). The ALM transfer table receiver (280) updates the client transfer table (236) with the received transfer table (FT) message (ST914).

  Next, the exchange sequence of the transfer table (FT) message will be described in detail with reference to FIG.

  Periodic metric information update (PIT message) is performed (ST1014). That is, the update of the metric information is continued between the metric provider (1012) of the client entity (1008) and the metric collector (1006) of the server entity (1002).

  The ALM forwarding table distributor (1004) of the server entity (1002) forwards the forwarding table (FT) message including the forwarding table contents to the ALM forwarding table receiver (1010) of the client entity (1008) (ST1016). Alternatively, the ALM forwarding table receiver (1010) can make an update request for the client forwarding table (236) using the forwarding table (FT) message (ST1018). In response to the request, the ALM transfer table distributor (1004) transfers a transfer table (FT) message including transfer table contents to the ALM transfer table receiver (1010) (ST1020).

  Next, the execution method of the ST-908 N-tree algorithm shown in FIG. 9 will be described with reference to FIG. The initial trigger for calling the N-tree algorithm is performed by the ALM forwarding table constructor (218) (ST1102). Next, access to a specific rule in the empirical rule database (206) is performed by the ALM forwarding table constructor (218) (ST1104). Subsequently, the optimum stream path information table (204) is accessed by the ALM transfer table constructor (218) (ST1106). Thereafter, the ALM_FRC guarantee by the N-tree algorithm is applied to the metric (ST1108). Then, the server transfer table (202) is updated by the ALM transfer table constructor (218) using the generated operated metric (ST1110).

  Next, a method of periodic transfer table information reception and packet transfer by the client node will be described with reference to FIG. The ALM transfer table distributor (278) periodically transmits the contents of the selected server transfer table (202) to a specific client via a transfer table (FT) message (ST1202). The ALM forwarding table receiver (280) of the client entity (256) receives these forwarding table (FT) messages and updates the client forwarding table (236) (ST1204).

  Next, detailed steps of executing the N-tree algorithm by applying the “water diffusion” policy will be specifically described with reference to FIG. The whole process is simply that in the ALM session, from every source node (root node) to every node in the N-tree, all the source nodes are connected to each other to form N trees. It can be expressed by diffusion. For example, an ALM session {a,. . , N} is encoded with a specific color (color is for uniquely identifying the source node), and all links belonging to the tree with root node j have color j. Initially, each node j includes a color set S [j] = {j} indicating a stream connection or reachability status. When stream k is “flowing” and reaches node j, node j adds color k to its color set S: S [j] = S [j] + {k}.

  Initially, for each color k, DONE {a, (k = FALSE),. . , N} is set to FALSE. When stream k reaches all other nodes in the ALM session (a tree rooted at k is completed), (DONE {a, (k = True), ..., n} is set to True Add k to the DONE list, so that all members in the DONE {a, ..., n} set are set to True until all N trees are complete. By using this process, one link is added to the tree repeatedly in one loop.

  First, the ALM forwarding table constructor (218) identifies all members in the N-tree session based on the metric collector (220) information and group membership (222) information, and stores " An “initial list” is created (ST1302). Next, the ALM forwarding table constructor (218) selects members from the "initial list" and, for all nodes in the "initial list" of all N-tree nodes, from the empirical rule database (206). Based on the above rule, the member is added to the “ordered list” in the list database (260) (ST1304). The rules from the empirical rule database (206) are, for example, descending numbers of the remaining unused upload slots.

  The ALM forwarding table constructor (218) conforms to the “water diffusion” policy example, and in ST1304, DONE {a, (k = False),. . , N} in its own set S, select the node with the highest number among the remaining slots still having color k. These nodes are added in the “ordered list”. Of the remaining slots, the next highest number that repeatedly satisfies the HRDB rule of the upload / download slot constraint is iteratively selected.

  Next, the ALM forwarding table constructor (218) did not establish links to all nodes of a given N-tree by satisfying specific rules from the empirical rule database (206). Select a node from the "ordered list" starting with. Next, the ALM forwarding table constructor (218) adds the node (for example, the highest numbered node of unused slots) to the “rule-based ordered list” (ST1306). The above process is continued for all nodes that match the rule in ST1306 until all nodes are ordered (ST1308).

  Next, the ALM forwarding table constructor (218) selects an ordered node from the “rule-based ordered list” in order to form a node pair (ST1310). Then another node from the “initial list” that satisfies a particular rule from the empirical rule database (206) (eg, the minimum latency between the two nodes) is selected, and the pair node is identified as “rule-based pair It is added to “list” (ST1312).

  Next, the ALM forwarding table constructor (218) selects a pair node from the “rule-based pair list”, and orders the pair nodes based on the specific rule from the empirical rule database (206) (ST1314). In the specific rule, for example, a pair is matched based on link information of the highest number in the order list among unused upload slots. By these steps (ST1306 to ST1314), link selection for N-tree formation is performed.

  The ALM forwarding table constructor (218) conforms to the example of “water diffusion”, and in these steps (ST1306 to ST1314), a legal tunnel (route that logically connects nodes) in which water can diffuse. Select. Legitimate tunnels (p1, p2) are tunnels with several streams that have already flowed to their starting point p1, but do not yet exist in p2. Large tunnels or high numbered slots are preferred.

  In the next step, for all links (p, q) having member p in the “ordered list” of the previous step, all links (i, j) satisfying the latency constraint and the set S [I], S [j] ≠ [radix] is selected. Next, the ALM forwarding table constructor (218) adds them to the “rule-based ordered list” (ST1306). In the next step, the ALM forwarding table constructor (218) selects the highest slots from the "rule-based ordered list" and adds them to the "rule-based pair list" in the list database (260). (ST1312).

  Next, the ALM forwarding table constructor (218) selects members from the "rule-based pair list" that satisfy additional rules from the empirical rule database (206) (eg, the lowest latency between pairs). (ST1316). Next, the ALM forwarding table constructor (218) forms an N-tree connecting the nodes from the “initial list” by adopting specific rules from the empirical rule database (206) (ST1318). The specific rule, for example, selects a plurality of node pairs that form a connection from one node of the N-tree to all nodes in the “initial list”, except itself, without violating the maximum latency rule. The rule is adopted. This process is {a, i ,, n} {a ,, j, n} {a so that the link status is set to True for all nodes (the complete connection state of the links in the tree). ,,, K}. . Continue for all members from the "rule-based pair list" by matching all nodes, n, members of set S to (ST1320).

  Consistent with the “water diffusion” example, in this step, the ALM forwarding table constructor (218) selects the shortest tunnel if there are many legal tunnels to select. That is, in this process, the ALM forwarding table constructor (218) has the least latency between the node pair (r, v) for all links (i, j) in the "rule-based pair list". Select the link (r, v).

  The ALM forwarding table constructor (218) sets the point from the source k when there are many streams that have already reached the start point r and can be spread to the end point v after selection of the tunnel (r, v). Choose a stream that gives an acceptable distance to v. The ALM forwarding table constructor (218) adds the link (r, v) to the node r tree when the node v has not yet received a stream having the node r as a root. Otherwise, the ALM forwarding table constructor (218) has a waiting time (k, v) of maximum waiting time for one random color k linking nodes from the sets S [r] and S [v]. In the following case, a link (r, v) is added to a tree having k as a root. The ALM forwarding table constructor (218) selects the other link with the next largest capacity left if there are no selectable links (r, v) (eg due to latency violation). Try.

  The ALM forwarding table constructor (218) repeats this process until a complete N trees are built for all nodes in the "rule-based ordered list" (ST1322). All nodes in one N-tree are for N sources so that a complete N tree can reach any other node in the other N-tree from any one node. Defined as connected to other nodes in N trees. This may be accomplished by fully satisfying all defined rules from the empirical rule database (206) or by fully satisfying partial rules from the empirical rule database (206).

  Finally, N-tree optimization is performed as shown in FIG. 13B (ST1324 to ST1330). First, N-tree link movement is performed (ST1324). This process attempts to move links between paired nodes to minimize the height of N trees, as shown in FIG. 14A. When N trees are changed after moving the link, N trees need to be reconstructed by the following steps starting from step 1312 (ST1326). When reconstruction of N trees fails due to link movement, link rearrangement is performed (ST1328). The link reordering process (ST1328) is continued (ST1330) until complete N trees are constructed.

  The update information based on the N-tree algorithm is reflected in the optimum stream path information table (204) and the server forwarding table (202) for delivery between client entities (ST1332).

  Next, two N-tree optimization techniques, link movement, and link rerouting will be described with reference to FIGS. 14A and 14B.

  First, the link rearrangement process of the N-tree algorithm will be described in detail with reference to FIG. 14A. First, when the ALM forwarding table constructor (218) receives a trigger for performing N-tree rearrangement (ST1402), it selects a member from the optimum stream path information table (204) (ST1404). If there is a link with unused bandwidth between two nodes in the optimal stream path information table (204), the ALM forwarding table constructor (218) determines that the new link delay constraint and bandwidth are With respect to the rules, the link is moved to another node so that it is still valid (ST1406). The relationship between the two nodes is not related to the parent-child relationship, grandparent-child relationship, etc. in the optimum stream route information table (204). This enables N-tree link rearrangement for all nodes in the optimum stream path information table (204) (ST1408).

  Next, the link movement process of the N-tree algorithm will be described in detail with reference to FIG. 14B. First, when receiving a trigger for link movement (ST1410), the ALM forwarding table constructor (218) selects a member from the optimum stream path information table (204) for link movement (ST1412). The ALM forwarding table constructor (218) has two or more logical links regardless of the parent-child relationship, grandparent-child relationship, etc. between two identical nodes in the optimal stream path information table (204), The first link and the second link are compared so that the delay constraint and the bandwidth between the two nodes on the link are the same. Furthermore, if there are more intermediate nodes, the ALM forwarding table constructor (218) replaces the former link with the latter link so as to have a shorter N-tree (ST1414). As a result, the height of the N-tree can be minimized for all the nodes in the optimum stream path information table (204) (ST1416).

  In exceptional cases, after trying for every candidate node during N-tree construction, both link movement and reordering for optimization may fail. In such a scenario, the N-tree computation is incomplete and the N-tree algorithm fails to complete. However, N-tree simulation results show that the N-tree algorithm succeeds in 99% of the given cases when the tree is calculated using link movement, link reordering techniques, and empirical rule database information. Proved. On the other hand, without link movement, link reordering technology, and empirical rule database information, the success rate of N-tree construction was 95% for a given case. The N-tree computation time is less than 50 milliseconds for 12 nodes in a normal personal computer specification.

  N-tree height minimization / link movement can be described as follows, based on a given “water diffusion” example. In a tree having an arbitrary node j as a root, the link from the parent node of node k to node k and the replacement from the link of node k to node p do not violate the waiting time of tree j. In this case, the second parent of the node k can be defined as a node p that can replace the existing direct parent of the node k. Thus, a node could have N-1 second parent sets (corresponding to N-1 trees rooted at other nodes). This is done by changing {a ,, k = true ,, n} to {a ,, k = false, n} to link “k” (link between paired nodes) from the “rule-based ordered list”. And select the link “k ′” from the “rule-based ordered list”. In this case, k 'is selected because it is the second parent.

  Next, a packet transfer method by a client node will be described with reference to FIG. First, the client node receives a source packet from any other node (ST1502). The client node then identifies two destinations (IP address and port pair) or three tuples (IP address, port, or ALM) as previously described to identify the destination IP address and port number. The client forwarding table (236) is searched based on a flag indicating a packet or a set of classifications (ST1504). The flag or classification indicating that the packet is an ALM packet may be either a DSCP (Differentiated Services Code Point) area of the IPv6 header, a flow label, or both of the IPv4 ToS (Type of Service) field. A combination may be used. Alternatively, the flag or section indicating that the packet is an ALM packet may be a specific area defined in the UDP payload area.

  Next, the client node changes the source IP address and port number of the ALM packet to the current node IP address and port number (ST1506). This ensures that all network packets follow an N-tree constructed path resulting in bandwidth fairness between nodes. Finally, the client node transfers the network packet to each next destination node (ST1508).

  Next, a specific example of the N-tree algorithm of the present invention will be described with reference to FIG. FIG. 16 illustrates the step-by-step application of the N-tree algorithm for the same network conditions as FIG. 17 and the resulting network link. The tree of each node [NY, SG, JP, VN] base is from the largest to the least according to some defined empirical rules, with a defined latency and bandwidth rate set / slot. Search for the best network link. The bandwidth rate set / slot is, for example, 8 Mbps, 4 Mbps, 3 Mbps, or 2 Mbps.

  The empirical rules in this case are rule 1 to rule 4 above. Thereby, as shown in FIG. 16, the route is selected in order from the loop 1.

  Here, if one path construction is not successful, backtracking is performed in order to start processing for maintaining a fair bandwidth for all nodes in addition to complying with the delay constraint.

  In the case of FIG. 16, when the N-tree algorithm of the present invention is executed, each of the paired N distributed trees for N nodes (N = 4 in this example) is calculated as follows.

(1) First, a route for directly transmitting a packet from MY having the largest allowable bandwidth to a VN having a bandwidth of 8 Mbps and the smallest delay with MY is selected (loop 1), and the delay with MY is the second. A route for directly transmitting a packet to a few SGs is selected (loop 2) (FIG. 16A). However, since the delay D = 300 ms between MY and JP exceeds the maximum latency of 250 ms, a route for directly transmitting a packet from MY to JP cannot be selected. As a result, the remaining bandwidth is 20 Mbps for MY, 12 Mbps for SG, 16 Mbps for JP, and 6 Mbps for VN.

(2) Next, a route for directly transmitting a packet from the SG with the second largest remaining bandwidth to the VN having the smallest delay with the SG with a bandwidth of 4 Mbps is selected (loop 3). A route for directly transmitting a packet to the JP with the second smallest delay is selected (loop 4). At this point, since the remaining bandwidth of JP is larger than that of JP, a route for directly transmitting a packet from JP to a VN with a bandwidth of 4 Mbps and the smallest delay with JP is selected (loop) 5) (FIG. 16B). As a result, the remaining bandwidth is 20 Mbps for MY, 12 Mbps for SG, 12 Mbps for JP, and 6 Mbps for VN.

(3) Next, a route for directly transmitting a packet to the SG with a bandwidth of 4 Mbps is selected from the JP having the second largest bandwidth (loop 6) (FIG. 16C). As a result, the remaining bandwidth is 20 Mbps for MY, 12 Mbps for SG, 8 Mbps for JP, and 6 Mbps for VN.

(4) Next, a route for directly transmitting a packet to MY with a bandwidth of 4 Mbps is selected from the SG with the second largest remaining bandwidth (loop 7) (FIG. 16D). As a result, the remaining bandwidth is 20 Mbps for MY, 8 Mbps for SG, 8 Mbps for JP, and 6 Mbps for VN.

(5) Next, from MY having the largest remaining bandwidth, a route for transmitting packets to the JP via the SG with a bandwidth of 4 Mbps is selected (loop 8) (FIG. 16E). As a result, the remaining bandwidth is 20 Mbps for MY, 4 Mbps for SG, 8 Mbps for JP, and 6 Mbps for VN.

(6) Next, the route from which the remaining bandwidth is the second from JP to 4 Mbps and the packet is transmitted to the MY via the SG is selected (loop 9) (FIG. 16F) . As a result, the remaining bandwidth is 20 Mbps for MY, 0 Mbps for SG, 8 Mbps for JP, and 6 Mbps for VN.

(7) Next, a route for directly transmitting a packet from the VN to the JP with a bandwidth of 4 Mbps is selected (loop 10), and further, from the VN to the SG via the JP with a bandwidth of 4 Mbps. A route for transmitting a packet is selected (loop 11) (FIG. 16G). As a result, the remaining bandwidth is 20 Mbps for MY, 0 Mbps for SG, 4 Mbps for JP, and 2 Mbps for VN.

  Here, since the remaining bandwidth of SG is 0 Mbps and the remaining bandwidth of VN is 2 Mbps, it is not possible to find a route for reaching MY from VN with a bandwidth of 4 Mbps.

  Therefore, the N-tree algorithm of the present invention moves to the backtracking process and reviews the route. Specifically, MY is selected, the link (SG-VN) is deleted, and two links, that is, (SG-MY) and (MY-VN) paths in the SG tree are selected instead (loop) 12) (FIG. 16H). As a result, the remaining bandwidth is 16 Mbps for MY, 4 Mbps for SG, 4 Mbps for JP, and 2 Mbps for VN.

  As a result, it becomes possible to select a route for transmitting packets to MY from VN via JP and SG with a bandwidth of 4 Mbps (loop 13) (FIG. 16 (I)). As a result, the remaining bandwidth is 16 Mbps for MY, 0 Mbps for SG, 4 Mbps for JP, and 2 Mbps for VN.

  Thus, the N-tree algorithm of the present invention builds N source trees at the same time as compared to an MST-based method that builds a source-based tree and biases it towards the shortest delay path. Bandwidth can be allocated fairly for each stream.

  For example, the conventional method shown in FIG. 18 can only secure a bandwidth of VMbps of 2 Mbps, whereas the N-tree construction method of the present invention shown in FIG. 16 can secure a bandwidth of 4 Mbps at all nodes. it can. That is, each node (MY, SG, JP, VN) can simultaneously construct an N-tree based on fair bandwidth and allowed delay constraints.

  The present invention is suitable for use in AV conferencing applications in which small group participants exchange AV streams or content with a high level of interaction and good AV quality through a session between all participants on the network.

102 server 104, 106, 108, 110 client 112 network 202 server forwarding table 204 optimal stream path information table 206 empirical rule database 210 server entity 218 ALM forwarding table constructor 220 metric collector 222 group membership 226 metric database 230 membership database 236 Client forwarding table 240 ALM forwarding table executor 244 Metric provider 246 Group membership 250 Metric database 254 Membership database 256 Client entity 260 List database 262 Routing engine 278 ALM forwarding table distributor 280 ALM transfer table receiver

Claims (4)

a. Ordering ALM members in the initial list with reference to metric collector information and group membership information for all nodes;
b. Ordering the members of the initial list into an ordered list based on a first empirical rule for all nodes in all N-trees;
c. Selecting from the members of the ordered list those satisfying the link connection status for other nodes in the N-tree based on a second empirical rule, and ordering into a rule-based ordered list for each N-tree When,
d. Two members having a logical connection between two nodes satisfying a third empirical rule are selected as a pair node from the members of the rule-based ordered list and the members of the initial list, and the rule-base pair Ordering in the list;
e. Selecting a node satisfying a fourth empirical rule from among the node pairs in the rule base pair list;
f. Forming an N-tree of a given root node selected from the initial list for all receiving nodes in the group based on a fifth empirical rule;
g. Ensuring that all links between nodes of a particular N-tree are connected;
h. Guaranteeing all nodes in a particular N-tree based on the nodes of the initial list included in the N-tree structure;
i. Guaranteeing all N-trees based on the nodes of the initial list included in a plurality of N-tree structures.
N-tree construction method.   j. The method of claim 1, further comprising the step of optimizing the N-tree by link movement and link reordering. The step of optimizing the N-tree includes:
If a link exists between two nodes with unused bandwidth / slots, the link of the N-tree is moved by moving the link to another node so that the sixth empirical rule is still valid Rearranging, and
If there are two or more logical links between two identical nodes and the first link has more intermediate nodes compared to the second link, the seventh empirical rule is satisfied and the former Performing a link move of the N-tree such that the link is replaced by the latter link so that it has a short N-tree.
The N-tree construction method according to claim 2.   k. The N-tree construction method according to any one of claims 1 to 3, further comprising a step of updating the constructed N-tree information to the optimum stream path information table and the server bandwidth fairness forwarding table.

Images