WO2004066569A1

WO2004066569A1 - Communication device, network system, and link generation method

Info

Publication number: WO2004066569A1
Application number: PCT/JP2003/012549
Authority: WO
Inventors: Akihiro Ito; Tomoki Sekiguchi; Naoki Utsunomiya
Original assignee: Hitachi, Ltd.
Priority date: 2003-01-21
Filing date: 2003-09-30
Publication date: 2004-08-05
Also published as: JPWO2004066569A1; JP4231002B2; AU2003271071A1

Abstract

Each node sets the maximum number of its links according to its performance and resource in advance. A node which tries connection to a network cluster determines a node to be actually connected from the node basic information including the number of used links and the maximum number of links of the node constituting the network cluster.

Description

Communication device, network system, and link generation method

Technical field

The present invention relates to a (Peer-to-Peer) type autonomous distributed system, and more particularly, to a network forming method. Akita

Background art

In a P2P network system in which nodes on the network communicate directly with each other without a centralized server intervening in the communication, search request messages are sent to the network when searching for the entire network. Must be broadcast to the whole. Normally, broadcast processing is performed by transferring a search request message or the like from node to node. In order to enable such processing, it is necessary to form an inter-node link such that all nodes on the network are connected to each other through the link, and to perform appropriate message routing based on the link.

As a link formation method, A Scalable Content-Addressable Network (SlGCOMM 2001 Proceeding) discusses an algorithm for selecting a node that forms a link by considering a tiled division of a virtual space. I have. Japanese Patent Application Laid-Open No. 2002-171283 discloses a method in which nodes that are geographically close to each other form the above-mentioned link using a GPS (Global Positioning System) device.

In a P2P network system, when a node moves or disappears, the network Network topology needs to be changed. Particularly in a mobile environment represented by wireless LAN, network topology changes frequently occur. A In the Scalable Content-Addressable Network, the process of changing the network topology when a node disappears is complicated.

Normally, a TCP link is used as the above link. However, small terminals with poor computer resources have strict upper limits on the number of TCP links that can be used. Even if the number of TCP links is sufficient, if the number of links is not limited, centralized access to specific terminals will occur, which may lead to performance degradation. Therefore, although the control of the number of links is important in a P2P network system, these problems are not considered in the conventional technology, and the number of links cannot be set for each terminal.

It has not been sufficiently considered to easily form an inter-node link that minimizes the number of hops (the number of message transfers when a certain message is transmitted).

Furthermore, in Japanese Patent Application Laid-Open No. 2002-171283, since the physical location information is used when selecting a node for establishing a link, it cannot be applied to a system having no means for obtaining the physical location information. DISCLOSURE OF THE INVENTION An object of the present invention is to flexibly cope with a change in network topology at the time of node generation and disappearance with a simple algorithm.

It is another object of the present invention to provide a P2P network system capable of easily forming a link in consideration of a limit on the number of links between nodes.

Another object of the present invention is to minimize the number of hops required for message transfer. The goal is to automatically generate an optimal network topology that will be smaller in a P2P network system.

According to the present invention, in the P2P network system, when the first node establishes a link with the second node, the first node uses the second basic information based on the node basic information including the maximum number of links of the node existing on the network and the current number of links. Determine the node. By this link establishment means, it is possible to construct an efficient network topology considering the number of links used by each node.

The maximum number of links in the basic node information is determined based on the node performance including CPU performance and network performance, the amount of data held by the node, and the quality of the data, and is held by each node. This makes it easy to establish a network topology that takes into account the limited number of links for terminals that should limit the number of links, and to a network that takes into account terminal performance such as concentrating access to high-performance terminals that can withstand loads. Control and application efficiency can be improved.

Also, based on the basic node information, among the new nodes whose current number of links does not exceed the maximum number of links, the node with the highest number of current links is selected as the partner node for link establishment, and message transfer is performed. Reduce the number of hops required.

Further, in the present invention, each node constituting the network cluster exchanges the network information including the node name and the destination of the node in the network cluster with each other to update the routing information in the network class. keep. This network information is transferred to the entire network cluster using the inter-node link when the inter-node link is established or when the inter-node link is lost. In order to carry out routing efficiently, the above network information includes the number of hops of the inter-node link until reaching the target node. Further, the present invention makes it possible for each node to broadcast the node basic information so that nodes existing in a range where the broadcast reaches can reliably recognize each other and establish a node-to-node link. I do.

Further, in a modified example of the present invention, by including the node basic information in the network information, the node basic information of the nodes constituting the network cluster is linked to the entire network class. Transfer using. Then, some of the nodes constituting the network cluster broadcast the node basic information. By limiting the nodes that perform broadcast processing, the network load can be reduced. The node that performs the broadcast process is determined from the network information.

Further, in another modified example of the present invention, the above broadcast processing is unnecessary by inquiring the we U-known node about the node basic information of the nodes constituting the network cluster.

;

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a diagram showing a configuration of a node according to an embodiment of the present invention.

FIG. 2 is a diagram schematically illustrating a connection state of a node group.

FIG. 3 is a diagram showing a node basic information table.

Fig. 4 (a) and Fig. 4 (b) show the link objects generated in a certain node.

Figure 5 shows the node-to-node generation process. G_chart1. . —

FIG. 6 is a flowchart 2 showing an inter-node link generation process.

FIG. 7 is a diagram showing a data structure of a connection request.

FIG. 8 is a diagram showing a data structure of a connection response. Figure 9 shows a link object configured in another node.

Figure 10 shows another link object.

FIGS. 11A and 11B show a network information update message transmitted by a certain node.

Figure 12 is a flow chart showing the network information update message transfer process.

FIGS. 13 (a), 13 (b), and 13 (c) are diagrams showing link objects configured in different nodes.

Figure 14 shows another network information update message.

Figure 15 is a diagram showing the inter-node link deletion process when a node disappears.

Figs. 16 (a) and 16 (b) show different network information update messages.

Fig. 17 (a) is a flowchart showing the transmission process of an application message.

Fig. 17 (b) is a flowchart showing the process of receiving an application message. BEST MODE FOR CARRYING OUT THE INVENTION

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. Figure 1 is a node

A hardware configuration of a portable terminal as an example of 101 and a software configuration for node management provided in each node are shown.

The node management program 1Q_2 stored in the storage device 107 includes a message processing unit 103, a link generation unit 104, and a link management unit. The message processing unit 103 delivers the message sent through the inter-node link to the component that actually performs the message processing. For example, If the above message is an inter-node link management control message, it is delivered to the link management unit 105, and if it is an application data message, it is delivered to the application. The link generator 104 performs a process of generating a link between nodes. The link management unit 105 stores the basic information (node name, maximum number of links (maximum number of usable links), number of used links (number of currently used links), destination) of the nodes existing in the network. It manages the node basic information table and the linked objects.

The above-mentioned node basic information table is used as a hint for selecting an inter-node link establishment partner when newly establishing an inter-node link. A link object is an object corresponding to a node-to-node link, and network information of all nodes that can communicate using the link (node name, approximate node is reached using the node-to-node link). Number of hops required to perform the operation, and the approximate node destination). This link object is created at both nodes that establish the link between the nodes. The link object represents a virtual link between nodes, and the actual means of communication between nodes is arbitrary (TCP, UDP, etc. can be used).

The maximum number of links is a parameter that can be set for each node. For example, by setting the maximum number of links of a node that can withstand a larger load (a node with high CPU performance / high network performance) larger than that of a normal node, efficiency in performance can be improved. In addition, by increasing the maximum number of links for nodes that have a lot of information such as a data base server and nodes that have high-quality information, application efficiency can be improved. It can be planned.

The CPU 108 sequentially reads out the node management program 102 from the storage device 107 via the backup memory 109 and executes it. The result of data processing performed by the CPU 108 is visually output to the display 110 as necessary. Is done. When a user's input is required for the processing of the node, the user's input is received from the keyboard 112 and processed by the CPU 108. 1 1 1 is a wireless communication device for communicating with other nodes, and 1 1 3 is a node management program that reads out from a storage medium such as a CD-ROM when the node management program is stored in the storage medium. Used when storing in a storage device. The components included in the node are connected to each other via a bus 114 so as to be able to communicate with each other. Although the configuration of the mobile terminal is shown in FIG. 1, a keyboard, a display medium reader / writer, and the like may not be provided as long as the node management program can be operated. In this embodiment, the node management is realized by the program and the processing of the CPU. However, each functional module of the node management program may be formed as an integrated circuit and constituted by a hardware circuit group.

FIG. 2 schematically shows an example of a connection form of a node group by an inter-node link. The solid line represents the link between the nodes that has already been established, and nodes 1 to 5 constitute a network cluster. Node 6 is trying to connect to node 4 (establishing an inter-node link is called a connection), but the inter-node link has not yet been established. At this time, a node basic information table and a link object as described below are configured in the node 4.

FIG. 3 shows a node basic information table 300 held by the node 4. This table is generated by the link manager of each node collecting the basic information (node name, maximum number of links, number of used links, destination) of the nodes to be broadcast on the network. Therefore, all nodes existing within the reachable range of the broadcast hold the same node basic information table. In the node basic information table 300, a URL according to the HTTP protocol is used as the destination of the node. However, this can be anything that can be used as the destination, for example, an IP address and a port. A set of G numbers may be used.

FIG. 4 shows a link object held by the link management unit of the node 4.

Link object 410 corresponds to the internode link established between nodes 4 and 1, and link object 402 corresponds to the internode link established between node 4 and node 5. Corresponding to From the viewpoint of node 4, node 2 and node 3 are connected to each other by an inter-node link, so link object 401 contains information on nodes 2 and 3. Which node in the link object is the link establishment partner of the node can be determined from the fact that the hop number of the node is one. Each node has a link object corresponding to the inter-node link that it has established, and the link object has all the nodes connected by the inter-node link ahead of the corresponding link. Since the network information is included, each node can enumerate all the nodes in the network class by enumerating the network information included in all the link objects in its own node.

Next, the process of generating a link between nodes when a new node connects to the network cluster will be described using an example in which node 6 connects to node 4. First, the node basic information table management processing will be described. In this processing, means for enabling all nodes to communicate with each other irrespective of the presence or absence of an inter-node link, for example, communication means using a broadcast address, is used as inter-node communication means. Link management unit of each node periodically blow H key _£ scan the basic information Jinoichido the network.

Node 6 receives the broadcast message and registers the information of the node in the node basic information table. Nodes that have not received the above broadcast message for a certain period of time are sent to the node basic information. Delete from table. In this way, the node basic information table in node 6 is always kept up to date. For example, when node 6 appears in the network shown in FIG. 2, the node basic information table in node 6 becomes as shown in FIG. Since this node basic information table management process is performed by all nodes, all nodes hold the same node basic information table.

FIG. 5 is a flow chart showing the operation of the CPU 108 reading and executing the link generation unit. Hereinafter, a process in which the node 6 generates a new inter-node link will be described with reference to FIG. For the sake of explanation, the subject of the operation is simply referred to as a link generation unit (the same applies to the link management unit / message processing unit described below). The link generation unit of the node 6 confirms that the number of used links of the own node is smaller than the maximum number of links of the own node. If the number of used links ≥ the maximum number of links, the link generation processing is stopped and a predetermined time is waited for (501). Next, the link generation unit of node 6 checks whether a new node exists in the network. The new node is a node that is included in the node's own node basic information table but is not included in any of the link objects. In other words, a node that does not exist in the network cluster including its own node but is included in the basic node table becomes a new node. At this point, since the node 6 is not connected to any node, all other nodes included in the node basic information table become new nodes (503).

Next, the link generation unit of the node 6 selects a node for which the actual connection / connection process is to be performed from the new nodes detected in step 503. The selection method is to select the node with the largest number of used links from the nodes with the smaller number of used links. By adopting this selection method, a network topology with a small number of hops at the time of communication through an inter-node link is used. Can be generated. Since node 6 holds the basic node information table as shown in Fig. 3, node 6 selects node 4 with the maximum number of links of 3 (common to each node) and the number of used links of 2 as the connection partner (5 0 4 ), Next, the link generation unit of the node 6 transmits a connection request 701 to the node 4 as a connection partner. FIG. 7 shows the data structure of the connection request 7001. This connection request includes the network information of all nodes in the network cluster of which the connection request sending node is a part. At this time, since node 6 is not connected to any node, the network cluster including node 6 is composed of node 6 only. Therefore, the connection request transmitted by the node 6 is only the network information of the node 6 itself (505).

If the link generation unit of the node 4 receives the connection request and permits the connection, the link generation unit of the node 4 generates a link object corresponding to an inter-node link with the node 6 within the node 4, and A connection response 8 01 (shown in Fig. 8) is returned to 6.

Connect response 81 contains network information for all nodes in the network cluster of which node 4 is a part. Detailed processing in node 4 will be described later. Upon receiving the connection response 801, the link generation unit of the node 6 generates a link object 901 corresponding to the inter-node link with the node 4. The network information to be registered in the link object 901 can be obtained by adding 1 to the number of hops of the network information in the connection response 801. If the connection response is rejection or no response, the link generation unit of the node 6 waits for a certain period of time and then returns to step 501 (steps 50_ to 5Q8).

If node 6 has established an existing inter-node link (inter-node link corresponding to a link object other than the link object generated in step 508), the link generation unit of node 6 Between links And send a network information update message. In this example, since node 6 has no link between existing nodes, there is no processing in this step. The specific processing content of this step is the same as that of step 606 described later. After waiting for a certain period of time, the process returns to step 501 (509).

In step 502, a certain period of time is waited until the next link generation processing is performed. This wait time may be varied depending on parameters such as the number of used links and the maximum number of links. For example, if the number is varied in proportion to (number of used links / maximum number of links), nodes with a small number of connections will actively try to connect to other nodes, and nodes with a large number of connections will wait for connections from other nodes. More. This makes the inter-node link generation process more efficient. Also, in order to avoid a situation where a plurality of nodes issue connection requests simultaneously, it is desirable to use a random number in determining the waiting time.

Next, the processing of the node 4 that has received the connection request from the node 6 will be described with reference to FIG.

When the link management unit of node 4 receives the connection request 701 from node 6, the link generation unit of its own node (node 4) is requesting a connection (processing other than the waiting process in step 502) If not). If a connection request is being made, a connection refusal is returned to node 6. This process is necessary for the normal operation of the inter-node link generation process even when multiple nodes send connection requests simultaneously (step 60016). Next, the link management unit of node 4 confirms that the number of used links is smaller than the maximum number of links ^ _ If (used _ number of leans) _ ≥ _ (maximum number of JJs). (If (number of used links) <(maximum number of links) is not satisfied), the link manager returns a connection refusal (603).

Next, the link management unit of the node 4 transmits the network The link object 1001 is generated by adding 1 to the hop number of the link information.

As a result, node 4 holds three link objects 401, 402, and 1001 (604).

Next, the link management unit of the node 4 returns a connection response 801 to the link request transmission node (node 6). The connection response 8001 includes the network information held by the link object other than the link object generated in the previous step and the network information of the own node. That is, the connection response includes the network information held by the link objects 401 and 402 and the network information of the own node (605).

Next, the link management unit of the node 4 transmits a network information update message to the existing inter-node link (inter-node link corresponding to the link objects 401 and 402). If there is no link between nodes to be transmitted, the above network information update message is not transmitted. The above network information update message includes the network information of all link objects other than the link object corresponding to the link between the nodes to be transmitted and the network information of its own node (node 4). . For example, the message sent to the inter-node link corresponding to link object 401 contains the network information held by link objects 402 and 1001, so the network information The update message looks like 1 1 0 1. Similarly, the link management unit of the node 4 transmits a network information update message 112 to the internode link corresponding to the link object 402 (606). With the above processing, the link between nodes 4 and 6 is established.

After establishing the link between nodes, each node forwards the network information update message to the network class using the existing link between nodes. Hereinafter, this network information update processing will be described using the above specific example with reference to FIG.

Upon receiving the network information update message 111 from node 4, the link management unit of node 1 confirms that the link generation unit of its own node (node 1) is not requesting a connection. If a connection request is being made, the process waits until the connection request processing is completed (steps 1201 to 1203).

Next, the link management unit of the node 1 once stores the link object corresponding to the inter-node link (the inter-node link established with the node 4) that has received the network update message 111. Discard. Then, a link object 1301 is generated by the method described in step 60.4. The above link object 1301 is a new link object replacing the link object discarded earlier. As a result, the link object in node 1 is as shown in Fig. 13.

If there is no change in the network information in the link object, the network information update processing ends (steps 1204 and 125).

Next, the link management unit of the node 1 sends a network information update message to an internode link corresponding to a link object other than the link object from which the network update message 111 was transmitted. Send. If there is no link object to send, the above message is not sent. The method of creating the network information update message to be transmitted is the same as the method of creating the network information update message in step 606. For example, the link management unit of the node 1 transmits a network information update message 1441 to the inter-node link _ corresponding to the link object 13. The link management unit similarly creates and transmits a network information update message to the internode link corresponding to the link object 1303 (step 1206 to 1208).

As described above, when a new node connects to a network cluster, a link object corresponding to the new internode link is generated, and then the network is connected to the entire network cluster through the existing internode link. Transfer the information update message. As a result, the link objects in all nodes that make up the network cluster are always kept up to date.

Next, processing when a node disappears from the network cluster will be described. Hereinafter, an example in which the node 5 disappears from the network class will be described with reference to FIG. When node 5 disconnects from the network, it sends a disconnect message to the internode link established by node 5. The node (node 4) that has established the link between node 5 and node 5 receives the above message, detects the disconnection of node 5, and if node 5 cannot send the message due to node crash, etc. Node 4 detects disconnection of node 5 due to communication failure with node 5 (1501).

Upon detecting the disappearance of node 5, the link management unit of node 4 confirms that the link generation unit of its own node (node 4) is not requesting a connection. If a connection request is being made, the process waits until the connection request processing is completed (steps 1502 and 1503). Next, the link management unit of the node 4 deletes the link object 402 corresponding to the inter-node link with the node 5 (1504).

Next, the link management unit of the node 4 transmits a network information update message to the established link between the nodes (the link object corresponding to the link object 401 and __10.01). The method for creating the network information update message is the same as the processing in step 606. In other words, node 4 links network information update message 16 01 The link is transmitted to the internode link corresponding to the object 1001, and the network information update message 1662 is transmitted to the link between the nodes corresponding to the link object 401 (step 1505). From step 1507).

The link management units of the nodes 1 and 6 that have received the network information update message execute the processing of steps 1221 to 1228. As described above, when a node disappears from the network class, the network information update message is transferred to the entire network cluster starting from the node where the node disappearance is detected. As a result, the link objects in all the nodes that make up the network class are always kept up to date.

Next, a process of transmitting an application message (hereinafter, referred to as an AP message) to a specific node using an inter-node link (message routing process) will be described. When transmitting the AP message, the message processing unit adds a data header consisting of the transfer route, destination node name, TTL (maximum number of hops), and route mode. There are two types of routing modes: “hop-by-hop mode” and “routing mode”. When “hop-by-hop mode” is used, each node that transfers the AP message determines the transfer route on demand, and the transfer route in the data header indicates the history of the AP message transfer. When using the "route specification mode", since the AP message creation node specifies the complete transfer path, the transfer path in the data header indicates the path from which the AP message will be transferred. Normally, in the RPC-type communication model, when sending a request message, the route to the destination node is unknown, so a hop-no-get is used, but when a reply message is sent, the request message follows. The route specification mode is used because the route to the destination node can be known from the received route history.

Below, Node 4 sends the AP message to Node 2 in hop-by-hop mode The case where the transmission is performed will be described as an example. Figure 17 shows the processing flow. First, the message processing unit of the node 4 selects a link object including the network information of the node 2 from the link objects held by the link generation unit. If there is more than one, select the link object containing the network information that minimizes the number of hops (1701). If there is no link object that satisfies the conditions, the transmission processing is regarded as an error and the actual transmission is not performed (1705). In this example, the link object 4 0 1 is selected.

The message processing unit generates the next data header (transfer route, destination node name, TTL, route mode) = (node4, node2, 5, hop-by-hop) and adds it to the AP message (17) 0 2), and transmits the AP message to the inter-node link corresponding to link object 4 0 1 (1704). In the above example, TTL is set to 5, but TTL is an item that the application can set freely. If the TTL is smaller than the number of hops in the network information of node 2 registered in link object 401, the transmission processing is considered to be in error and actual transmission is not performed (1703, 1705) .

The message processing unit of the node 1 that has received the above message performs retransmission processing when confirming that the destination node name is not its own node (1706). First, the message processing unit selects an inter-node link to which data is to be transmitted next. This processing is the same as the inter-node link selection processing in step 1701. Therefore, the message processing unit selects the internode link corresponding to the link object 1302 (1707).

Next, the message processing unit of the node_.1 adds the information of its own node to the transmission route of the data header and subtracts 1 from the TTL. That is, the data header is changed as follows: (transfer route, destination node name, TTL, route mode) = ((node4, nodel), node 2, 4, hop-by-hop). (1 7 0 8) Next, the message processing unit of node 1 confirmed that the TTL in the data header was equal to or greater than the number of hops in the network information of the destination node (node 2) registered in link object 1322. Thereafter (1709), an AP message is transmitted to the inter-node link selected in step R4 (1710). If the TTL is equal to or less than the above number of hops, the transmission process is regarded as an error and actual transmission is not performed (1771 1).

Upon confirming that the destination node is the own node, the message processing unit of the node 2 receiving the AP message delivers the data to the corresponding application (1712).

When the application returns the return AP message to the node 4 that transmitted the AP message, the application transmits the return AP message using the routing mode. Hereinafter, this processing will be described. The message processing unit of node 2 adds the following data header to the reply message.

(Transfer route, destination node name, TTL, route mode) = ((nodel, node4), nodel, 5, route specification)

The transfer path specifies the transfer path information of the data header included in the AP message received by the message processing unit in step R6 in reverse order. In the above example, the destination node name and TTL are specified. However, in the routing mode, since the complete route to the destination node is included in the transfer route, it is not actually necessary to specify the destination node name and TTL. There is no (1 7 1 3).

The message processing unit of node 2 recognizes that the inter-node link strength t has been established with the next transfer destination node (node 1), and uses the corresponding inter-node link. To send a reply AP message (1 7 1 4).

The message processing unit of the node 1 that has received the reply AP message deletes the information of its own node from the transfer route, and communicates with the next transfer destination node (node 4). Check that the inter-node link is established during Then, a reply AP message is transmitted using the corresponding inter-node link (steps 1706 to 1170: the judgment of 1709 is omitted when replying). By transferring data according to the transfer path described in the data header in this way, a desired message can be transmitted to a target node.

If the network cluster becomes large and the network information registered in the link object does not spread quickly enough on the network, a closed circuit may be formed in the link between nodes.

If a loop is formed, the AP message will continue to cycle until the TTL of the above data header becomes zero. To avoid an infinite loop of AP messages in a closed circuit, the transfer route in the above data header may be checked. In other words, when the message processing unit receives an AP message via the link between nodes, it confirms that its own node is not already registered in the data header transfer path. If the self-node has been registered, it indicates that the above-mentioned data has gone around a closed circuit, and the data is discarded.

The above checks need to be performed not only for AP messages but also for all messages sent and received using inter-node links. In other words, the message processing unit adds the above-mentioned data to the network information update message and performs an infinite loop avoidance check of the network information update message in a closed circuit. In this case, the above-mentioned data header is used to record the transfer route. Other information included in the f-data header, which is the destination zone name, is not used. The message processing unit discards the network update message that has gone around the cycle without delivering it to the link management unit. In addition, the transfer processing of the network information update message is performed by the link processor. Therefore, the message processing unit does not perform the message transfer processing.

When a closed circuit is formed, one of the inter-node links constituting the closed circuit may be disconnected (closed circuit disconnection processing). Hereinafter, this closed-circuit disconnection processing will be described.

The message processing unit adds a data header to all messages (AP messages, network information update messages) that transfer the link between nodes, and checks the routing information in the data header as described above. Thus, the closed circuit is detected. When a closed circuit is detected, the route information in the data header actually detected indicates a closed circuit. The message processing unit delivers this route information (hereinafter referred to as “closed information”) to the link management unit. Upon receiving the above-mentioned closed circuit information, the link management unit selects one inter-node link (a scheduled inter-node link) constituting the closed circuit. The link selection processing is performed by the link management units of all the nodes constituting the closed circuit, but as a result, the link management units of all nodes select the same link. After selecting the link between the nodes scheduled to be disconnected, the link management unit checks whether its own source has established the link. A node that has established a link to be disconnected disconnects the link.

Hereinafter, a method of selecting a link to be disconnected will be described. The link management unit obtains the ID value of the inter-node link using an appropriate hash function from the IP addresses (or node names) of the nodes at both ends of the inter-node link. Among the inter-node links that constitute a closed circuit, the inter-node link having the maximum (or minimum) ID value is the inter-node link to be disconnected.

If there are a plurality of inter-node links having the maximum ID value, the next inter-node link having the largest I_D value is the inter-node link: T—. In this way, comparison of ID values is performed until one inter-node link is determined. In the above selection method, all nodes perform the same calculation, so that all nodes determine the same internode link as the link to be disconnected. Theoretically, in the above method, it may not be possible to determine the only link between nodes until the end, but practical problems do not occur by using a useful hash value calculation algorithm such as MD5.

In the first embodiment, all the nodes broadcast the basic information of the own node. Although this method can reliably detect nodes, the network load increases in proportion to the number of nodes. In the second embodiment, a network configuration method is provided in which it is sufficient if at least one node performing the broadcast exists in the network class. The following mainly describes differences from the first embodiment.

Using the example shown in Fig. 2 (example where node 6 connects to node 4), a method for connecting a new node to a network cluster will be described. Each node that makes up the network cluster registers the basic information of all nodes in the network cluster of which it is a part in its own node basic information table. A node (broadcast node) in the network class broadcasts basic information of all nodes forming the network cluster. In this example, since the nodes 1 to 5 form a network cluster, the broadcast node broadcasts the basic information of the nodes 1 to 5.

The selection of the broadcast node may be performed by any method that can determine a specific node within the network class. For example, among the nodes registered in the node basic information table, a method in which the node with the smallest IP address becomes a broadcast node can be used. Alternatively, it may be determined by a node name or the like.

Upon receiving the above broadcast message, the link manager of node 6 merges the received basic information into its own node basic information table. I do. The link management unit of the node 6 executes the processing of steps 501 to 504, selects the node 4 as a link establishment partner between the nodes, and transmits and receives the connection request 701 and the connection response 801. Then, a link between the nodes is established (steps 505 and 506). In the present embodiment, the node 6 further includes the contents of the node basic information table of its own node in the connection request 7001. Node 4 also includes the contents of the node basic information table of its own node in connection response 8001. Upon receiving the connection request and the connection response, the link management units of the nodes 4 and 6 merge the basic information of the nodes contained therein into the respective node basic information tables.

After establishing the inter-node link, nodes 4 and 6 forward the network information update message to the entire network cluster through the existing inter-node link in the same way as in steps 1201 to 1208. I do. At this time, the link management unit of each node includes the contents of the node basic information table of its own node in the transferred network information update message. The link management unit of the node that has received the network information update message merges the basic information of the node included in the network information update message with the node basic information table of the own node.

In the first embodiment, when the node disappears from the network class, the network information update message is transferred to the entire network cluster, thereby updating the network information in the link object. In the second embodiment, the contents of the node basic information table are further included in the network information update message.

In other words, when the node disappears, the node disappears when the node disappears. The network information including the contents of the node basic information table after deleting the disappeared node from its own node basic information table. Send an update message to the internode link. Link management of the node that received the network information update message After updating the node basic information table of its own node in accordance with the network information update message, the management unit transmits a network information update message including the contents of the node basic information table to the nodes other than the receiving route. To the inter-link.

As described above, in this embodiment, when a new node is connected or an existing node disappears, the basic information of the node is included in the messages (connection request, connection response, network information update message) transmitted and received through the inter-node link. Thus, the node basic information table of each node constituting the network cluster is updated without relying on the broadcast processing. As a result, at most one node that performs broadcast processing in the network class is sufficient, and as a result, the network load can be reduced.

In the first and second embodiments, the basic information of the nodes constituting the network cluster is transmitted to the new node by using the broadcast. However, broadcast can usually be used only between nodes existing in the same network (segment), and is not suitable for communicating with nodes located in remote locations. The third embodiment provides a network configuration method that does not use any broadcast.

In this embodiment, it is assumed that there is at least one node (welt known node) that does not disappear (or does not frequently disappear) from the network class. Then, the node connected to the network class evening knows the destination of the above we l-known no H in advance :? T must have there _a _ - hereinafter with the method of the new node connects to the network cluster, will be described focusing on differences between the actual施例2. First, the link management unit of the node (node 6) that wants to connect to the network cluster is based on a well-known node. Send an information request message. The we known node that receives this message returns the contents of its own network basic information table. The message transmission and reception here is performed by a method that does not use the link between nodes (simple UDP data transfer, etc.).

5 Upon receiving the network basic information table, the link management unit of the node 6 updates the network basic information table of its own node to detect a node forming a network cluster. The detection method is the same as the method described in step 503. Thereafter, in the same manner as in the second embodiment, the node 6 establishes a link between the nodes and connects to the network cluster.

0 The basic information and network information of node 6 are transferred to the entire network cluster by the network information update message transfer process. Since the we U-known node is part of the network cluster, the above information is also transferred to the known node. Therefore, the next time another node sends a basic information request message to the node kno wn, the well-known node returns the basic information of the node group including the information of node 6. Become.

Node disappearance processing is the same as in the second embodiment.

As described above, according to this embodiment, by assuming the existence of a well-known node, the network system described in Embodiments 1 and 2 can be used even when broadcast is not available. It becomes. It is sufficient that at least one 0 we U_known node exists in the network cluster. By increasing the number of we11-known nodes, the load on the we1known node can be reduced.

_ —— In addition, the method of Embodiment 1 is adopted if the presence of a network class can be detected by broadcasting, and the method of this embodiment is adopted otherwise. How to use it is also conceivable. As a result, the number of queries to the we ll-known node is reduced, and consequently As a result, well-known node load, network load, etc. can be reduced.

As described above, according to the present invention, an optimum network topology that minimizes the number of hops required for message transfer can be automatically generated in a P2P network system. Further, according to the present invention, it is possible to flexibly cope with a change in network topology at the time of node generation and disappearance with a simple algorithm.

Further, according to the present invention, the maximum number of links can be set for each node, so that the above network topology takes into consideration the performance and resources of the node.

Claims

The scope of the claims

1. A communication device that communicates with another device to configure a network, and that stores node basic information including the maximum number of links of a node existing on the network and the current number of links,

A selecting unit for selecting a candidate of a partner node for establishing a link based on the basic information.

A communication device, comprising:

2. The communication according to claim 1, wherein the selecting unit sets a node having the largest current link number among nodes whose current link number does not exceed the maximum link number as a link partner node candidate. apparatus.

3. The device according to claim 1, further comprising a link generation unit that establishes a communication link with another device, wherein the link generation unit configures a partner device selected by the selection unit and a network to which each device already belongs. Communication equipment for exchanging network information including node names and destinations of a set of nodes

4. In claim 1, further comprising a link generation unit for establishing a communication link with another node, wherein the link generation unit determines the current number of links of its own device when receiving a connection request from another node. A communication device characterized by comparing a maximum number of links and transmitting a connection permission response to another node to establish a link when the current number of links does not indicate a maximum link number ________________________ or not a link.

5. Claim 3 further includes a link management unit, and the link management unit A communication device for transmitting and receiving network information using an established link, and holding network information obtained through the link in association with the link.

6. When the network information acquired through the link and the network information held by the communication device in association with the link have the same content, the link management unit suppresses retransmission of the network information. 6. The communication terminal according to claim 5, wherein:

7. The link management unit merges the network information acquired through the link with the network information already held by the communication device, and merges the merged network information with a link other than the link described above. The communication device according to claim 5, wherein the communication device is retransmitted.

8. The communication device according to claim 5, wherein the link management unit deletes network information associated with the lost link when detecting that the established link has disappeared.

9. The link management unit obtains the same result on all the nodes described above, when a cycle is formed in the path formed by the link, by identifying the IDs of all the inter-node links constituting the cycle. By selecting a link from the links between the communication terminals that constitute the closed circuit by calculating the link, and disconnecting the selected link, the link The communication terminal according to claim 5, wherein the communication terminal is disconnected.

10. The communication terminal according to claim 1, wherein the node held by the storage unit is A communication terminal that broadcasts basic information to a network periodically.

11. The link management unit determines the maximum number of links based on node performance including CPU performance and network performance, the amount of data held by the node, and the quality of data. The communication terminal of claim 1, wherein

1 2. In a network system in which each node autonomously determines the partner node to establish a link,

A storage unit for storing node basic information including a maximum number of links of the nodes existing on the network and a current number of links, and selecting a candidate node to establish a link based on the basic information; Selection section

A network system characterized by comprising:

13. The node selecting the node having the largest current link number among nodes whose current link number does not exceed the maximum link number as a link partner node candidate. Item 12 network system.

14. In claim 12, the node further includes a link generation unit that establishes a communication link with another node, wherein the link generation unit includes a partner node selected by the selection unit and each node. A network system that exchanges network information, including the node names and destinations of nodes that already belong to a network, and the network information.

15. The node according to claim 12, wherein the node further comprises a communication link with another node. A link generation unit that establishes a link and, when receiving a connection request from another node, compares the current number of links of the own device with the maximum number of links, and determines that the current number of links is the maximum number of links. A network system for transmitting a connection permission response to another node to establish a link when the number does not exceed the limit.

16. In claim 14, wherein the node further comprises a link management unit, the link management unit transmits and receives network information using an established link, and transmits the network information acquired through the link to the node. A network system characterized by linking and maintaining links.

17. The network system according to claim 12, wherein the node basic information stored in the storage unit is periodically broadcast to a network.

18. Some of the nodes that make up the network generate node basic information including the maximum number of links of the node and the current number of links from the network information held by the node, and form the network cluster. 13. The network system according to claim 12, wherein basic information of a node to be updated is periodically broadcast.

19. The communication system according to claim 12, wherein the selection unit is a node that configures a network cluster in a kjiown node on je1 with little change in destination—a node that configures a network class. And a candidate node to establish a link based on the node basic information received from the well-known node.

20. In a network system where each node autonomously determines the partner node to establish a link,

Each node holds the basic node information including the maximum number of links of the node existing on the network and the current number of links.

A link generation method characterized by selecting a candidate of a partner node for establishing a link based on the basic information.

21. The method according to claim 20, wherein, in the selecting step, a node having the largest current number of links among nodes whose current number of links does not exceed the maximum number of links is set as a candidate for a link partner node. Link generation method.

22. In claim 21, further, the network information including the partner node selected by the selection unit and the node names and the destinations of the node groups that constitute the network to which each node already belongs. A link generation method characterized by exchanging each other.

23. In claim 20, when a connection request is received from another node, the current number of links of the own device is compared with the maximum number of links, and if the current number of links does not exceed the maximum number of links. A link generation method characterized by transmitting a connection permission response to another node to establish a link.