JP2004274702A - Router device and packet transfer control method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To transfer a packet traffic to a user network which is multi-home-connected by efficiently using a plurality of links. <P>SOLUTION: Upon receiving a packet addressed to the user network (including P(X) in a destination), a distribution router (DR) provided in the Internet selects a route arriving at a border router (UR) of a user network through any one of connections (links A, B) provided to the user network by networks (ISP-A, B) of each of providers, and transfers the received packet to the selected route. A route may be selected on the basis of the traffic of each link or communication environment such as communication quality of each route. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

  The present invention relates to a service provided to a user network connected to a plurality of provider networks (performing a so-called "multi-home connection"), and more particularly to a mechanism for providing a path control for packet transfer from the Internet to a user network. . In this specification, the term “plurality” refers to a provider that provides “plurality” links (internet connection) to a user network, and a single Internet service provider (ISP) uses “ It is also possible to function as multiple providers. Similarly, each provider's network need not be physically separate, and a single network can function as a "plurality" of provider networks.

  In order for a user network (for example, a corporate network or a SOHO (Small Office Home Office) network) to obtain more reliable Internet connectivity, a technique called "multi-home connection" is used. This is because the user network receives an IP (Internet Protocol) address from each of the two or more ISPs, so that even if the link to one ISP goes down, the link to another ISP will go down. Thus, Internet connectivity can be ensured.

  As such a technique, two systems shown in FIGS. 1A and 1B have been proposed (for example, see Non-Patent Document 1). It is assumed that the user network is assigned a prefix A as an address from the ISP-A and a prefix B as an address from the ISP-B. The prefix refers to upper predetermined bits of the IP address, and specifies a network. The remaining bits of the IP address can be assigned to a plurality of hosts in the network as information for identifying each host.

  The user network includes a user border router-A provided at a connection point with ISP-A and a user border router-B provided at a connection point with ISP-B. These user border routers share routing information by IBGP (Internal BGP: BGP (Border Gateway Protocol) is one of the protocols for exchanging routing information in the Internet). Then, each user border router performs an address advertisement (announcement) to the Internet using the routing protocol BGP.

  In the method of FIG. 1A, the prefix A is normally advertised from the user border router-A via the ISP-A, and the prefix B is advertised from the user border router-B via the ISP-B. Therefore, a packet addressed to prefix A is transferred to the user network via ISP-A, and a packet addressed to prefix B is transferred to the user network via ISP-B.

  For example, when the connection between the user network and the ISP-B is lost, both the prefix A and the prefix B are advertised from the user border router A via the ISP-A. Then, the packet addressed to the prefix B is also transferred to the user network via the ISP-A, so that the Internet connectivity can be maintained.

  Even in the method shown in FIG. 1B, the prefix A is normally advertised from the user border router-A via the ISP-A, and the prefix B is advertised from the user border router-B via the ISP-B. Then, when the connection with the ISP-B is lost, the prefix B is advertised from the user border router-A, and the advertisement is sent to the border router of the ISP-B indirectly (ISP-A). BGP) (via the border router). Then, the packet addressed to the prefix B passes through the ISP-B, and then reaches the user network through the link to the ISP-A by tunneling from the border router of the ISP-B to the ISP-A. become.

There have been proposals to apply such a multi-homed connection technique not only to an IPv4 network but also to an IPv6 network (for example, see Non-Patent Document 2).
T. Bates et al., "RFC 2260: Scalable Support for Multi-homed Multi-provider Connectivity," January 1998, Internet Engineering Task Force (IETF) <URL: http: // www. ief. org / rfc / rfc2260. txt? number = 2260> J. Hagino et al., "RFC 3178: IPv6 Multihoming Support at Site Exit Routers," October 2001, IETF <URL: http: // www. ief. org / rfc / rfc3178. txt? number = 3178>

  In the method shown in FIG. 1A, since an address assigned by another ISP is allowed to be advertised via a certain ISP, the amount of routing information exchanged on the Internet may explosively increase. is there.

  In the method of FIG. 1B, even when the link between the user network and the ISP-B goes down, the prefix B is advertised from the ISP-B to the inside of the Internet, so that the routing information increases. Although this problem does not occur, as a practical matter, it is unlikely that the ISP will adopt such a method. This is because each ISP's user is charged according to the amount of traffic to and from the Internet provided by each ISP. That is, even though the link to ISP-B is down and ISP-A provides an alternative link, the user can be charged only for traffic between the user network and the Internet. This is because ISP-B is responsible, and ISP-A has only business disadvantages.

  Further, as described above, as a multi-homed user network, not only a connection of another ISP is used as a backup when a connection of one ISP goes down, but also a connection of both ISPs is maintained. It is desirable that both of these can be used effectively in the state. For this purpose, when there are a plurality of links to the ISP, a mechanism for distributing traffic between the user network and the Internet to the plurality of links is required.

  However, in the conventional method described above, in a normal time when a plurality of ISP links can be used, a packet addressed to prefix A reaches the user network via ISP-A, and a packet addressed to prefix B is sent to the user network via ISP-B. It is only possible to reach the network and controls to properly distribute traffic from the Internet to multiple hosts in the user network between the link from ISP-A and the link from ISP-B I couldn't.

  For example, if all hosts in the user network use the address assigned by ISP-A as their own address, no matter how much prefix B is advertised via ISP-B, all actual traffic will be -A, and traffic cannot be distributed. Also, even though half the hosts in the user network use addresses assigned from ISP-A and the other half use addresses assigned from ISP-B, most of the actual traffic is If the address of the ISP-A is for the host group, most of the traffic will be transferred via the ISP-A, and the traffic cannot be distributed.

  If the nodes in the user network use a unique address (for example, a PI (Provider Independent) address) instead of an address assigned by the ISP, the prefix of the unique address is advertised via both ISPs. As a result, there remains a problem that the amount of routing information exchanged on the Internet explosively increases.

  SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and has as its object to provide a multi-home connection method which is easy for an ISP to adopt as a real problem while preventing an explosive increase in the amount of routing information in the Internet. I do. Here, the multi-home connection means providing a plurality of links (connection to the Internet) from one or more ISPs to one user network. Another object of the present invention is to realize control for appropriately distributing traffic from the Internet to a plurality of links for a multihomed user network. Another object of the present invention is to provide a mechanism that can appropriately select a route that passes through any one of a plurality of links according to communication quality parameters such as delay, packet loss, jitter, and bandwidth.

  The present invention introduces a new router that controls packet transfer as a router configuring the Internet as follows. Here, the content distribution server and the like also correspond to the router according to the present invention. In the case of a server, a packet addressed to a user network is received not from outside the own device but from an application in the own device.

  When the router according to the present invention receives a packet destined for a user network provided with a plurality of connections to the Internet, based on the packet including the address of the user network, the router includes a plurality of connections provided to the user network. Select one of Then, in accordance with the address corresponding to the selected connection, the received packet is transferred to a route on which the packet is transferred from the router to the user network (a virtual tunnel is set). For this virtual tunnel, for example, the following three packet transfer methods can be used.

  According to the present invention, in order to realize multi-homed connection, only a part of a large number of routers constituting the Internet, which can function as the router according to the present invention, have the address of the user network and each connection. Since the information for knowing the correspondence with the address corresponding to the Internet need only be sent, the amount of routing information in the Internet does not increase explosively. Furthermore, according to the present invention, a route for transferring a packet from the router according to the present invention to the user network according to the address corresponding to the selected connection is used, so that a packet arriving at the user network through a certain connection (link) is used. Since the link is transferred via the network of the provider that provides the link, each provider can charge for the service provided by the provider.

  The router according to the present invention, when receiving a packet addressed to a user network provided with a plurality of connections to the Internet, based on the received packet and a communication environment provided to the user network using each connection. Then, a configuration may be adopted in which a route to the user network is selected, and the received packet is transferred to the selected route. The route selection determines which of the plurality of links (connections) the packet should reach the user network (from which of the plurality of provider networks the packet should be forwarded to the user network). By doing so. As a result, when transferring a packet to a multihomed user network, a packet transfer path can be selected based on the communication environment.

  The selection of the path is performed based on the received packet. For this purpose, the destination address, port number, transmission source address, port number, protocol ID, and the like, which are information in the received packet, can be referred to. For example, when referring to the destination address, even if the traffic is addressed to the same user network, for a traffic addressed to a certain host or a group of hosts, a link provided by a certain provider is selected as a route, and the traffic addressed to another host group is selected. As for, it is possible to select a link provided by another provider as a route.

  Further, by referring to information in a layer higher than the network layer, more detailed control is possible. For example, if you refer to the protocol ID, even if the traffic is addressed to the same host, select a route through one provider for voice call traffic and select a route through another provider for other traffic. Will be able to

  When the route is selected based on a communication environment provided to the user network using each connection (link), an example of the communication environment is a traffic between each provider network and the user network. Amount can be used. This is the amount of traffic between the Internet and the user network of each link, but may be the sum of both input and output traffic, or may be the amount of input traffic to the user network. These traffic volumes can be observed in the user network. Thus, the router according to the present invention selects, for example, if the traffic volume of a certain link is large, for a part of the packet group that has selected the route through the link, a route through another link is selected. It is possible to switch.

  As another example of the communication environment, communication quality parameters such as delay, packet loss rate, jitter, and band of each path from the router according to the present invention through each connection (link) to the user network can be used. . These communication quality parameters are measured for each route and may be observed in the user network, or may be obtained by the router according to the present invention exchanging messages with the user network. good. Thus, the router according to the present invention selects, for example, if a path through a certain link has a large delay, selects a path through another link for a group of packets carrying delay-sensitive data (for example, voice communication or video streaming). Can be switched.

  Here, the dynamically changing traffic volume and communication quality parameters are taken as an example of the communication environment in the present invention. However, a static value, that is, a value previously stored in the router according to the present invention, is used as the communication environment. You can also. For example, if it is known in advance that the bandwidth of a link provided by a provider to a user network is larger than the bandwidth of a link provided by another provider, the router according to the present invention refers to these bands as a communication environment. Then, a group of packets addressed to the user network can be distributed to a large number of routes and a small number of packets to another route according to the bandwidth corresponding to each route. Also, when it is known in advance that a route via a certain provider has a larger delay than a route via another provider, the router according to the present invention refers to these bands as a communication environment, and transmits packets addressed to the user network. For a group that carries delay-sensitive data, a path with a small delay can be selected.

  Also, for each connection (link) provided to the user network, whether the packet transfer path using the connection is operating normally (the link is disconnected or the reachability to the link is lost due to a failure in the provider network). It is also possible to select a packet transfer path based on whether or not the packet transfer path has been used. For this purpose, for example, the router according to the present invention transmits a confirmation message to an address corresponding to each connection (link), and determines whether or not there is a reply message to the address or the content of the response message, and the router using each link. What is necessary is just to confirm whether the route from to the user network is operating normally.

  By transmitting the above confirmation message at predetermined intervals, the current status of each route can be checked. However, the transmission interval may be changed based on the communication quality of each route. For example, when the bandwidth of a certain route is large, the transmission interval is shortened and the confirmation message is transmitted frequently, and when the bandwidth is small, if the transmission interval is extended and the transmission frequency is suppressed, the test traffic such as the confirmation message can be reduced. Therefore, it is possible to appropriately avoid excessively squeezing a band that can be used by actual traffic.

  For example, the following three methods can be used to transfer a packet along a selected route (a route according to an address corresponding to a selected connection). In this example, it is assumed that the address of the user network is written in the destination address field of the received packet, and the address of the source network is written in the source field. Here, as the address of the user network, an address assigned by any one provider may be used, or a user-specific address such as a PI address may be used.

  The first uses encapsulation of IP packets, and the router according to the present invention uses an address (preferably a border router of a user network) assigned by a provider that provides a selected connection in accordance with the connection. (The address assigned to the (user router)), and the received packet is wrapped and forwarded with a new packet whose source is the address of the own device. Upon receiving the encapsulated packet, the user router decapsulates the packet, refers to the destination address of the original packet, and transfers the packet to the destination host.

  The second one uses a technique called MPLS (Multi-Protocol Label Switching), and starts from the router according to the present invention as a starting point, passes along each connection (link) to a user network, along each path. It is assumed that a label switching path through which a packet is transferred is set based on a label that is lower layer information than the network layer. The router according to the present invention adds the label of the label switching path corresponding to the selected route to the received packet and transfers the packet. The user router receiving the packet from the label switching path removes the label and refers to the destination address of the packet on the network layer to transfer the packet to the destination host. Note that the label switching path is set according to the address assigned to the user router by the provider that provides each connection, and the user router, when setting up the label switching path, is a router that is the starting point of the label switching path (the present invention). Since the address of the router according to the present invention is notified, the address of the router according to the present invention can be obtained by referring to the label given to the received packet.

  Third, focusing on the fact that a host in a user network cannot communicate with the Internet unless through a user router, the router according to the present invention translates the destination address of a packet. That is, the router according to the present invention rewrites the destination address field of the packet based on the address assigned by the provider providing the selected connection corresponding to the connection (preferably to the user router to which the connection is provided). Transfer. Here, the information written in the original packet is lost, but in the user router which must pass through, the address is reversely converted and the information is restored. That is, the user router rewrites the destination address field of the received packet based on the address of the user network and transfers the packet to the destination host.

  The first and second methods have an advantage that the border router of the user network can know the address of the router (router according to the present invention) from which the route of the received packet has been selected. The third method has an advantage that the packet size (MTU: Max Transfer Unit) of the IP packet does not change.

  As information of each route to be selected in the router according to the present invention, for example, a user network address and a provider that provides each connection correspond to each connection (preferably to each user router to which each connection is provided). Information indicating the correspondence with the assigned address (this information is referred to as router registration information) is stored in the router according to the present invention.

  When this information is transmitted to another router that configures the Internet and that can function as the router according to the present invention, the other router that has received the transmission, like the router according to the present invention, An operation of selecting one from a plurality of routes to the network and transferring the packet is performed.

  On the other hand, when the router according to the present invention transmits information indicating the correspondence between the address of the user network and the address of its own device to the other routers constituting the Internet, the other router that has received the transmission transmits Since the packet addressed to the network address is transferred to the router according to the present invention, the router according to the present invention selects one of the transferred packets from a plurality of routes to the user network. And perform the transfer operation.

  The present invention introduces a new router as described above as a router configuring the Internet, while introducing a new router that controls packet transfer as follows as a user router. The user router is a router provided at a connection point with the Internet in a user network provided with a plurality of connections with the Internet. The user router according to the present invention may be configured as a device having a plurality of network interfaces so as to correspond to all of the plurality of connections, or as a device having a network interface corresponding to a part of the plurality of connections. You may comprise. In the latter case, each user router device corresponding to each connection may share routing information by, for example, IBGP or OSPF, and perform the function of a border router for multihome connection with the Internet as an aggregate.

  The user router according to the present invention is configured to provide the identification information of the user network to which the own device belongs and the address assigned to the own device among the addresses assigned corresponding to each connection (the address corresponding to the connection provided by the own device). ), And receives a packet addressed to the user network via a route according to an address corresponding to one of the plurality of connections. The route used for this reception is selected by the router according to the present invention, which constitutes the Internet. By transmitting the above information by the user router according to the present invention, the router according to the present invention can obtain information on each route to be selected.

  Then, the user router according to the present invention further observes a communication environment provided to the user network (via a link provided to the own device) by a provider network corresponding to the own device, and obtains information obtained by the observation. Is transmitted, the router according to the present invention can use this information as the above communication environment to select a route.

  As another method, the user router according to the present invention determines which one of a group of packets destined for the user network based on information obtained by observing the communication environment provided to the user network via each link. It is also possible to generate information instructing which part should be transferred along a path corresponding to an address corresponding to which link (specific selection criteria), and transmit this instruction information. In this case, the router according to the present invention can select a route by following this instruction information.

  As yet another method, the user router according to the present invention should be configured such that a part of a group of packets destined for the user network is transferred to the user network via a link that provides what communication environment to the user network. (Abstract policy) may be generated, and this instruction information may be transmitted. In this case, the router according to the present invention identifies the one of the plurality of links by comparing the instruction information with the information on the communication environment provided to the user network through each link obtained separately from the instruction information. By doing so, a route can be selected.

  When the router according to the present invention adopts the first or second of the three packet transfer methods described above, the user router according to the present invention performs route selection of a received packet. The user can know the address of the router (router according to the present invention), and by using this address, designates a specific router, and observes information on the communication environment provided to the user network, instruction information based on the communication environment, and the like. Can be sent. That is, it is possible to specify a router that needs such information and notify the router.

  A user router according to the present invention transmits, to the Internet, identification information of a user network, information on an address assigned corresponding to a connection provided by the own router, and these information (referred to as user registration information). For example, the following two methods can be used. Note that the router registration information used by the router according to the present invention for selecting a route is obtained based on the user registration information transmitted here.

  One is a method of transmitting the above-mentioned user registration information for a plurality of connections via one connection. In this case, when the other connection goes down, if the information on the address assigned to the user router corresponding to the downed connection is included in the above-mentioned user registration information, delete this information so that it is not transmitted. I do. As a result, the router configuring the Internet does not recognize the route using the down connection as a selection target, so that the connectivity of the user network to the Internet can be ensured.

  The other is a method of transmitting the above-mentioned user registration information for each of a plurality of connections via the corresponding connection. In this case, when a certain connection goes down, only the above-mentioned user registration information about the connection does not automatically arrive. Therefore, the router configuring the Internet does not recognize the route using the down connection as a selection target, and can ensure the connectivity of the user network with the Internet.

  As the user network identification information included in the user registration information, the address information of the user network may be used, or an identifier different from the address of the user network may be used. In the former case, the router according to the present invention can directly obtain the router registration information, but the user network declared by the user router must be registered so that incorrect routing information does not enter the Internet. It is desirable to confirm that the address is correct. In the latter case, the router according to the present invention obtains router registration information by obtaining the address of the user network based on the identifier transmitted from the user router.

  The present invention described above can be used not only as a packet transfer control method, a router device, and a user router device, but also at a computer functioning as a router configuring the Internet, or at a boundary of a user network provided with a plurality of connections to the Internet. The present invention can be implemented as a program that is installed and executed on a computer that functions as a user router, or as a storage medium that stores such a program.

  According to the present invention, a destination is a user network provided with a connection to the Internet from a plurality of providers (which may be a plurality of ISPs or a single ISP may provide a plurality of links). When transferring a packet, it is possible to appropriately select a route using any one of a plurality of connections.

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

  FIG. 2 illustrates the state of packet transfer in one embodiment of the present invention. ISP-A to ISP-E represent provider networks, and each network has a number of routers (not shown) to provide IP-based communication. The Internet is composed of such ISP-A to ISP-E and a number of routers (not shown) connecting these provider networks. In the Internet, routers (distribution routers) DR1 to DR4 newly introduced according to the present invention are provided.

  The user network is provided with a connection to the Internet from the ISP-A via the link A, and provided with a connection to the Internet via the link B from the ISP-B. A user router UR is provided at the boundary of the multihomed user network, and the UR has a first interface corresponding to link A and a second interface corresponding to link B. . The following description proceeds with an example in which one UR has two interfaces, but a first UR corresponding to the link A and a second UR corresponding to the link B are provided at the boundary of the user network, The same can be realized in an example in which necessary information is shared by two URs.

  There are many hosts (not shown) in the user network, and each host has an IP address X having P (X) as a prefix (predetermined high-order bit). This P (X) may be an address prefix A assigned from the ISP-A, an address prefix B assigned from the ISP-B, or a PI address prefix unique to the user network.

  When the present embodiment is applied to an IPv4 network, the address used by each host may be, as described above, an address assigned from ISP-A, an address assigned from ISP-B, or a PI address. good. In the case of an IPv6 network, route aggregation-type routing advertisement is performed, so it may be difficult to use a PI address. However, even in such a case, if an address assigned by any ISP is used, the present embodiment is used. Can be applied. Of course, the present embodiment may be applied using a PI address in an IPv6 network.

  The first interface of the UR has an address URa assigned from the ISP-A, and the second interface of the UR has an address URb assigned from the ISP-B. The upper predetermined bit of URa is A, and the upper predetermined bit of URb is B.

  In the present embodiment, a source host S1 that wants to transmit a packet to a host of a user network writes an address X having P (X) as a prefix at the destination of the packet, and transmits the packet to the Internet. Each of DR1 to DR4 advertises P (X) on the Internet separately from its own address. Therefore, the packet of the destination P (X) that has issued S1 is sucked into the DR that has arrived first (the closest DR from S1, and in the example of this figure, DR1). The packet delivery from S1 to DR1 utilizes a technique called anycast.

  A search table as illustrated in FIG. 3 is stored in DR1. When a packet destined for P (X) is received as described above, information (destination address / port, transmission source) in the packet is received. The search table is referred to using at least a part of the address / port, protocol ID, etc. as a search key (details will be described later). Then, DR1 transmits the received packet addressed to P (X) to the path from DR1 to UR through ISP-A and link A (solid arrow in FIG. 2), and from DR1 to ISP-B and link B. It can be understood which of the routes to the UR through the route (dotted arrow in FIG. 2) should be selected. Then, the UR receives the packet transferred through the link of the selected route.

  Similarly, when the source host S2 wants to transmit a packet to the host of the user network, if a packet addressed to P (X) is sent to the Internet, the DR2 closest to S2 receives this packet. Since DR2 also stores a search table similar to that of DR1 for the packet addressed to P (X), DR2 is linked to ISP-A from DR2 according to the result of referring to the information in the packet as a search key. Select one of a route from A to the UR (dotted arrow in FIG. 2) and a route from DR2 to the UR through ISP-B and link B (solid arrow in FIG. 2). I do.

  Also when the transmission source host S3 wants to transmit a packet to the host of the user network, the DR3 stores the same search table as that of the DR1 or DR2, and the route is selected in the same procedure.

  In addition, the route selection performed in each DR described above means that the tunnel is actually exited from the DR to the UR through a virtual tunnel (from the link A and the link B, Which way to reach the UR). This virtual tunnel suffices if an entrance (DR) and an exit (the first or second interface of the UR) are defined, and a plurality of routers (not shown) to be routed between the entrance and the exit (DR) are entered. It is not necessary to specify in (DR). Such a virtual tunnel can be formed by, for example, three methods shown in FIGS.

  FIG. 4A uses encapsulation of an IP packet. For example, a new IP header is added to a packet received by DR1 (X is a destination and S1 is a transmission source) as follows. If URa is obtained as a tunnel destination (packet transfer destination) as a result of referring to the search table in DR1, the destination of the IP header to be added is URa and the transmission source is DR1. Then, the router in the Internet forwards the packet based on the destination URa, and the packet arrives at the UR through the link A. The UR removes the IP header newly added by the DR, refers to the destination X of the original packet, and transfers this packet to the destination host in the user network.

  FIG. 4 (b) uses the MPLS technology. From the entrance to the exit of the tunnel, packets are referred to by referring only to information (labels) of a layer lower than the network layer (without referring to an IP address). A label switching path is set so that the packet can be transferred. In the label switching path (LSP), only two paths from the DR to the URa (the first interface of the UR) and the path from the DR to the URb (the second interface of the UR) are always set. Alternatively, various LSPs may be set as needed in accordance with a packet flow having a finer granularity.

  In the former case, there are two types of label values stored as the tunnel destination (packet transfer destination) in the DR search table, and selecting the label A means selecting the route through the link A. , Label B means selecting a path through link B.

  In the latter case, even if a packet flow passes through the same URa (link A), an LSP dedicated to a packet flow to a specific destination host is set separately from an LSP accommodating a packet flow to another destination host, An LSP dedicated to a packet flow (of a specific protocol ID) carrying voice call data is set separately from an LSP accommodating a packet flow carrying other data, and a destination / source, an address / port, a protocol ID, A dedicated LSP for any combination can be set. In this case, in the DR search table, information that defines a packet flow of a fine granularity is stored as a search key, and the label of each LSP is stored corresponding to each packet flow.

  Then, for example, in DR1, when a packet to which X is the destination is received, the packet is decomposed into lower-layer (for example, data link layer) frames, and a label (for example, link A, LSP that passes). A router in the Internet then forwards these frames at a lower layer than the network layer based on the label, and each frame arrives at the UR via link A. The UR removes the label, assembles the frame into the original packet, refers to the destination X of the original packet, and transfers this packet to the destination host in the user network.

  FIG. 4 (c) shows an example in which the address conversion is performed at the entrance of the tunnel and the reverse conversion is performed at the exit. For example, the destination address of the packet (X is the destination and S1 is the transmission source) received at DR1 is And so on. As a result of referring to the search table in DR1, assuming that A (address prefix assigned by the user network from the ISP-A) is obtained as the tunnel destination (packet transfer destination), the upper predetermined bits of the destination address X are determined. Replace P (X) with A. Then, since the router in the Internet forwards this packet based on the destination address prefix A, the packet arrives at the UR through the link A.

  In the UR employing the address translation method, when an IP packet including the prefix A in the destination address is received, predetermined upper bits of the destination address of the received packet are replaced with A to P (X). Then, since the destination address of the original packet is restored, this packet is transferred to the destination host in the user network with reference to this destination X. When the UR receives an IP packet including the prefix B in the destination address, the UR replaces the upper predetermined bits of the destination address from B to P (X). In the case of FIG. 4 (c), since the information of P (X) is lost from the packet while passing through the tunnel, the UR at the exit of the tunnel needs to use A or X to recover the lost information. An address conversion table for converting B into P (X) is stored in advance.

  In the address conversion method, the DR that converts P (X) to A or B and the UR that converts A or B to P (X) require mapping because the number of destination addresses is insufficient. Since the operation becomes difficult, the addresses from the ISP-A and the ISP-B are set so that the number of addresses included in each of the prefixes P (X), A, and B is the same (or A and B are larger). You will get an assignment. The advantage of the address translation method is that it does not add a header unlike IP packet encapsulation, so the amount of data in one packet does not change before and after entering the tunnel, and the additional header cannot be included. It is not necessary to divide data into many packets and send it.

  When referring to the search table of FIG. 3, each DR searches for the best matching entry using the contents of the header of the received packet as a search key (this method is called “best match”). In the example of FIG. 3, a packet addressed to P (X) = 10.1.1.x is received, and if the destination address of the packet is 10.1.1.1, URa is output as the tunnel destination and the received packet is If the destination address is 10.1.1.2, URb is output as the tunnel destination. Then, the packet is transferred by the method shown in FIG.

  Also, assuming that the label value of the LSP from DR to URb, which is set for the packet flow of Web communication, is 100, the port number and the protocol ID of the received packet are referred to FIG. If represents the Web communication, the label 100 is output as the tunnel destination. Then, the packet is transferred by the method shown in FIG.

  Furthermore, in the example of FIG. 3, if the source network of the received packet is 1.1.1.x, prefix A is output as the tunnel destination, and the source network of the received packet is 2.2.2.x. For example, prefix B is output as a tunnel destination. Then, the packet is transferred by the method shown in FIG.

  FIG. 3 shows an example of the search table in the case where each DR can handle a plurality of types of tunnels. However, the type of the tunnel which each DR can handle may be a single type. If it is a corresponding DR, either URa or URb will be output from the search table. In the above description, an example in which a table entry is selected based on the best match for the entire search key is described. However, only the destination address portion of the search key is the best match, and the remaining portions (ports, transmission sources, etc.) May be referred to in order from the top to select the first matching entry. Alternatively, an entry that matches at the beginning of the table for the entire search key (regardless of whether it is the best match) may be selected.

  In the example described with reference to FIG. 2, all DRs have a function of selecting one of two routes (links A and B) to the user network. That is, all DRs have a search table for packets addressed to P (X) as shown in FIG. In this method, since there are actually a plurality of user networks, as shown in FIG. 5A, all DRs have information on a plurality of routes (the search table in FIG. 3) for all the user networks. And the amount of information in the search table of each DR and the amount of message exchange between DRs described later increase.

  On the other hand, as shown in FIG. 5B, if not all DRs have a plurality of routes to one user network but a specific DR, a search table for each DR is reduced. be able to. That is, when a packet addressed to a certain user network is received by a device other than the specific DR, the packet is transferred to the specific DR. When the packet reaches a specific DR, the specific DR selects one of a plurality of routes to the destination user network. In this method, each DR has information on a plurality of routes (the search table in FIG. 3) for a user network in charge of itself, but only transfers a packet to another DR for other user networks. .

  FIG. 6 shows an example of packet transfer in the case of FIG. 5B. When the source hosts S1, S2, and S3 that want to send a packet to a user network host write an address X prefixed with P (X) to the destination of the packet and send it out to the Internet, the DR closest to the source becomes (DR1 from S1, DR2 from S2, DR3 from S3) is the same as in FIG.

  Since DR1 is the DR that is in charge of the user network P (X), referring to the search table in FIG. 3 based on the packet received from S1, the DR1 reaches the UR through the ISP-A and the link A through the link A. A packet is transferred by selecting one of a route (solid arrow in FIG. 6) and a route from DR1 to UR through ISP-B and link B (dotted arrow in FIG. 6).

  Since DR2 is also the DR that is in charge of the user network P (X), it refers to the search table in FIG. 3 based on the packet received from S2 and reaches the UR from DR2 through ISP-A and link A through link A. A packet is transferred by selecting one of a route (dotted arrow in FIG. 6) and a route from DR2 to UR through ISP-B and link B (solid arrow in FIG. 6).

  Since DR3 is not the DR that is in charge of the user network P (X), when receiving the packet addressed to P (X) from S3, the routing table is searched to obtain the address of DR2. Since DR2 has notified other DRs in advance that it will select a plurality of routes for a packet addressed to P (X), the routing table of DR3 searches DR2 using P (X) as a key. It is set so that the address is output. Therefore, the packet addressed to P (X) from S3 is transferred from DR3 to DR2, and the path is selected at DR2 as described above, and the packet is transferred to the UR through the selected path.

  Since DR4 is not the DR that is in charge of the user network P (X), when receiving a packet addressed to P (X) from a nearby source host, it searches the routing table and obtains the address of DR2. The correspondence between P (X) and DR2 is also set in the routing table of DR4 by notifying other DRs in advance that DR2 will select a plurality of routes for a packet addressed to P (X). Here, an example in which a packet addressed to P (X) is transferred directly from DR4 to DR2 has been described, but a packet is transferred from DR4 to DR3, and a packet is further transferred from DR3 to DR2. Is also good. To this end, DR2 does not notify all other DRs of the correspondence between P (X) and DR2, but notifies a nearby DR (for example, DR3), and DR3 that receives this notifies a further nearby DR (for example, DR3). DR (4) may be notified of the correspondence between P (X) and itself (DR3).

  The method of transferring a packet flowing into the user network has been described above with reference to FIGS. 2 and 6, but the transfer of a packet flowing from the user network to the Internet may be performed as follows. For example, when a host in the user network wants to send a packet to S2, a packet with the destination set to S2 and the source set to X is sent from the UR via the ISP-A from the first interface or via the second interface. Can be sent to the Internet via ISP-B.

  However, this is a case in which the ISP performs an input filter and is set so that packets other than a packet whose source address is the IP address assigned by the own ISP do not pass through the own ISP, and the address X is assigned by the ISP to be passed. If not, the UR sets up a tunnel to the DR in the upstream direction (opposite to the arrow in the figure) and transfers the packet there. The tunnel in the upstream direction is set, for example, when the prefix A of the ISP-A is used as P (X) and it is desired to transmit via the ISP-B, or when the P (X) is the PI address prefix. Is used. The tunnel in the upstream direction can be realized by the three methods shown in FIGS. 4A to 4C, in which the UR transfers a packet by the same operation as that of the DR shown in FIG. The operation will transfer the packet.

  FIG. 7 exemplifies a state of message exchange performed by the UR and the DR to realize the packet transfer as shown in FIGS. The message exchange may be performed by using a protocol defined for another purpose on the Internet, or a new protocol may be defined for the UR and the DR of the present embodiment. In each UR or DR, a module for network layer processing for performing the above-described packet transfer and a module for performing message exchange described later using a protocol operating at a layer above the transport layer are separately provided. May be.

  First, the UR transmits a message for registering the correspondence between its own P (X) and URa and URb to the DR. This registration message is transmitted using anycast. That is, the registration message issued by the UR is received at the DR closest to the UR and processed there. The UR may transmit the registration message by unicast instead of anycast. When unicast is used, transmission is performed with one or a plurality of DR addresses registered in the UR in advance as destinations.

  This registration message can be sent in the following two ways. One is that the UR determines the correspondence between the address URa of the first interface and P (X) from the link A (first interface), and the link UR of the second interface from the link B (second interface). This is a method of transmitting the correspondence between the address URb and P (X). The other is a method in which the UR selects one of the links (for example, link A), and transmits the correspondence between a plurality of addresses (URa and URb) and P (X) from the selected link.

  In either case, it is not necessary to register URa and URb at the same time. In the former case, the registration message from link A and that from link B can be transmitted at different timings. Two messages from A, that is, a message for registering the correspondence between P (X) and URa and a message for registering the correspondence between P (X) and URb may be transmitted separately.

  In the case of sending using both links, the registration message A (correspondence of P (X) and URa) that has issued the UR is sucked into the nearest DR1 via the ISP-A, and the registration message B (P (X) and URb correspondence) is sucked into the nearest DR2 via ISP-B. Since DR1 and DR2 share registration information as described later, the correspondence between P (X) and URa and URb is registered in both DR1 and DR2. Each DR can generate a search table as shown in FIG. 3 using the registration information, information on a communication environment described later, and instruction information from a UR.

  In the case of sending using one link (for example, link A), a registration message (correspondence between P (X) and URa and URb) that has issued a UR is sucked into the nearest DR1 via the ISP-A. Then, the correspondence between P (X) and URa and URb is registered in DR1, and the same correspondence is registered in DR2 because DR1 and DR2 share registration information as described later. . Each DR can generate a search table as shown in FIG. 3 based on the registration information in the same manner as described above.

  One method of maintaining or deleting the correspondence information once registered in each DR as described above is to periodically transmit a registration message having the same contents from the UR (this method is called a soft state. is there). Then, the DR maintains the registered content as long as the registration message is received, and deletes the registered content when the registration message is not received (after a predetermined period has elapsed since the last time the registration message was received). For example, when a registration message indicating the correspondence between P (X) and URb is no longer received, this correspondence is deleted (only the correspondence between P (X) and URa remains as registration information). Then, the destination of the tunnel corresponding to the route passing through the link B is deleted from the search table of FIG. 3, and the search table is rewritten so that the tunnel for the remaining link A is selected.

  When the soft state is adopted, when one of the links (for example, link B) goes down in the method of transmitting the registration message using both of the above links, the registration message B does not reach the Internet. Therefore, the correspondence information between P (X) and URb is deleted from each DR, the search table of each DR indicates the path of the remaining link A, and the packet addressed to the user network is transmitted via the ISP-A. Since the data is transferred, connectivity with the Internet is maintained.

  In the case where the above-mentioned registration message transmission method using one of the links is adopted in the soft state, when the UR detects that one link (for example, the link B) is down, the UR sets P (X), URa, and URb. Is stopped as a registration message, and only the correspondence between P (X) and URa is transmitted as a registration message. Then, the correspondence information between P (X) and URb is deleted from each DR, and the search table of each DR points to the path of the remaining link A.

  As another method for maintaining or deleting registration information, a message for explicitly requesting deletion may be transmitted from the UR (this method may be called a hard state). Then, the DR keeps the registered contents until the deletion message is received, and deletes the registered contents when it is received. In the case of this method, even when both links go down or the UR itself goes down and communication becomes impossible, a delete message from the UR is not received, so that the registered contents are maintained until such a case. The DR periodically checks the existence of the UR so that the UR does not occur. For this confirmation, the existence confirmation by the protocol operating separately from the message exchange of the present embodiment may be shared.

  In the case of adopting the hard state, when one of the links (for example, link B) goes down in the method of sending the registration message using the above two links, the DR responds to the existence confirmation that is periodically transmitted to the URb by the DR. Will not return. Thereby, the correspondence information between P (X) and URb can be deleted from the DR, and the route indicated by the search table of each DR can be changed to the remaining link A.

  In the hard state, when the registration message transmission method using one of the above links is employed, when the UR detects that one link (for example, link B) has gone down, the UR determines the correspondence between P (X) and URb. Is sent to DR.

  In the above, an example has been described in which the pair of the user network prefix P (X) and the URa or URb is included in the registration message transmitted from the UR to the DR. However, the registration message includes the identifier Ix of the user network and the URa or A combination of URb and URb may be included, and a correspondence between P (X) and URa or URb may be registered by obtaining a prefix P (X) from Ix in the DR that receives the set. Not using the prefix P (X) as information included in the registration message from the UR has the following advantages.

  Since P (X) becomes routing information advertised to many other routers in the Internet, if there is an error in P (X), it is highly likely that a malfunction will occur over a wide range and the network will not function. In order to reduce such a risk, it is desirable to avoid using information declared by a UR, a user outside the Internet, as it is as routing information. As described above, the DR identifies the user network x in charge of itself by the identifier Ix. The UR declares Ix instead of P (X), and the DR converts Ix to P (X). By doing so, the risk of using the wrong P (X) can be reduced. As Ix, for example, an address of a UR in a user network (an address of a UR having P (X) as upper predetermined bits) can be used. In this case, the DR knows in advance the number of bits of P (X) of the user network in charge of itself, and obtains P (X) by extracting upper bits by the number of bits from the received Ix. As the protocol for transmitting the registration message, for example, Mobile IP standardized by the IETF can be applied. The UR is regarded as a mobile node, the DR is regarded as a home agent, and Ix in the registration message is represented as the home address of the UR. , URa or URb as a care-of address of the UR, the registration processing for the DR to transfer the packet addressed to P (X) to the URa or URb can be realized.

  Also, even when the UR reports P (X), the correct prefix P (X) of the user network in charge of itself is stored in advance in the DR receiving the registration message, and the registration message is stored in the DR. It is desirable to check whether the included P (X) matches the stored P (X). As a result of the check, if they match, the registration processing of the correspondence between P (X) and URa or URb is started, but if they do not match, an error message is returned to the UR of the registration message transmission source.

  As described above, the correspondence between the P (X) and the address of the UR is registered in at least one DR by the registration message issued by the UR. Next, this registration information is transmitted between the DRs. In the example shown in FIG. 7, a registration information server acting as a BGP route reflector is introduced, and the registration information server centrally manages connections between DRs for exchanging registration information. As a result, even when the registration information is transmitted to all DRs in full mesh, the load on each DR can be reduced, and network management can be facilitated. As another example, registration information can be propagated between DRs, such as a routing protocol called OSPF (Open Shortest Path First).

  Here, as described with reference to FIG. 5, in the case of (a) in which all DRs select a route for a UR, all the information transmitted between the DRs includes P (X) and a plurality of URs. Address. On the other hand, in the case of (b) in which the specific DR selects a route for the UR, the information transmitted between the DRs corresponds to a plurality of addresses of P (X) and the UR for the specific DR. , P (X) and the address of a specific DR for other DRs.

  By the message exchange described above, registration information (correspondence between P (X) and URa and URb) on which each DR creates a search table as shown in FIG. 3 is stored in each DR. Hereinafter, a method of determining which of the plurality of transfer paths to select for the DR will be described. Since the route to be selected determined here is written in the search table, the route is selected by referring to the search table at the time of packet transfer. When this search table is rewritten, the selected route is changed.

  When the traffic of each link and the communication quality (for example, delay, jitter, band, packet loss, etc.) of each path are detected as shown in FIG. And optimal route selection. The traffic of each link can be observed at the first and second interfaces of the UR, and only the sum of the traffic of all the packet flows from all the DRs may be observed, or for each specific packet flow. The traffic may be observed, or in the case of a tunnel by IP packet encapsulation or MPLS, the traffic may be observed for each DR at the entrance of the tunnel that can be known by the UR. When the UR observes traffic for each DR, it is possible to selectively notify a route change to a DR having a large traffic.

  Even if only the sum of the traffic is observed, for example, if the traffic of the link A is higher than that of the link B, the packet flow appropriately determined (for example, randomly) (for example, a set of a specific destination address and a source address) is determined. An instruction can be given to change the route (exit of the tunnel) from link A to link B. This instruction is realized by, for example, performing the processing up to the identification of the packet flow whose path is to be changed using the UR that has observed the traffic, and transmitting an instruction message including information specifying the identified packet flow from the UR to the DR. it can. As another example, the UR can send a communication environment notification message including the value of the observed traffic to the DR, and based on the information of this message, the DR can specify the packet flow whose route is to be changed.

  Further, if traffic is observed for each specific packet flow at the UR, for example, when the traffic of the link A is greater than that of the link B, if the route is changed for any packet flow, It can be determined whether the dispersion is optimal. In this case, the communication environment notification message method in which the packet flow whose path is changed is specified on the DR side may be used. Therefore, there is a possibility that an instruction message method in which the UR side specifies the packet flow to be route-changed is more suitable.

  The above-mentioned instruction message and communication environment notification message issued by the UR can be delivered to the nearest DR using anycast and transmitted between the DRs in the same manner as described for the registration message in FIG. In the case of a tunnel method in which the UR can know the addresses of the individual DRs, the UR may transmit the above message to a specific DR, and do not transmit the messages between the DRs.

  When an instruction message is sent using anycast, all the DRs or all the DRs in charge of the UR may change the path to be selected in the same manner, or the address of the DR to be changed in the instruction message. By including the list, it is also possible to change the route of only a specific DR. When a communication environment notification message is sent using anycast, since each DR makes an individual determination based on the observation result included in this message, there is a possibility that each DR will have different changes.

  If an instruction message is sent to a specific DR and transmission is not performed between DRs, in the case of a configuration in which all DRs select a route to a UR (FIG. 5A), even if the packet flow is the same, the DR passes. The route to be selected differs depending on the DR, and the load can be distributed accordingly. In the case of a configuration in which a specific DR selects a route to a UR (FIG. 5B), it is not necessary to send a message for changing the selected route to a DR other than the specific DR, but the UR is not connected to the entrance of the tunnel. If an instruction message or a communication environment notification message is sent to a specific DR (DR for route selection) by detecting an address, unnecessary message exchange can be omitted only by eliminating transmission between DRs. .

  By discriminating from which DR the traffic is sent from the traffic observation in the UR, it is possible to specify to which DR the above instruction message or communication environment notification message should be transmitted. Further, when the same packet flow is sent via a plurality of DRs, the above instruction message or communication environment notification message is transmitted only to some DRs, thereby requesting a path change only to some DRs. This also makes it possible to appropriately distribute the traffic on the link A and the link B.

  As an example of the communication quality of each route shown in FIG. 8, a description will be given taking delay as an example, but the present embodiment can be similarly applied to jitter, band, packet loss, and the like. One method is that, in the case of a tunnel method in which the UR can know the address of each DR, the UR observes the delay of a packet from a specific DR, and notifies the delay message indicating the traffic or the communication message. Like the environment notification message, it feeds back to the specific DR.

  Another method is applicable to any tunnel method. In addition to performing the packet transfer described with reference to FIGS. 2 and 6, the DR can measure the delay for a plurality of routes to the UR. A message is sent, and in response to this, delay information included in the message returned from the UR is obtained.

  When the DR receives the instruction message from the UR, the DR changes the route of the packet flow specified by the message as specified, and when the DR receives the message including the delay information from the UR, the DR itself changes the route. Identify the packet flow.

  As still another method, the administrator (operator) of the user network sets a route selection policy in the UR, such as sending a voice call (VoIP: Voice over IP) packet through a tunnel with a minimum delay. There may be a method of transmitting to the DR as an instruction message. For example, if the DR or the registration information server shown in FIG. 7 has a function as a Web server and the UR has a function as a Web client, the operator can also input a policy on a Web page to input the policy. it can. When the DR functions as a Web server, the DR receives an instruction message from the UR and transfers the content to another DR. When the registration information server of FIG. 7 functions as a Web server, the registration information server of FIG. 7 delivers the contents of the instruction message from the UR to another DR.

  The DR can select a route according to the user's request based on the instructed policy and the information on the delay of each route measured as described above. Specifically, for example, the destination of the tunnel with the minimum delay is stored in the search table of FIG. 3 corresponding to the search key indicating the packet flow of the voice call.

  FIG. 9 and FIG. 10 are functional block diagrams showing an example of the internal configuration of the DR device 100 and the UR device 200 according to the present embodiment, respectively. Here, an example will be described in which any of the above-described methods for determining the selection path can be handled. However, in practice, some of the blocks shown in the figure may be selected and implemented as needed. It should be noted that double arrows in the figure indicate the flow of packets, and other arrows indicate the flow of control information.

  In the DR device 100, the packet receiving unit 102 receives a packet transmitted from the source host, or in the case of FIG. 5B, a packet transferred from another DR, and the route selecting unit 104 receives the packet. Referring to the search table storage unit 124 using the information in the packet as a key, one of a plurality of routes to the destination user network is selected based on the search result, and the packet transmission unit 106 The packet is transferred to the selected route (the interface of the UR serving as the exit of the tunnel) by any one of the methods (1) to (c). In the case of the method of FIG. 4C, the address conversion unit 128 is used. In the case of FIG. 5B, the search table storage unit 124 may indicate another DR as a transfer destination. In this case, the packet transmission unit 106 transfers the packet to the DR.

  In the UR device 200, if a packet is transferred from the link A, the ISP-A packet receiving unit 202 receives the packet. If a packet is transferred from the link B, the ISP-B packet receiving unit 204 receives the packet. Is received, and the process on the UR side in any of FIGS. 4A to 4C is performed. In the case of FIG. 4C, the address conversion unit 206 performs address conversion with reference to the correspondence between P (X) and A and B stored in the registration information storage unit 210. Then, the packet transmitting unit 208 transfers the original packet generated from the received packet to the destination host.

  The registration information transmission unit 212 of the UR 200 transmits the correspondence between P (X) stored in the registration information storage unit 210 and URa and URb as a registration message to the DR 100 as described with reference to FIG. 108 receives this. When the communication environment observation unit 214 detects that any link is down, the registration information transmission unit 212 changes the content of the registration message to be transmitted as described above (including an example of transmitting a deletion message). ). The content of the registration message received by the DR 100 is stored in the registration information storage unit 112, and when there is a change as described above, the registration information update unit 110 rewrites the stored content.

  When adopting a configuration in which the correspondence between the network identifier Ix and URa and URb is sent instead of the correspondence between P (X) and URa and URb as the registration message, the registration information transmission unit 212 of the UR 200 stores the registration information in the registration information storage unit 210 A registration message is created using Ix instead of the stored P (X). Then, the registration information updating unit 110 of the DR 100 obtains P (X) by referring to the prefix storage unit 130 based on the received registration message Ix, and stores P (X) and URa in the registration information storage unit 112. And URb are stored. When the correspondence between P (X) and URa and URb is transmitted as a registration message, the registration information update unit 110 of the DR 100 refers to the prefix storage unit 130 based on P (X) of the received registration message. If the received P (X) is correct, the registration information can be stored, and if it is incorrect, an error message can be returned to the UR without registration.

  The communication environment observation unit 214 of the UR 200 monitors the packet reception units 202 and 204, and obtains information such as traffic volume and communication quality. Then, the observation information transmission unit 216 transmits the observed information to the DR as a communication environment notification message, and the UR observation reception unit 114 receives the information. Alternatively, based on the information observed by the communication environment observing unit 214, the instruction information generating unit 218 determines which packet flow path should be changed and where, and the instruction information transmitting unit 220 instructs the information including the determination result. The message is transmitted to the DR, and the UR instruction receiving unit 118 receives the message.

  The instruction information generation unit 218 determines a packet flow whose route is to be changed based on the policy input by the operator and the observed communication environment, and may use the specific determination result as the instruction information, The policy itself input by the operator may be used as the instruction information without making a determination based on the policy. In the former case, the search table updating unit 122 rewrites the contents of the search table storage unit 124 according to the contents of the instruction message received by the UR instruction receiving unit 118. In the latter case, the content of the instruction message received by the UR instruction receiving unit 118 is stored in the policy storage unit 120, and the search table updating unit 112 determines the route based on the policy and the output of the communication environment detecting unit 116. Determine the packet flow to be changed. When the DR functions as the Web server described above, the UR instruction receiving unit 118 has a Web server function.

  Even when the UR instruction receiving unit 118 does not receive the instruction message, the communication environment detecting unit 116 detects the communication environment information output from the UR observation receiving unit 114, and the search table updating unit 112 outputs the information of the communication environment detecting unit 116. , The packet flow whose route is to be changed can be determined and the search table can be rewritten. The communication environment detection unit 116 can use not only the information on the communication environment observed at the UR but also the information on the communication environment acquired by the DR exchanging messages with the UR.

  By referring to the rewritten search table storage unit 124, the route selected by the route selection unit 104 (the interface of the UR serving as the exit of the tunnel) is changed, and traffic distribution and selection of the optimal route are possible. It becomes. The DR device 100 includes a DR control information transmitting / receiving unit 126, and can exchange control information such as the content of the above-described instruction message, communication environment information, and registration information with another DR. That is, the DR control information transmission / reception unit 126 is connected to the search table update unit 122, and the information received by the DR device 100 from the UR via the reception units 108, 114, and 118 to another DR if necessary. After processing, transfer. Further, the DR control information transmitting / receiving unit 126 receives the above-described control information from another DR or the registration information server in FIG. 7, and the search table updating unit 122 receives the control information from the other DR or the like based on The contents of the search table storage unit 124 can be rewritten.

  FIG. 11 shows that the DR confirms whether or not each of the plurality of routes to the UR is operating normally, and sets a normally operating route as a route for transferring a packet addressed to P (X). 2 shows a mechanism for selecting. The DR in the figure has the same function as the DR 100 shown in FIG. 9, and the registration information storage unit 112 stores a path A using the link A by the ISP-A for the user network P (X). , Path B using link B by ISP-B are registered. This registration information is reflected in the DR search table, and the route selection unit 104 selects a route to the user network with reference to the search table. The DR in this example includes a confirmation packet transmission unit 132, a reply packet reception unit 134, and an interval control unit 136, in addition to the above.

  The acknowledgment packet transmitting unit 132 refers to the registration information storage unit 112, and periodically sends the route A (virtual tunnel whose exit is URa) and the route B (virtual tunnel whose exit is URb). Then, a packet called a keep-alive packet (test traffic different from actual traffic) is transmitted. The reply packet receiving unit 134 examines a packet returned for each of the transmitted keep-alive packets, and checks whether each of the tunnel from DR1 to URa and the tunnel from DR1 to URb are functioning effectively. Check. Then, it instructs the route selection unit 104 not to select a route that has been found to be not functioning effectively. For example, when instructed not to use the route A, the route selecting unit 104 refers to the search table, selects a route other than the route A (for example, the route B), and transfers a packet addressed to P (X).

  The interval control unit 136 can change the interval at which the confirmation packet transmitting unit 132 transmits a keep-alive packet to each route according to the communication quality (for example, band) of each route. If the transmission interval is short, it is possible to grasp the status of each route in real time, but if the bandwidth is narrow, frequent transmission of test packets such as keepalive can transmit packets of real traffic Since the band is further narrowed, the keep-alive transmission interval is controlled to be longer as the band becomes narrower. For example, if the bandwidth of the route A is larger than the bandwidth of the route B, the interval at which keepalives are transmitted to the route A is shorter than the interval at which keepalives are transmitted to the route B.

  By applying a mechanism known as pathchar, the communication quality of each of the above paths can be estimated from the contents of a reply to the above keep-alive packet. According to this mechanism, the address of the router on the route from DR1 to the tunnel exit (URa or URb) is returned as a response to the keep-alive packet, as in the case where DR1 designates the exit of the tunnel and executes traceroute. , DR1 in that order, and the time required to each router on the route can also be measured. This measurement is repeated many times while changing the size of the keep-alive packet to be transmitted. Then, by measuring the RTT (Round Trip Time) for each of the measurement packets having different sizes and obtaining the regression curve, the band of the tunnel can be estimated. The bandwidth can be estimated from the slope of the regression curve, and the propagation delay can be estimated from the RTT when the packet size indicated by the regression curve is 0.

  In the above description, an example has been described in which the operation of the tunnel from the DR to the UR (downstream direction) is checked, and a path in which the DR is operating normally is selected to transfer the packet to the user network. Is also applicable when using a (upstream) tunnel from the UR to the DR for packets flowing from the user network to the Internet. That is, the UR is composed of a route from the interface of URa to DR (shown by a dotted line next to the route A in FIG. 11) and a route from the interface of URb to DR (shown by a dotted line next to the route B in FIG. 11). For each of them, a keepalive packet is transmitted, the response is checked, a normally operating tunnel is selected, and the packet leaving the user network is forwarded. The same applies to the case where the keepalive transmission interval can be changed according to the communication quality of the route.

  In the conventional system shown in FIG. 1, it is necessary for a network administrator to operate BGP (protocol for the Internet) which is not normally handled in a corporate network or the like, and to advertise an address to the Internet. According to this, even without introducing such a BGP, a multi-home connection can be realized by installing a UR having a function of transmitting the above-described registration message in a corporate network or the like.

  Although the embodiments of the present invention have been described above, it is needless to say that those skilled in the art can variously modify and apply the above embodiments within the scope of the present invention. For example, in the above description, a case where one user network is provided with a link from each of two ISPs has been described. However, even when one user network is provided with a link from each of three or more ISPs, The same multi-home connection, traffic distribution, and the like are possible. In the above description, one ISP provides one link to the user network (two links are provided by two ISPs). However, two or more links are provided by one ISP to the user network. (For example, the network of ISP-A and ISP-B in FIG. 2 is operated by one ISP, and two links A and B are provided to the UR from the one ISP network). Home connection, traffic distribution, etc. are possible.

  As described above, if the router and the user router according to the present invention are used, for example, a route can be selected so that traffic of a plurality of links is appropriately distributed, and a communication quality suitable for the property of data carried by a packet can be selected. Multi-home connection to the Internet is possible, such as selecting a route provided by the Internet, selecting a route that is operating normally, and selecting a route to meet the needs of the user network. Useful for

FIG. 3 is a diagram for explaining a conventional multihome connection method. FIG. 4 is a diagram for explaining an example of packet transfer according to an embodiment of the present invention. FIG. 4 is a diagram illustrating an example of a search table used for packet transfer according to the embodiment. In the present embodiment, a path through which a packet is transferred is configured by (a) encapsulation of an IP packet, (b) by a label switching path, and (c) by a destination address conversion of an IP packet. FIG. Of the router devices DR according to the present embodiment that exist on the Internet, (a) all DRs transfer to the user network, and (b) some DRs transfer to the user network. FIG. FIG. 7 is a diagram for explaining another example of packet transfer in the embodiment. The figure which illustrates a mode that the message for registering the information required for control in the router device DR concerning this embodiment is exchanged. FIG. 4 is a diagram for explaining an example of a communication environment detected in the embodiment. FIG. 2 is a diagram illustrating an example of an internal configuration of a router device DR according to the embodiment. FIG. 3 is a diagram showing an example of an internal configuration of a user router device UR according to the embodiment. FIG. 4 is a diagram illustrating a configuration example of a router device having a function of checking whether a path (virtual tunnel) through which a packet is transferred in the present embodiment is operating normally, and a diagram for explaining the function;

Explanation of reference numerals

100 Router device DR according to this embodiment
102 packet reception unit 104 route selection unit 106 packet transmission unit 108 UR registration reception unit 110 registration information update unit 112 registration information storage unit 114 UR observation reception unit 116 communication environment detection unit 118 UR instruction reception unit 120 policy storage unit 122 search table update Unit 124 search table storage unit 126 DR control information transmission / reception unit 130 prefix storage unit 132 confirmation packet transmission unit 134 reply packet reception unit 136 transmission interval control unit 200 user router device UR according to the present embodiment
202 Packet receiving unit from ISP-A 204 Packet receiving unit 128, 206 from ISP-B Address conversion unit (optional)
208 Packet transmission unit 210 Registration information storage unit 212 Registration information transmission unit 214 Communication environment observation unit 216 Observation information transmission unit 218 Instruction information generation unit 220 Instruction information transmission unit

Claims (28)

A method of controlling packet transfer in a router constituting the Internet,
Receiving a packet destined for a user network provided with a plurality of connections to the Internet,
Selecting one of a plurality of connections provided to the user network based on the packet including the address of the user network;
A packet transfer control method, comprising: transferring the received packet to a path from the router to the user network according to an address corresponding to the selected connection. In the destination address field of the received packet, the address of the user network is written, and in the source field of the packet, the address of the source network is written, respectively.
The transfer of the received packet is performed by encapsulating the packet using the address assigned by the provider that provides the selected connection as the destination and the address of the router as the source, respectively. 2. The packet transfer control method according to claim 1, further comprising: Label switching in which a packet is transferred based on a label, which is lower layer information than a network layer, along a route to the user network by using one of the plurality of connections starting from the router. The path is set,
2. The packet transfer control method according to claim 1, wherein the transfer of the received packet is performed by giving a label of a label switching path corresponding to the selected connection to the packet. In the destination address field of the received packet, the address of the user network is written, and in the source field of the packet, the address of the source network is written, respectively.
2. The transfer of the received packet according to claim 1, wherein the transfer is performed by rewriting the destination address field based on an address assigned by the provider providing the selected connection in accordance with the connection. Packet transfer control method.   The selection is performed by referring to information of a layer higher than a network layer among information in the received packet, and part of a group of packets received by the router destined for the user network, of the plurality of connections. One of the plurality of connections is transferred on the path to the user network, and the other part is transferred on the path to the user network by using the other one of the plurality of connections. 2. The packet transfer control method according to claim 1, wherein the control is performed in such a manner.   The selecting selects which of the plurality of connections the packet should be forwarded to the user network based on a communication environment provided to the user network using each connection. 2. The packet transfer control method according to claim 1, wherein: The communication quality of a packet transfer in a route to the user network by using a certain one of the plurality of connections from the router as a starting point, and the other of the plurality of connections from the router as a starting point. Measure the communication quality of packet transfer on the path to the user network by using one,
7. The packet transfer control method according to claim 6, wherein the selection is performed by using each measured communication quality as a communication environment provided to the user network using each of the connections. Packet traffic observed by the user network between the Internet and the user network over one of the plurality of connections and between the Internet and the user network over the other one of the plurality of connections; Receiving information about the packet traffic of
7. The packet transfer control method according to claim 6, wherein the selection is performed by using the received information on each traffic as a communication environment provided to the user network using each connection. For each connection provided to the user network, using the connection, a route in which the transfer from the router to the user network is performed, confirms whether the network is operating normally,
2. The packet transfer control method according to claim 1, wherein the selection is performed such that the received packet is transferred to a path confirmed to be operating normally. For each connection provided to the user network, using the connection, a route in which the transfer from the router to the user network is performed, confirms whether the network is operating normally,
2. The packet transfer control method according to claim 1, wherein in the selection, control is performed so as not to select a connection corresponding to a path determined to be not operating normally by the confirmation. The confirmation is performed by transmitting a confirmation message from the router to each of the connections at predetermined intervals,
11. The packet transfer control method according to claim 9, wherein an interval at which the confirmation message is transmitted is changed based on communication quality of a path on which transfer is performed using each connection. Which part of the packet group destined for the user network, determined by the user network, based on the communication environment provided to the user network using the respective connections observed by the user network, Receiving information indicating which one of a plurality of connections should be used for transfer,
2. The packet transfer control method according to claim 1, wherein the selection is performed according to received instruction information. Which part of the group of packets destined for the user network determined by the user network should be forwarded to the user network by using a connection that provides what communication environment to the user network Receiving the information indicating
The selection is performed by comparing received instruction information with information on a communication environment provided to the user network using each connection obtained separately from the instruction information, and selecting one of the plurality of connections. 2. The packet transfer control method according to claim 1, wherein the method is performed by specifying the following. Storing information indicating a correspondence between the address of the user network and an address assigned corresponding to each connection by a provider that provides each connection;
The selection is performed based on the stored information,
2. The packet transfer control method according to claim 1, wherein the stored information or information indicating a correspondence between the address of the user network and the address of the router is transmitted to another router. Receiving, from the user network, identification information of the user network and information on an address assigned to each connection by a provider that provides each connection;
Calculating an address of the user network from the received identification information;
Storing information indicating a correspondence between the address of the user network and an address assigned corresponding to each connection by a provider that provides each connection;
2. The packet transfer control method according to claim 1, wherein the selection is performed based on the stored information. A method for controlling packet transfer in a user router provided at a connection point with the Internet, in a user network provided with a plurality of connections to the Internet,
A router provided in the Internet and having a function of selecting which connection among a plurality of connections provided to the user network should be used to transfer a packet. Transmitting identification information and information on an address assigned to the user router among addresses assigned corresponding to each of the plurality of connections;
A packet transfer control method, comprising: receiving a packet addressed to the user network transferred through a route according to an address corresponding to a connection provided to the user router among the plurality of connections.   17. The packet transfer according to claim 16, wherein the transmitted identification information of the user network is address information of the user network or information that can be converted into an address of the user network by a router having the function. Control method.   Observing a communication environment provided to the user network via a connection provided to the user router among the plurality of connections, and transmitting information obtained by the observation to a router having the function. 17. The packet transfer control method according to claim 16, wherein:   Based on the observation information of the communication environment provided to the user network via each of the plurality of connections, a part of a packet group destined for the user network is assigned to any one of the plurality of connections. 17. The packet transfer control method according to claim 16, wherein information indicating whether transfer is to be performed through a path according to a corresponding address is generated, and the instruction information is transmitted to a router having the function.   Generating information indicating which part of a group of packets destined for the user network should be transferred to the user network by using a connection that provides a communication environment to the user network; 17. The packet transfer control method according to claim 16, wherein said instruction information is transmitted to a router having said function.   An address specifying a router having the function is obtained based on the received packet, and a communication environment provided to the user network through a connection provided to the user router among the plurality of connections is obtained. 17. The packet transfer control method according to claim 16, wherein the received information or information representing an instruction generated based on the communication environment is transmitted to the router specified by the obtained address. An address corresponding to the selected connection is written in a destination address field of the received packet,
17. The packet transfer control method according to claim 16, wherein the destination address field is rewritten based on an address of the user network, and the received packet is transferred to a node in the user network. The transmission is to transmit, using one of the plurality of connections, identification information of the user network and information on an address assigned corresponding to each connection,
17. The method according to claim 16, wherein when at least one of the plurality of connections is disconnected, transmission of information on an address assigned to the at least one other connection is stopped. Packet transfer control method.   The transmission of the identification information of the user network and the information on the address assigned to the user router is performed using a connection corresponding to the address assigned to the user router among the plurality of connections. Item 18. A packet transfer control method according to Item 16. A program for controlling transfer of packets, which is executed in a computer functioning as a router configuring the Internet,
When the router receives a packet including a destination of a user network address provided with a plurality of connections to the Internet, the router selects one of the plurality of connections provided to the user network based on the packet. ,
A packet transfer control program for causing the router to transfer the received packet to a path on which the packet is transferred to the user network according to an address corresponding to the selected connection. A program for controlling the transfer of packets, which is executed on a computer functioning as a user router provided at a connection point with the Internet, of a user network provided with a plurality of connections to the Internet,
The user router may be a router provided in the Internet and having a function of selecting which connection among a plurality of connections provided to the user network should be used to transfer a packet. Transmitting the identification information of the user network and the information on the address assigned to the user router among the addresses assigned corresponding to each of the plurality of connections;
A packet transfer control program for causing a user router to receive a packet addressed to the user network transferred through a route according to an address corresponding to a connection provided to the user router among the plurality of connections. Receiving means for receiving a packet destined for a user network provided with a plurality of connections to the Internet;
Storage means for storing an address corresponding to each of the plurality of connections, corresponding to the address of the user network;
Control means for determining, based on the packet received by the receiving means, which of the plurality of addresses stored in the storage means should be transferred by a path according to any of the addresses;
A router for transferring the received packet to a route to the user network selected by the control unit. A storage means for storing at least one address assigned to at least one of the plurality of connections to its own device belonging to a user network provided with a plurality of connections to the Internet,
A router provided in the Internet, the router having a function of selecting which connection of the plurality of connections should be used to transfer a packet. Transmitting means for transmitting information about the address assigned to the
Receiving means for receiving a packet addressed to the user network transferred by a route according to an address corresponding to a connection provided to the own device among the plurality of connections;
Transfer means for transferring a packet received by the receiving means to a node in the user network.

Images