JP5022412B2 - Route information management system, route information management method, and program - Google Patents

Description

  The present invention relates to a technique for distributing and holding inter-domain route information in an AS (Autonomous System) in an IP (Internet Protocol) network.

  An IP network such as the Internet is formed by connecting a number of ASs operated by different organizations. The ASs are connected by boundary routers of each AS. Since the AS needs to transfer a packet destined for the outside of the AS, the Internet route information is exchanged between the ASs using BGP (Border Gateway Protocol). All routers in the AS must hold all external route information outside the AS. Specifically, the AS boundary routers can forward external route information (a set of IP addresses) to the AS external destination prefix (a set of IP addresses) by eBGP (external BGP). Information on whether to reach the destination) is exchanged. The external route information is advertised to the router in the AS by iBGP (internal BGP). However, in order to advertise between routers using iBGP, the routers need to be connected to each other by IGP (Interior Gateway Protocol).

  The router that has received the advertised external route information indicates, based on the received external route information, a transfer route that indicates which border router can be reached for each destination prefix, that is, an optimal transfer route. Was calculated and held. Therefore, the router has to store external route information corresponding to the number of destination prefixes and transfer route information calculated by route calculation. Also, the router needs to advertise external route information corresponding to the number of destination prefixes as a message.

  In recent years, with the rapid expansion of the Internet, the number of destination prefixes existing on the Internet has increased. Therefore, network operators such as Internet service providers have to deal with expansion of storage capacity of route information to be retained and increase of processing load of route calculation in routers etc. that are network equipment. Increased cost burden. In addition, the increase in the number of routes to be held per router, the increase in the number of messages that need to be transmitted, and the increase in the processing load of route calculation are as follows. This may increase the time until the determination, which may greatly affect the stability of network control.

  In the prior art, iBGP advertises using iBGP by introducing a route reflector (RR) (see Non-Patent Document 1) or AS confederation (see Non-Patent Document 2) in order to reduce the processing load of the router. The number of sessions is being reduced. The route reflector (RR) transfers external route information only between route reflector clients that have established peers, and does not transfer between route reflector clients that have not established peers. Therefore, the route reflector (RR) can reduce the number of peers to be established in the AS compared to a case where each router in the AS is connected with a full mesh using iBGP. Also, in AS confederation, the number of peers can be reduced as compared to the case of full mesh by configuring several sub-ASs in the AS. As described above, in the route reflector (RR) or AS confederation, the number of peers can be reduced, so that an increase in processing load on the router can be suppressed.

  In order to reduce the storage capacity of the router, Compact Routing (see Non-Patent Documents 3 and 4) has been proposed. Compact routing is a technology that transfers packets once to a router that becomes an aggregation point of routes, and transfers the packets from the router that becomes the aggregation point to a destination address. In other words, Compact Routing reduces the storage capacity required for holding route information of each router by aggregating routes.

T. Bates, E. Chen, and R. Chandra, “BGP Route Reflection: An Alternative to Full Mesh Internal BGP (IBGP)”, RFC 4456, IETF, April 2006 P.Traina, D.McPherson, and J.Scudder, “Autonomous System Confederations for BGP”, RFC 5065, IETF, August 2007 D. Krioukov, kc claffy, K. Fall, and A. Brady, “On Compact Routing for the Internet”, SIGCOMM Comput. Commun. Rev., Vol.37, No.3, July 2007, p.41- 52 P.Francis, X.Xu, and H.Ballani, “FIB Suppression with Virtual Aggregation”, internet Draft, draft-ietf-grow-va-00.txt, May 2009

  However, in the method using the route reflector (RR) described in Non-Patent Document 1 and the method using AS confederation described in Non-Patent Document 2, the iBGP topology is not a full mesh, and external route information is exchanged. In the iBGP session, only one route information is advertised for each destination prefix (hereinafter also referred to as a prefix). Therefore, the number of iBGP sessions is reduced, and a certain degree of effect is recognized in reducing the processing load of the router and the storage capacity of the path information. However, since the router stores only a part of route information, route calculation is performed based on incomplete external route information, so the result of route calculation is not necessarily optimal, and the calculated route is not a loop. As a result, there is a problem that it may not converge.

  In addition, in the method using Compact Routing described in Non-Patent Documents 3 and 4, since packets are once collected in a router as a route aggregation point, the transfer route of packets is not necessarily optimal (shortest), and network resources are reduced. The problem of uselessness occurs.

  Therefore, an object of the present invention is to solve the above-mentioned problem, that is, satisfying that the route calculation is performed based on all the route information and the optimum transfer route is calculated, and the future expansion of the Internet. It is an object of the present invention to provide a technology that suppresses an increase in the number of routes (storage capacity) to be held per router, an increase in the number of messages to be transmitted, and an increase in processing load as a new routing architecture that can be supported.

  The present invention obtains external route information including a prefix, which is a set of IP addresses of other ASs, from other ASs in an AS (Autonomous System) that constitutes an IP (Internet Protocol) network, and other packets. Information management in which a plurality of routers that transfer to the AS and two or more route information management devices that collect external route information including the same prefix from the router and perform route calculation are connected to each other so as to be able to communicate with each other In the system, the path information management device uses the external path information including the prefix for each prefix collected from the router, and uses the external path information including the prefix as a unit of a group of two or more routers in the AS. Transfer route to the packet transfer destination prefix in at least one of the routers belonging to the set of routers Select a predetermined number of routers belonging to the router set that transmits the route calculation result information including the route calculation result for each router set calculated by the route calculation unit and the route calculation unit that executes the route calculation A storage destination selection unit, and an input / output unit that transmits the route calculation result information to the predetermined number of routers selected by the storage destination selection unit, and the router has received a packet to be transferred When referring to the route calculation result information received from the route information management device, the transfer route determination unit for determining the transfer destination of the packet, and the transfer to the transfer destination determined by the transfer route determination unit And an input / output unit for transmitting a power packet.

  According to such a configuration, since the route information management apparatus stores the external route information related to the prefix, the number of routes (storage capacity) to be held in the router can be reduced. Further, since the route information management apparatus collects all the external route information related to the same prefix and executes route calculation, there is no problem that the route calculation result is not optimal. In addition, when a plurality of router sets are provided in the AS, the BGP route in the router set is common to the prefix, so the route information management device calculates a route for only one router in the router set. If you run Therefore, the route information management apparatus does not need to perform route calculation for all routers, and can reduce the processing load. In addition, since the router does not need to perform route calculation, the processing load can be reduced. Furthermore, since the route calculation result for each router set calculated by the route information management device is transmitted only to a predetermined number of routers in the router set, the number of messages when the route calculation result is transmitted as a message is reduced. be able to.

  In the present invention, the route information management device is further generated by a hash function as a router ID of at least one router for each router set after a bit string of the router set ID of the router set to which the router belongs. A storage unit for storing storage destination node information for storing a bit string obtained by concatenating a bit string of hash values, and a router set of the router set as a search key used for searching the storage destination of the route calculation result information for each of the router sets An ID bit string, the prefix bit string, a bit string of a mask value for the prefix, and a search key generation unit that generates a bit string obtained by concatenating all the bit strings of “0” as padding, and the storage destination selection unit includes: Distance between the router ID stored in the storage node information and the search key Is calculated as the numerical value of the number of bits counted from the right for the bit position where “1” first appears from the left in the bit string indicating the exclusive OR for each bit of the router ID and the search key. Then, a predetermined number of routers are selected as destinations from the set of routers in ascending order of the calculated distance.

  According to such a configuration, the bit string of the router ID includes the router set ID and the hash value, the bit string of the search key includes the router set ID and the prefix, and the distance between the router ID and the search key is in ascending order. A predetermined number of routers are selected. Therefore, a router having a hash value with a matching router set ID and a distance close to the prefix is selected. Therefore, the storage destination routers for the selected route calculation result information are almost equalized. That is, the storage destination routers are selected almost evenly in the router set, and the router storage capacity can be averaged. In other words, the storage capacity of one router can be reduced as compared with the storage capacity of one conventional router that stores all route calculation result information.

  The present invention is characterized in that the route calculation result information is information in which the search key is associated with a result of route calculation for each router set corresponding to a prefix included in the search key.

  According to such a configuration, the route calculation result information is information in which the search key is associated with the result of route calculation for each router set corresponding to the prefix included in the search key. Therefore, the route calculation result for each router set corresponding to the prefix can be stored in the router having the node ID close to the prefix of the search key using the search key as a medium. Accordingly, when searching for route calculation result information related to a prefix, if a search key including the prefix is used, a router having a node ID close to the search key and a router corresponding to the prefix stored in the router The result of route calculation for each set can be acquired. That is, the route calculation for each set of routers corresponding to the prefix included in the search key is performed using the search key from the route calculation result information stored in a distributed manner to reduce the storage capacity of the router per unit. It becomes possible to search the result.

  In the present invention, the router further stores acquisition destination node information for storing router IDs of at least one router excluding itself in the router set to which the router belongs, and the route calculation result information, When a packet to be transferred is received, a bit string of a router set ID to which the router belongs, a bit string of a destination IP address of the packet, a bit string of all “1” as a mask value for the destination IP address, and all “0” as padding An acquisition key generation unit that generates an acquisition key, formed by concatenating the bit strings, a router ID stored in the acquisition destination node information, and a distance between the acquisition key, the router ID, and the acquisition key In the bit string indicating the exclusive OR of each bit, the bit position where “1” first appears from the left is shown. An acquisition destination selection unit that calculates a numerical value of the number of bits counted from the right, and selects a predetermined number of routers as acquisition destinations in ascending order of the calculated distance, and the transfer path determination unit includes Determining whether or not the search key of the route calculation result information stored in the storage unit and the acquisition key match, and if there is a match, the router associated with the search key Based on the route calculation results for each set, determine the transfer destination of the packet, and if there is no match, send the acquisition key to the router selected as the acquisition destination by the acquisition destination selection unit, Obtaining a route calculation result for each router set in which the search key and the acquisition key of the route calculation result information stored in the router match, and obtaining the route calculation for the obtained router set And determining a forwarding destination of the packet based on the results.

  Further, according to the present invention, when the transfer route determination unit further has no match between the search key of the route calculation result information and the acquisition key, the last non-zero bit of the destination IP address portion of the acquisition key Is changed to zero, the right bit including the change position of the destination IP address portion of the mask value portion of the acquisition key is changed to zero, the changed bit string is set as a new acquisition key, and the new acquisition key is set. Using the acquisition key, the acquisition destination router selected by the acquisition destination selection unit of the acquisition destination router and the route calculation result information selected from the acquisition destination router selected by the acquisition destination selection unit of itself Repeatedly acquiring, determining whether or not the search key of all the acquired route calculation result information matches the new acquisition key, and the route calculation result All the bit strings of the destination IP address portion of the acquisition key determine the forwarding destination of the packet based on the result of route calculation for each router set when the search key of the information and the acquisition key have the longest match It is characterized by being repeatedly executed until it reaches zero.

  According to such a configuration, when the router receives a packet to be transferred, first, the router calculates the route calculation result information in which the search key and the acquisition key of the route calculation result information stored in its storage unit match. Extract. If there is no match, search the route calculation result information stored in another router, and extract the route calculation result information in which the search key and the acquisition key of the route calculation result information match. Can do. That is, it becomes possible to search the route calculation result information distributedly managed in order to suppress the storage capacity of each router by checking the acquisition key and the search key. In addition, as a process when the route calculation result information in which the search key and the acquisition key match is not found, the search key and the acquisition key are reduced by shortening the effective digit of the destination IP address portion of the acquisition key. Can be extracted from itself and other routers. Therefore, even if the router itself that receives the packet to be transferred does not store the necessary route calculation result information, the route calculation result information more suitable for transfer can be acquired from other routers. That is, since the route calculation result information is distributed and managed by the routers in the router set, the storage capacity of each router is the storage capacity of one conventional router that stores all the route calculation result information. Compared to, it can be reduced.

  In the present invention, the acquisition destination node information divides the router ID into a plurality of ranges, and sets the range of the division narrower as the router ID is closer to the distance from the router ID of its own router. The range of the division is set wider as the ID, the router ID of the router belonging to the range is stored for each of the ranges, and the acquisition destination selection unit sets the acquisition key and the router ID of the boundary of the range. In comparison, the range including the acquisition key is extracted, the distance between the acquisition key and a router ID of a router belonging to the extracted range is calculated, and a predetermined number of routers are acquired in the order of the closest distance. It is characterized by selecting.

  According to such a configuration, since the router ID is divided into a plurality of ranges and stored, it is only necessary to calculate the distance between the router ID belonging to the range including the acquisition key and the acquisition key. Therefore, it is not necessary to calculate the distance between all the router IDs stored in the router and the acquisition key, and the calculation processing amount (processing load) can be reduced.

  The present invention is a route information management method for a route information management device used in the route information management system, wherein the route information management device has a group of two or more routers in the AS as a unit. For each set of multiple routers, a bit string obtained by concatenating a bit string of a hash value generated by a hash function after a bit string of a router set ID of the router set to which the router belongs, as a router ID of at least a predetermined number of routers. A storage unit for storing storage destination node information to be stored, and for each of the prefixes collected from the router, the external route information including the prefix is used as a unit of a group of two or more routers in the AS. In at least one of the routers belonging to a set of routers set in the AS A route calculation step for performing a route calculation for calculating a transfer route to a prefix of a packet transfer destination, and a route calculation including a route calculation result for each router set calculated by the route calculation step for each router set As a search key used for searching the storage destination of the result information, a bit string of a router set ID of the router set, a bit string of the prefix, a bit string of a mask value for the prefix, and a bit string obtained by concatenating all bit strings of “0” as padding In the bit string indicating the exclusive OR for each bit of the router ID and the search key, the search key generation step to generate, the distance between the router ID stored in the storage node information and the search key, Number from the right for the bit position where "1" first appears from the left A storage destination selection step of calculating a numerical value of the number of digits of the bits, and selecting a predetermined number of routers as a transmission destination from the router set in ascending order of the calculated distance; and the storage location selection step of the route calculation result information And an input / output step for transmitting to the predetermined number of routers selected by

  The present invention is also a route information management method for routers used in the route information management system, wherein a plurality of routers are set in the AS in units of a group of two or more routers in the AS. For each router set, as a router ID of at least a predetermined number of routers, an acquisition destination for storing a bit string obtained by concatenating a bit string of a hash value generated by a hash function after a bit string of a router set ID of the router set to which the router belongs A storage unit for storing node information, route calculation result information indicating information relating a search key transmitted from the route information management device and a route calculation result for each router set, and a packet to be transferred When received, the bit string of the router set ID to which the router belongs and the destination IP address of the packet An acquisition key generating step for generating an acquisition key, which is formed by concatenating a bit string of all “1” as a mask value for the destination IP address and a bit string of all “0” as padding, and the acquisition destination node In the bit string indicating the exclusive OR for each bit of the router ID and the acquisition key, the distance between the router ID stored in the information and the acquisition key first appears as “1” when viewed from the left. The bit position is calculated as a numerical value of the number of bits counted from the right, and an acquisition destination selection step for selecting a predetermined number of routers as acquisition destinations from the router set in ascending order of the calculated distance and a packet to be transferred are received. When the search key of the route calculation result information stored in the storage unit and the acquisition key match, If there is a match, the packet transfer destination is determined based on the result of route calculation for each router set associated with the search key. If there is no match, the acquisition destination selection unit The acquisition key is transmitted to the router selected as the acquisition destination, and the route calculation result for each router set in which the search key of the route calculation result information stored in the router matches the acquisition key is acquired. A transfer route determination step for determining a packet transfer destination based on the obtained route calculation result for each router set; and an input / output step for transferring the packet to the transfer destination determined in the transfer route determination step; It is characterized by performing.

  According to such a configuration, since the route information management apparatus stores the external route information related to the prefix, the number of routes (storage capacity) to be held in the router can be reduced. Further, since the route information management apparatus collects all the external route information related to the same prefix and executes route calculation, there is no problem that the route calculation result is not optimal. In addition, when a plurality of router sets are provided in the AS, the BGP route in the router set is determined in common for the prefix, so the route information management device calculates a route for only one router in the router set. If you run Therefore, the processing load of the route information management device can be reduced and route calculation at the router becomes unnecessary. Further, since the result of route calculation for each router set calculated by the route information management device is distributed and managed in the router set, it may be stored (transmitted) in a predetermined number of routers for each router set. Therefore, it is possible to reduce the number of messages when the route calculation result is transmitted as a message. Further, when receiving a packet to be transferred, the router first extracts the route calculation result information in which the search key and the acquisition key of the route calculation result information stored in its storage unit match. If there is no match, search the route calculation result information stored in another router, and extract the route calculation result information in which the search key and the acquisition key of the route calculation result information match. Can do. That is, it becomes possible to search the route calculation result information distributedly managed in order to suppress the storage capacity of each router by checking the acquisition key and the search key. Therefore, even if the router itself that receives the packet to be transferred does not store the necessary route calculation result information, the route calculation result information more suitable for transfer can be acquired from other routers. That is, since the route calculation result information is distributed and managed by the routers in the router set, the storage capacity of each router is the storage capacity of one conventional router that stores all the route calculation result information. Compared to, it can be reduced.

  In the present invention, the route information management method is a program for causing a route information management device or router as a computer to execute.

  A computer in which such a program is installed can realize functions based on this program.

  According to the present invention, an increase in the number of routes (storage capacity) to be held per router and transmission can be achieved while satisfying that route calculation is performed based on all route information to calculate an optimum transfer route. It is possible to provide a technique for suppressing an increase in the number of messages to be processed and an increase in processing load.

It is a figure showing the outline | summary of this embodiment. It is a figure which shows an example of the function of the route server in this embodiment. It is a figure which shows an example of the function of the router in this embodiment. It is a figure which shows the example of the determination method of a route server. It is a figure which shows the definition of the distance in this embodiment. It is a figure which shows an example of the storage destination node DB of a route server. It is a figure which shows an example of the acquisition destination node DB of a router. It is a figure which shows an example of route calculation result DB of a router. It is a figure which shows the flow of the storage process of the route calculation result in this embodiment. It is a figure which shows the flow of a process which acquires transfer path information from the other router in this embodiment. It is a figure which shows operation at the time of the bit change of the acquisition key kd.

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

≪Overview of this embodiment≫
First, an outline of the present embodiment will be described with reference to FIG. As shown in FIG. 1, the route information management system 30 includes in AS1 two or more route servers (route information management devices) 10A and 10B (10) and border routers that collect external route information and manage routing. A plurality of routers 20 (20A, 20B,..., 20L) are included. Each route server 10 is communicably connected to the router 20. The routers 20 are connected so that they can communicate with each other. Each router 20 belongs to one of the PoPs (Point of Presence) indicated by a one-dot chain line. PoP represents a set of routers (router set) existing in the same location such as the same city or the same building. In FIG. 1, AS1 is composed of PoP1, PoP2, PoP3, and PoP4. However, the number of PoPs is not limited to four as long as the number of PoPs is two or more. Further, the number of route servers 10 and the number of routers 20 are not limited to the numbers shown in FIG. The connection between the routers 20 is not limited to the solid line shown in FIG.

Here, in the route information management system 30, three features provided for suppressing the storage capacity of the router 20, the number of messages that need to be transmitted, and the processing load will be described.
(Characteristic 1) The external route information related to the prefix is collected in the route server 10.
(Characteristic 2) The route server 10 performs route calculation related to the prefix for each PoP, and stores (transmits) the result of the route calculation in some routers 20 belonging to the PoP for each PoP.
(Feature 3) Within PoP, route calculation results are distributed and managed between routers 20 belonging to the PoP, and route calculation results necessary for packet transfer are mutually acquired.

(About feature 1)
As shown in FIG. 1, it is assumed that the router 20A in the AS1 has received external route information related to the prefix P1 from the AS2 by eBGP. Then, the external route information related to the received prefix P1 is collected by the route server 10A. Further, it is assumed that the router 20K receives the external route information related to the prefix P7 from the AS 3 by eBGP. Then, the external route information related to the received prefix P7 is collected by the route server 10B. That is, each of the route servers 10A and 10B collects external route information related to different prefixes. Note that which route server 10 collects the external route information related to which prefix is predetermined. The method for determining the route server that collects the prefix will be described later.

  In this way, the route server 10 can collect all the external route information for any of the prefixes, so that the route calculation can be performed based on all the external route information. In other words, the route calculation is executed based on some external route information as seen in the method using the route reflector (RR) and the method using AS confederation described in the background section. There is no case where the route calculation result is not optimal or the route calculation does not converge. Further, since each route server 10 is responsible for distributing external route information for each prefix, by increasing the number of route servers 10 corresponding to the increase in the number of routers 20 in the AS 1, one route server 10 is provided. It is possible to suppress an increase in storage capacity per unit and an increase in processing load.

  In addition, the conventional border router includes an RIB (Routing Information Base) that stores a reception list and a transmission list of external route information, and an FIB (Forwarding Information Base) that stores a packet transfer destination. However, in the border router 20 according to the present embodiment, the reception list and the transmission list of the external route information are aggregated in the route server 10, so that RIB is unnecessary. That is, the number of paths (storage capacity) to be held in the router 20 can be reduced.

(About feature 2)
The route server 10 may perform route calculation for one router in the PoP for each prefix. The reason for this is as follows. That is, normally, the IGP link cost of the link spanned between the PoPs in the AS 1 is set to a value larger than the link cost in the PoP. In the route calculation, the route having the smallest value obtained by summing up the link costs assigned to the links constituting the route is preferentially selected. In this way, packets destined for the same PoP are prevented from going out of the PoP, and useless transfer such as going back and forth between PoPs is prevented. Therefore, routers in the same PoP take the same route (next hop) for BGP routes destined for the same prefix. That is, for all routers in PoP, the AS external transfer path is common. Therefore, route calculation with the same prefix as the destination need only be performed for one router in the PoP. Hereinafter, this is referred to as “route calculation for each PoP”. The route calculation for each PoP is performed by the route server 10 instead of the route calculation in the router 20, and a route calculation method using a known link cost can be used. However, the route calculation for each PoP is different from the known route calculation method in that the route server 10 does not execute the route calculation from itself but starts from one router 20 in the PoP. Only. Therefore, it can be said that the route calculation for each PoP has a reduced processing load compared to the conventional method in which the route calculation is performed for all the routers 20 in the AS 1.

  The route server 10 stores (transmits) the route calculation result in a part of routers of each PoP (dotted line in FIG. 1). This is possible because of the feature 3 described later. In this embodiment, the number of messages can be reduced as compared with the conventional case where the route calculation results are distributed as messages to all the routers 20.

(About feature 3)
When the border router 20 receives a packet to be transferred and does not store a route calculation result necessary for the transfer of the packet in its storage unit 23 (see FIG. 3), the border router 20 A necessary route calculation result is acquired from the router 20 (solid arrow in FIG. 1). Then, the border router 20 can transfer the received packet based on the acquired route calculation result. In other words, the router 20 in the PoP manages the route calculation result stored (transmitted) from the route server 10 in a distributed manner. That is, since each router 20 need only store a part of the route calculation results, the storage capacity can be reduced as compared with the related art. Details of this distributed management method will be described later.

<< Functions of the route server 10 >>
Next, functions of the route server 10 used in the route information management system 30 according to the present embodiment will be described with reference to FIG. Note that FIG. 2 shows only some functions of the route server 10 of the present embodiment, and descriptions of functions related to BGP, IGP, and the like are omitted.

  As shown in FIG. 2, the route server 10 includes an input / output unit 11 that controls input / output of various types of information such as external route information and route calculation results, a processing unit 12 that calculates a route based on external route information, and a storage unit. 13. The input / output unit 11 inputs and outputs information with other routers 20 and the like. Input / output of information in the input / output unit 11 is performed based on an instruction from the processing unit 12.

  The processing unit 12 includes a CPU (Central Processing Unit) (not shown) included in the route server 10 and a main memory (not shown). The CPU develops an application program in the main memory and executes it to implement various arithmetic processes. The application program is stored in the storage unit 13. The processing unit 12 includes an external route information collection unit 121, a route calculation unit 122, a storage destination selection unit 123, and a search key generation unit 124.

  The external route information collection unit 121 receives external route information advertised from the border router 20 via the input / output unit 11. Note that all external route information related to the same prefix is collected in the same route server 10. The external route information collection unit 121 stores the received external route information in the external route information DB 131 of the storage unit 13. The route calculation unit 122 reads the external route information newly stored in the external route information DB 131 and executes the above-mentioned “route calculation for each PoP” for each prefix, and results of the route calculation for each PoP, that is, the optimum Calculate the transfer path and suboptimal transfer path. The reason why the suboptimal transfer path is calculated in advance is that when the optimum transfer path cannot be used due to congestion or failure, the suboptimal transfer path is immediately switched to be used.

  The storage destination selection unit 123 uses the router ID of the router 20 stored in the storage destination node DB 132 and a search key including a prefix to be described later as a result of route calculation for each PoP calculated by the route calculation unit 122. The router 20 in the PoP that is the storage (transmission) destination is selected. Then, the storage destination selection unit 123 stores (transmits) the result of route calculation for each PoP in the router 20 selected for each PoP via the input / output unit 11. In addition, the search key generation unit 124 generates a search key used when storing the result of route calculation for each PoP. The storage method and search key generation will be described later.

  The storage unit 13 stores an external route information DB 131 and a storage destination node DB 132. The external route information DB 131 stores external route information related to the prefix advertised by the border router 20. For example, the external route information includes a prefix and path attribute information associated with the prefix. The path attribute information is the link cost assigned to the link leading to the AS-path or the next hop. The AS-path is for recording the AS that has passed through, and an AS number is added each time the AS is passed through. The route with the smaller number of AS numbers recorded in the AS-path is preferentially used. The external route information stored in the external route information DB 131 is used when the route calculation unit 122 executes route calculation. The storage destination node DB 132 stores a router ID that can identify the router 20. This router ID is preset by the administrator of the AS. The storage destination node DB 132 is referred to when the storage destination selection unit 123 selects the storage destination router 20. A specific example of the storage destination node DB 132 will be described later.

<< Function of Router 20 >>
Next, the function of the router 20 of this embodiment will be described with reference to FIG. FIG. 3 shows only a part of the functions of the router 20 of the present embodiment, and descriptions of functions related to BGP, IGP, and the like are omitted.

  As shown in FIG. 3, the router 20 includes an input / output unit 21 that controls input / output of information transmitted / received between the route server 10 and other routers 20, a processing unit 22, and a storage unit 23. . Input / output of information in the input / output unit 21 is performed based on an instruction from the processing unit 22.

  The processing unit 22 includes a CPU (Central Processing Unit) (not shown) included in the router 20 and a main memory (not shown). The CPU develops an application program in the main memory and executes it to implement various arithmetic processes. The application program is stored in the storage unit 23. The processing unit 22 includes a route server selection unit 221, a transfer path determination unit 222, an acquisition destination selection unit 223, an encapsulation processing unit 224, and an acquisition key generation unit 225.

  The route server selection unit 221 refers to the route server ID information stored in the route server ID DB 231 of the storage unit 23 and advertises external route information received from another AS based on the route server determination method described later. The route server 10 is selected. Then, the route server selection unit 221 advertises external route information received from another AS to the selected route server 10 via the input / output unit 21.

  When the transfer route determination unit 222 receives a packet to be transferred, the transfer route determination unit 222 searches the route calculation result information stored in the route calculation result DB 232 of the storage unit 23 and determines the transfer destination of the packet. However, the route calculation result DB 232 may store only a part of route calculation result information necessary for route calculation. In that case, the transfer route determination unit 222 acquires the missing route calculation result information from the other routers 20 selected by the acquisition destination selection unit 223. Then, the transfer route determination unit 222 determines the transfer route (transfer destination) of the packet based on the acquired route calculation result information.

  When the acquisition destination selection unit 223 receives a packet to be transferred, the acquisition destination selection unit 223 refers to the information related to the router 20 stored in the acquisition destination node DB 233 and selects the router 20 that holds the insufficient route calculation result information. . In addition, the encapsulation processing unit 224 executes a process for encapsulating the packet to be transferred with the header of the address of the border router 20 serving as an exit. In the egress border router 20, the encapsulation processing unit 224 executes a process of removing the header of the address of the egress border router 20 from the encapsulated packet. The acquisition key generation unit 225 generates an acquisition key used when selecting the router 20 from which the route calculation result information is acquired. The acquisition destination selection method and the acquisition key generation method will be described later.

  The storage unit 23 relates to a route calculation result DB 232 that stores route calculation result information distributed from the route server 10, a route server ID DB 231 that stores an ID number of the route server 10, and a router 20 that distributes and manages the route calculation result information. An acquisition destination node DB 233 for storing information is stored. The route calculation result DB 232 is referred to in the route calculation of the transfer route determination unit 222. The route server ID DB 231 is referred to when the route server selection unit 221 selects the route server 10 that advertises external route information. The acquisition destination node DB 233 is referred to when the acquisition destination selection unit 223 selects the acquisition destination router 20 of the necessary route calculation result information. Specific examples of the route calculation result DB 232, the route server ID DB 231 and the acquisition destination node DB 233 will be described later.

≪Route server determination method≫
Here, a route server determination method in the router 20 will be described with reference to FIG. 4 (see FIG. 3 as appropriate). FIG. 4 shows a route server 10 indicated by a circle and a hash space covering it. The numbers in the circles in FIG. 4 indicate the ID number of the route server 10. FIG. 4 shows a case where three route servers 10A, 10B, and 10C exist in the hash space. For example, it is assumed that the ID number of the route server 10A is 5, the ID number of the route server 10B is 10, and the ID number of the route server 10C is 15. Note that information related to the ID numbers of the route servers 10A, 10B, and 10C is stored in the route server ID DB 231 of the storage unit 23 in FIG.

  First, a function called next (m) is introduced. next (m) returns the ID of the next node in the hash space in the clockwise direction (that is, in the direction of increasing the ID). For example, next (2) = 5, next (6) = 10, and next (15) = 15. That is, next (m) returns an ID that is the smallest value in a range of a certain value m or more in the hash space. Here, m is a hash value calculated by substituting the prefix value into the hash function. However, 0 <m ≦ 15 in the hash space shown in FIG.

  In FIG. 4, the sharing range of the route server 10A in the hash space is 0 to 5 and below, the sharing range of the route server 10B is 5 to 10 and the sharing range of the route server 10C is 10 to 15 and below. In this way, if the ID number of the route server 10 is determined in advance, it becomes possible to identify the route server 10 that transmits the route information regarding the prefix.

≪Distributed management of route calculation result information≫
As described above, since the router 20 in the PoP takes the same route as the BGP route, the route calculation result information can be distributed and managed by the routers 20 in the PoP. Then, by managing the route calculation result information in a distributed manner, the number of routes (storage capacity) that the router 20 should hold can be suppressed. The distributed management method will be described below.

(Framework)
First, a framework for distributed management will be described. The router 20 in the AS is assigned a unique router ID by the AS administrator in order to identify the router. The route calculation result information is formed by associating a search key with transfer route information (optimum transfer route and suboptimal transfer route) of the route calculation result so that distributed management is possible. However, the router ID and the search key are expressed with the same bit length.

  First, the search key is used in the route server 10 to select which router 20 stores (transmits) the route calculation result information. Then, the route server 10 selects k routers having router IDs closer to the search key in the order of closer distance based on the distance between the search key and router ID described later. The meaning of selecting k is to ensure redundancy. Therefore, if redundancy is not considered, k = 1 may be used. The route server 10 stores (transmits) the route calculation result information in the selected k routers 20.

  Here, the definition of the distance will be described with reference to FIG. In FIG. 5, in order to simplify the description, the bit length of the router ID and the search key is represented by 5 bits. Assume that the router IDs of the routers a, b, c, and d are 10011, 10101, 10110, and 11000, respectively. Assume that the search key is 10111. In this case, an exclusive OR between the router ID and the search key is calculated. Then, in the bit string of the exclusive OR operation result, the numerical value of the number of digits counted from the right at the position where “1” first appears from the left is defined as “distance”. That is, for each router a, b, c, d, the distance between the router ID and the search key is 3, 2, 1, 4 respectively. Therefore, it is determined that the search key is closest to the router c. When redundancy is taken into consideration and k = 2, it is determined that the router b is the next closest to the router c. In this way, the route calculation result information formed by the search key and the transfer route information associated with the search key is stored in the routers b and c. By using this distance, the closer the distance, the longer the matching length of the bit strings counted from the left, and when the distance is 0, there is an effect of indicating that the bit strings are completely matched.

  Next, a method for selecting the acquisition destination router 20 when the router 20 acquires necessary route calculation result information from another router 20 will be described. In this case, the router 20 storing the necessary route calculation result information can be identified using the same distance as when the route calculation result information is stored. Hereinafter, configurations of the router ID, the search key, and the transfer route information, and examples of the storage destination node DB 132, the acquisition destination node DB 233, and the route calculation result DB 232 will be described with reference to FIGS. In the present embodiment, for example, the bit length is described as 160 bits.

The configuration of the router ID is shown in Equation (1).
<Router ID> :: = <PoP ID><HashValue> (1)
However, the PoP ID is an ID for identifying the PoP to which the router 20 belongs, and is 32 bits. The PoP ID is unique identification information set by the AS administrator. The hash value is, for example, an MD5 hash value calculated by MD5 (Message Digest 5) based on a random value, which is one of hash functions, and is 128 bits.

The structure of the search key kr used when selecting the router 20 that is the storage destination of the route calculation result information is shown in Expression (2).
<Kr> :: = <PoP ID><prefixvalue><maskvalue><padding>
..Formula (2)
However, the PoP ID is the same as described above. That is, the PoP ID identifies which PoP the search key is for, and makes it easy to find a node ID including the same PoP ID. The prefix value is a prefix included in the external route information, and is 32-bit IPv4 (Internet Protocol version 4). The mask value is a mask value for the prefix value and is 32 bits. This mask value corresponds to a subnet mask used in IPv4. The padding is 64 bits with all values being “0”. Note that “:: =” indicates that the bit string on the left side and the bit string on the right side are the same, and the right side indicates that the bit strings of <> are connected in this order.

The configuration of the transfer route information value associated with the search key kr that forms the route calculation result information is shown in Equation (3).
<Value> :: = <version number><n><NH_1><priority1> ...
・ ・ <NH_n><Priorityn> ・ ・ Formula (3)
Here, the version number is given each time the route server 10 executes route calculation. n is the number of transfer routes (NH_ *) that are the result of route calculation stored in value. NH_ * represents a transfer path (next hop information) corresponding to the prefix. The priority designates the priority order using the transfer path, and is used in descending order of priority. Since the transfer route information value stores a plurality of transfer routes, the transfer route with the next highest priority can be used immediately when a failure occurs in the selected transfer route.

  Next, an example of the storage destination node DB 132 (see FIG. 2) in the storage unit 13 of the route server 10 will be described with reference to FIG. The storage node DB 132 stores, for each PoP, the router name, IP address, and router ID of the router 20 belonging to the PoP. Since the route calculation result information is distributed and managed in PoP, the storage destination node DB 132 does not have to store information related to all the routers 20 in the AS.

  An example of the acquisition destination node DB 233 (see FIG. 3) in the storage unit 23 of the router 20 will be described with reference to FIG. Note that the acquisition destination node DB 233 illustrated in FIG. 7 represents what is stored in a certain router 20. The range indicates a range in which the router ID is divided into several parts. For example, if the hash space shown in FIG. 4 indicates the entire ID used as the router ID, the shared range of the route server 10 corresponds to “range”. The range is set so that the range close to the router ID is narrow and the range far is set wide. FIG. 7 shows a case where the number of acquisition destination routers k = 2 is provided with redundancy, and shows a state in which information related to two acquisition destination routers is stored for each range.

  An example of the route calculation result DB 232 (see FIG. 3) in the storage unit 23 of the router 20 will be described with reference to FIG. The route calculation result DB 232 stores a part of route calculation result information in PoP. The route calculation result information is represented by a set (kr, value) of a search key kr and transfer route information value.

(Flow of storage processing of the route server 10)
Next, the flow of storage processing by the route server 10 in this embodiment will be described with reference to FIG. 9 (see FIG. 2 as appropriate). In step S901, the route calculation unit 122 refers to the external route information DB 131, performs route calculation for each PoP for each prefix, and generates transfer route information value. In step S <b> 902, the search key generation unit 124 generates a search key Kr including the prefix received from the router 20. In step S903, the storage location selection unit 123 extracts k (predetermined predetermined number) router IDs in order of decreasing (small) distance between the search key kr and the router ID for each PoP. In step S904, the storage destination selection unit 123 transmits a request find_node (kr) for acquiring k pieces of information related to the router of the acquisition destination stored in the router 20 to the router 20 having the extracted router ID. . The acquisition destination selection unit 223 of the router 20 that has received the find_node (kr) compares the search key kr with the router ID at the boundary of the range (see FIG. 7), and extracts the range including the search key kr. The distance between the search key kr and the router ID of each of the routers 20 belonging to the extracted range is calculated, and information related to k acquisition destination routers is returned to the route server 10 in order of increasing distance.

  In step S905, the storage destination selection unit 123 acquires k pieces of information related to the acquisition destination router from the transmission destination router 20 of the find_node (kr). In step S906, the storage destination selection unit 123 determines whether there is a router that has not transmitted find_node (kr) among the routers acquired in step S905. If there is an untransmitted router (Yes in step S906), in step S907, the storage destination selection unit 123 transmits find_node (kr) to the untransmitted router. Then, the process returns to step S905. If there is no unsent router (No in step S906), in step S908, the storage destination selection unit 123 selects and selects k router IDs in order of distance from the search key kr for each PoP. Route calculation result information (kr, value) is stored (transmitted) in the routers with k router IDs. Note that the route calculation result information (kr, value) stored in step S908 is deleted when the route server 10 re-executes route calculation. In addition, the router ID newly acquired by the route server 10 from the transmission destination of the find_node (kr) is stored in the storage destination node DB 132 and updated.

(Process for obtaining transfer route information from other routers 20)
Next, an outline of processing for acquiring transfer path information from another router 20 will be described (see FIG. 3 as appropriate). First, when a packet to be transferred arrives, the acquisition key generation unit 225 of the router 20 generates an acquisition key kd including the destination IP address of the packet to be transferred in order to determine the transfer destination. Details of the acquisition key kd will be described later. Then, the transfer route determination unit 222 refers to the route calculation result DB 232 stored therein and searches for the search key kr included in the route calculation result information that matches the acquisition key kd. When the search key kr matching the acquisition key kd is found, the transfer route information value associated with the search key kr is read and used for packet transfer.

  Further, when the packet to be transferred arrives, the acquisition destination selection unit 223 selects the routers 20 with k router IDs that are close to the acquisition key kd. This distance is the same definition as the distance described above. Then, the acquisition destination selection unit 223 calculates, for the selected router 20, information on k routers that are close to the acquisition key kd held by the router 20 and the route calculation of the search key kr that matches the acquisition key kd. When there is result information (kr, value), a request find_value (kd) for acquiring route calculation result information (kr, value) is transmitted. Then, the transfer route determination unit 222 calculates the longest match with the acquisition key kd (for example, the distance between the plurality of search keys kr and the acquisition key kd) from the acquired plurality of route calculation result information (kr, value). The route calculation result information (kr, value) including the search key kr is selected. Next, the transfer route determination unit 222 reads transfer route information value from the selected route calculation result (kr, value) and uses it for packet transfer.

(Configuration of acquisition key kd)
Here, the configuration of the acquisition key kd is shown in Equation (4).
<Kd> :: = <PoP ID><IPaddress><Maskvalue><Padding>
..Formula (4)
However, the PoP ID is the ID of the PoP to which the router that received the packet to be transferred belongs. That is, the PoP ID identifies which PoP the acquisition key is for. The IP address is a destination IP address of a packet to be transferred and is 32 bits. The mask value is a mask value for the prefix value and is 32 bits. This mask value corresponds to a subnet mask used in IPv4, and is all set to “1”. The padding is 64 bits with all values being “0”.

  As described above, when the packet to be transferred arrives, the border router 20 needs to acquire the transfer route information value that is distributed and managed in order to determine the transfer destination. Therefore, the border router 20 only needs to be able to generate the search key kr that forms a pair with the transfer route information value, but only the destination IP address can be acquired from the packet to be transferred. The bit string indicating the destination IP address and the bit string indicating the destination prefix do not always match completely. That is, when the boundary router 20 generates the acquisition key kd, it cannot always generate the same bit string as the search key kr used when storing the route calculation result information (kr, value). Therefore, the flow of transfer path information acquisition processing performed by the router 20 using the acquisition key kd will be described with reference to FIG. 10 (see FIG. 3 as appropriate).

(Flow of transfer route information acquisition processing)
As shown in FIG. 10, in step S1001, the acquisition key generation unit 225 generates an acquisition key kd including the destination IP address of the packet to be transferred. In step S1002, the acquisition destination selection unit 223 compares the acquisition key kd with the router ID stored in the acquisition destination node DB 233, and selects k routers in order of distance from the acquisition key kd. In step S1003, if the acquisition destination selection unit 223 includes, in the selected router, information related to k routers close to the acquisition key kd and a search key kr that matches the acquisition key kd, the search key kr A request find_value (kd) for acquiring the transfer route information value of the route calculation result information (kr, value) including the is transmitted.

  In step S <b> 1004, the transfer route determination unit 222 determines that the information about the k routers 20 close to the acquisition key kd and the search key kr that matches the acquisition key kd from the destination router 20 of find_value (kd). If it exists, transfer route information value of route calculation result information (kr, value) including the search key kr is acquired. The acquisition destination selection unit 223 of the router 20 that has received find_value (kd) compares the acquisition key kd with the router ID at the boundary of the range (see FIG. 7), and extracts the range including the acquisition key kd. The distance between the acquisition key kd and the router ID of each of the routers 20 belonging to the extracted range is calculated, and information related to k acquisition destination routers is returned to the transmission source router 20 in order of increasing distance.

  In step S1005, the transfer route determination unit 222 determines whether transfer route information value has been acquired. When the transfer route information value is acquired (Yes in Step S1005), in Step S1006, the transfer route determination unit 222 determines that the transfer route included in the transfer route information value (NH_ * in Expression (3)) and the current Compare which of the used transfer routes is the longest match with the IP address of the packet to be transferred, and set the transfer route with the longest match as the currently used transfer route. However, if both the transfer path included in the transfer path information value acquired in step S1005 and the currently used transfer path have the same distance from the IP address of the packet to be transferred, the transfer path information value Set the transfer route that is currently used to the latest version number. In step S1007, the transfer path determination unit 222 sets the currently used transfer path to the FIB (not shown). Then, the process proceeds to step S1008. If transfer path information value is not acquired in step S1005 (No in step S1005), the process skips to step S1008.

  In step S1008, the transfer path determination unit 222 determines whether or not there is a router that has acquired find_value (kd) in step S1004 that has not been transmitted. When there is a router acquired in step S1004 that has not been transmitted (Yes in step S1008), in step S1009, the acquisition destination selection unit 223 transmits find_value (kd) to the router 20 that has not been transmitted. Then, the process returns to step S1004. If there is no unacquired router acquired in step S1004 (No in step S1008), the process proceeds to step S1010.

  In step S1010, the transfer path determination unit 222 changes the last (rightmost) non-zero bit of the IP address part of the acquisition key kd to zero, and at the same time includes the bit at the change position of the IP address in the mask value portion. The right bit is changed to zero, and the changed bit string is set as a new acquisition key kd. A specific example of the operation when changing the bit of the acquisition key kd will be described with reference to FIG. As shown in FIG. 11, it is assumed that the bit string of the IP address part before the bit change and the bit string of the mask value part are <10101100> <11111111>, respectively. The last (rightmost) non-zero bit of the IP address part is the third from the right. The third bit from the right is changed to zero. For the mask value portion, the right three bits including the third bit from the right are changed to zero.

  In step S1011, the transfer path determination unit 222 determines whether the host part of the IP address part of the acquisition key kd is all zero or not. If a part of the host part of the IP address part of the acquisition key kd is not zero (No in step S1011), the process returns to step S1002. If the host part of the IP address part of the acquisition key kd is all zero (Yes in step S1011), in step S1012, the transfer path determination unit 222 replicates the route calculation result information including the currently used transfer path. Is executed for the router having the router ID closest to the (k + 1) th distance from the initial acquisition key kd. Then, the transfer route information acquisition process ends.

  Note that the packet to be transferred is transferred based on the transfer path set in the FIB in step S1007. At this time, the encapsulation processing unit 224 encapsulates the packet to be transferred with a header whose destination is the IP address of the border router at the exit, and transfers the packet using the IGP. The effect of this encapsulation is that it is not necessary for the router 20 to hold the BGP route information of other PoPs, and that it is possible to prevent the router 20 passing in the AS from forwarding a packet to an unintended route. .

  The reason why the bit string of the IP address part of the acquisition key kd is changed to zero in the last (rightmost) non-zero bit in step S1010 and the search returns to step S1002 to continue the search is as follows. , Value) is managed based on a prefix which is a set of destination IP addresses, whereas the acquisition key kd is generated using the destination IP address. That is, it is considered that the possibility of finding the search key kr that directly matches the acquisition key kd is low. This is because the bit string in the IP address portion is searched for the bit string with the longest match while changing the last (rightmost) non-zero bit to zero.

  In step S1006, when a plurality of longest matching transfer paths are found, the latest version number included in the transfer path information value is selected. In step S1012, the route calculation result information to be replicated is stored in the router 20 closest to the (k + 1) th distance. If already stored in the (k + 1) th router 20, the information is replicated to the (k + 2) th router. The replicated route calculation result information is deleted after a predetermined time has elapsed (after a timeout). The predetermined time is set in inverse proportion to the distance between the acquisition key kd and the router ID. That is, the predetermined time-out time is set shorter as the distance is longer.

(Reduction effect of storage capacity)
As described above, in the present embodiment, since the external route information stored in the router 20 is stored in the route server 10, it is possible to reduce the storage capacity compared to the conventional router. In addition, the route server 10 can also reduce the storage capacity per unit by increasing the number of prefixes that are shared and stored. Furthermore, in each conventional router, the information on the transfer route necessary for packet transfer (information stored in the FIB) has to be stored for all prefixes. In the present embodiment, By managing the distribution within the PoP, the storage capacity of the router 20 can be reduced. Quantitatively, the number of routes held per router 20 is calculated by adding the number of routes of information to be replicated to the value obtained by dividing the total number of routes by the number of routers in PoP. In the Internet, since most of the traffic tends to concentrate on a small number of destinations, the number of routes to be replicated is smaller than the value obtained by dividing the total number of routes by the number of routers in the PoP.

(Reduction effect of the number of messages and the amount of calculation processing)
Since the route server 10 performs route calculation, each router 20 may not execute route calculation. Therefore, in the router 20 of this embodiment, the amount of calculation processing can be reduced as compared with the conventional router. Furthermore, in the present embodiment, the router 20 in the AS belongs to one of a plurality of PoPs. Since the BGP route is the same transfer route in PoP, the route server 10 only has to execute route calculation for one router 20 for each PoP, and the route calculation result is used as a message in the PoP. What is necessary is just to transmit to k routers 20 fewer than the number of routers in the. Therefore, compared with the case where the route server 10 transmits all the route calculation results for each router 20, the number of messages and the amount of calculation processing (processing load) can be reduced qualitatively.

  Quantitatively, if the average number of routers in the PoP is Rpop, and the number of messages when the route calculation result (message) is transmitted from the route server 10 to all the routers 20 is M, The number of messages Rm transmitted by the route server 10 is M / Rpop + Nf. Here, Nf is the number of find_node messages. In the initial state, this Nf may be up to the number of routers in the PoP. However, when the find_node is repeated, the storage destination node DB 132 of the route server 10 is updated, and the route server 10 cannot know. Since the router 20 disappears, it finally converges to k (redundancy). Therefore, the number of messages is reduced.

  Further, the number of find_value messages depends on the distribution of destinations of packets to be transferred. In the Internet, most packet destinations are concentrated on a small part of the prefix, so that it is not necessary to make an acquisition request to another router by replication. Therefore, the number of find_value messages converges to almost zero. Therefore, in this embodiment, the number of messages can be reduced with the introduction of PoP and replication.

  As described above, according to the route information management system 30 of the present embodiment, since the route server 10 stores the external route information related to the prefix, the number of routes (storage capacity) to be held in the router 20 can be reduced. Further, since the route server 10 collects all the external route information related to the same prefix and executes route calculation, there is no problem that the route calculation result is not optimal. In addition, when a plurality of PoPs are provided in the AS, the BGP route in the PoP is common to the prefix, so that the route server 10 performs route calculation for only one router 20 in the PoP. Better. Therefore, the processing load on the route server 10 can be reduced, and route calculation at the router 20 becomes unnecessary. In addition, since the result of route calculation for each PoP calculated by the route server 10 is distributed and managed in the PoP, it may be stored (transmitted) in k (predetermined number) routers 20 for each PoP. Therefore, it is possible to reduce the number of messages when the route calculation result is transmitted as a message. Further, since the route calculation result information is distributed and managed by the routers 20 in the PoP, the storage capacity of each router 20 is the memory of one conventional router 20 storing all the route calculation result information. Compared to capacity, it can be reduced. Further, since the search for obtaining the necessary route calculation result is limited to only the router 20 in the PoP, the obtaining process is accelerated.

  In addition, this embodiment is not limited to these, It can implement in the range which does not change the meaning. For example, when the acquisition destination selection unit 223 selects k routers 20 that are close to the search key kr or the acquisition key kd, in this embodiment, k selected by the range including the search key kr and the acquisition key kd. Described as obtaining routers. However, the distance between the router ID stored in the storage unit 23 of the router 20 and the search key kr or the acquisition key kd may be calculated to select the k routers 20 having the closest distance. In addition, in FIG. 7, as an example, a case where information related to k routers for each range is stored in the acquisition destination node DB 233 of the storage unit 23 is shown. However, when information related to more than k routers 20 is stored for each range, k routers 20 having the shortest distance may be selected from the information.

  When the transfer route determination unit 222 searches for the search key kr that matches the acquisition key kd, the search range is narrowed down if the route calculation result information (kr, value) is stored in association with each range. The search time can be reduced. Further, the router 20 can speed up the search by storing the transfer route information value acquired by find_value (kd) in a cache memory (not shown) of its own storage unit 23. Note that the information stored in the cache memory is deleted after a predetermined time (after timeout). Note that replication and cache are managed by one Radix Tree that stores transfer path information.

  The router 20 may also serve as the function of the route server 10. Moreover, since this embodiment does not modify eBGP, which is a communication protocol between ASs, it can be applied by being closed in AS. And AS which applied this embodiment can be connected with other AS as usual.

  In the present embodiment, the processing of the units 11 and 12 of the route server 10 (see FIG. 2) may be realized by a program installed when the route server 10 is realized by a computer. In the present embodiment, the processing of the units 21 and 22 of the router (see FIG. 3) may be realized by a program installed when the router 20 is realized by a computer. This program can be provided via a communication line, or can be distributed by writing in a computer-readable recording medium such as a CD-ROM.

1 AS
10 Route server (route information management device)
11 Input / output unit 12 Processing unit 13 Storage unit 20 Router (border router, router)
22 processing unit 23 storage unit 30 route information management system 121 external route information collection unit 122 route calculation unit 123 storage destination selection unit 124 search key generation unit 131 external route information DB
132 Storage node DB
221 Route server selection unit 222 Transfer route determination unit 223 Acquisition destination selection unit 224 Encapsulation processing unit 225 Acquisition key generation unit 231 Route server IDDB
232 Route calculation result DB
233 Acquisition destination node DB

Claims (10)

In an AS (Autonomous System) constituting an IP (Internet Protocol) network, external route information including a prefix that is a set of IP addresses of the other AS is acquired from another AS, and the packet is transferred to the other AS. A route information management system in which a plurality of routers and two or more route information management devices that collect external route information including the same prefix from the router and perform route calculation are connected to each other so as to be able to communicate with each other ,
The route information management device
For each prefix collected from the router, at least one of a plurality of routers set in the AS with a group of two or more routers in the AS using external route information including the prefix as a unit A route calculation unit that performs route calculation for calculating a transfer route to a prefix of a packet transfer destination in one router;
A storage location selection unit that selects a predetermined number of routers belonging to the router set for transmitting route calculation result information including a route calculation result for each router set calculated by the route calculation unit;
An input / output unit that transmits the route calculation result information to the predetermined number of routers selected by the storage destination selection unit;
With
The router
A transfer path determination unit that determines a transfer destination of the packet with reference to the path calculation result information received from the path information management device when a packet to be transferred is received;
An input / output unit that transmits the packet to be transferred to the transfer destination determined by the transfer path determination unit;
A route information management system comprising: The route information management device further includes:
A storage location for storing a bit string obtained by concatenating a bit string of a hash value generated by a hash function after a bit string of a router set ID of the router set to which the router belongs, as a router ID of at least one router for each router set A storage unit for storing node information;
For each router set, as a search key used for searching the storage destination of the route calculation result information, the router set ID bit string of the router set, the prefix bit string, the mask value bit string for the prefix, and the padding are all “ A search key generation unit for generating a bit string obtained by concatenating bit strings of “0”;
With
The storage destination selection unit
The distance between the router ID stored in the storage destination node information and the search key is first set to “1” when viewed from the left in the bit string indicating the exclusive OR for each bit of the router ID and the search key. 2. The bit position where “” appears is calculated as a numerical value of the number of bits counted from the right, and a predetermined number of routers are selected as destinations from the router set in ascending order of the calculated distance. The route information management system described in 1. The route according to claim 2, wherein the route calculation result information is information in which the search key is associated with a result of route calculation for each router set corresponding to a prefix included in the search key. Information management system. The router further includes:
A storage unit for storing acquisition destination node information for storing router IDs of at least one or more routers excluding itself in the router set to which the router belongs, and the route calculation result information;
When a packet to be transferred is received, a bit string of a router set ID to which the router belongs, a bit string of a destination IP address of the packet, a bit string of all “1” as a mask value for the destination IP address, and all “0” as padding An acquisition key generation unit that generates an acquisition key formed by concatenating bit strings of
The distance between the router ID and the acquisition key stored in the acquisition destination node information is first set to “1” when viewed from the left in the bit string indicating the exclusive OR for each bit of the router ID and the acquisition key. `` Appears as a numerical value of the number of digits of the bit counted from the right for the bit position, and an acquisition destination selection unit that selects a predetermined number of routers as an acquisition destination from the router set in ascending order of the calculated distance;
With
The transfer path determination unit
It is determined whether or not the search key of the route calculation result information stored in its storage unit matches the acquisition key, and if there is a match, the router set associated with the search key The packet transfer destination is determined based on the result of each path calculation, and if there is no match, the acquisition key is transmitted to the router selected as the acquisition destination by the acquisition destination selection unit, and the router A route calculation result for each router set in which the search key of the route calculation result information stored and the acquisition key match are acquired, and based on the acquired route calculation result for each router set 4. The route information management system according to claim 3, wherein a transfer destination is determined. The transfer route determination unit further, when there is no match between the search key of the route calculation result information and the acquisition key,
The last non-zero bit of the destination IP address portion of the acquisition key is changed to zero, and the right bit including the change position of the destination IP address portion of the mask value portion of the acquisition key is changed to zero, and the change is made The obtained bit string is set as a new acquisition key, and using the new acquisition key, from the acquisition destination router selected by the acquisition destination selection unit of itself, and further by the acquisition destination selection unit of the acquisition destination router Repeat the acquisition of the selected acquisition destination router and route calculation result information, and determine whether or not the search key of all the acquired route calculation result information matches the new acquisition key The packet transfer destination is determined based on the route calculation result for each router set when the search key of the route calculation result information and the acquisition key have the longest match. Route information management system of claim 4 Rukoto a bit string of the destination IP address portion of the acquisition key and said repeatedly executes it until all zero. The acquisition node information divides the router ID into a plurality of ranges, and sets a narrower range for the router ID closer to the distance from the router ID of its own router, and sets the range for the router ID farther away. Set a wide range, and for each said range, store the router ID of the router belonging to that range,
The acquisition source selection unit compares the acquisition key with a router ID at the boundary of the range, extracts the range including the acquisition key, and obtains the acquisition key and a router ID of a router belonging to the extracted range. The route information management system according to claim 4, wherein the distance is calculated, and a predetermined number of routers are selected as acquisition destinations in order of increasing distance. A route information management method for a route information management device used in the route information management system according to claim 4,
The route information management device
For each set of routers set in the AS in units of a group of two or more routers in the AS, as router IDs of at least a predetermined number of routers, the router set ID of the router set to which the router belongs A storage unit for storing storage destination node information for storing a bit string obtained by concatenating a bit string of hash values generated by a hash function after the bit string;
For each prefix collected from the router, at least one of a plurality of routers set in the AS with a group of two or more routers in the AS using external route information including the prefix as a unit A route calculation step for performing a route calculation for calculating a transfer route to a prefix of a packet transfer destination in one router;
For each router set, as a search key used for searching a storage destination of route calculation result information including a route calculation result for each router set calculated by the route calculation step, a bit string of a router set ID of the router set, A search key generation step of generating a bit string obtained by concatenating a bit string of the prefix, a bit string of a mask value for the prefix, and a bit string of all “0” as padding;
The distance between the router ID stored in the storage destination node information and the search key is first set to “1” when viewed from the left in the bit string indicating the exclusive OR for each bit of the router ID and the search key. `` Appears as a numerical value of the number of digits of the bit counted from the right for the bit position where it appears, and a storage destination selection step of selecting a predetermined number of routers as a transmission destination from the router set in ascending order of the calculated distance;
A route information management method for a route information management device, comprising: performing an input / output step of transmitting the route calculation result information to the predetermined number of routers selected by the storage destination selection step. A route information management method for a router used in the route information management system according to claim 4,
The router
For each set of routers set in the AS in units of a group of two or more routers in the AS, as router IDs of at least a predetermined number of routers, the router set ID of the router set to which the router belongs Associating acquisition destination node information that stores a bit string obtained by concatenating a bit string of hash values generated by a hash function after a bit string, a search key transmitted from the path information management device, and a result of path calculation for each router set A storage unit for storing route calculation result information indicating the received information,
When a packet to be transferred is received, a bit string of a router set ID to which the router belongs, a bit string of a destination IP address of the packet, a bit string of all “1” as a mask value for the destination IP address, and all “0” as padding An acquisition key generation step for generating an acquisition key formed by concatenating the bit strings of
The distance between the router ID and the acquisition key stored in the acquisition destination node information is first set to “1” when viewed from the left in the bit string indicating the exclusive OR for each bit of the router ID and the acquisition key. Is obtained as a numerical value of the number of bits of the bit counted from the right for the bit position where `` appears '', and an acquisition destination selection step of selecting a predetermined number of routers as an acquisition destination from the router set in ascending order of the calculated distance;
When a packet to be transferred is received, it is determined whether the search key of the route calculation result information stored in the storage unit of the packet matches the acquisition key. A packet transfer destination is determined based on a route calculation result for each router set associated with the search key, and if there is no match, the router selected as the acquisition destination by the acquisition destination selection unit Sending the acquisition key, acquiring a route calculation result for each router set in which the search key of the route calculation result information stored in the router matches the acquisition key, and acquiring the obtained router set A transfer route determining step for determining a transfer destination of the packet based on the result of the route calculation of
A route information management method for a router, comprising: performing an input / output step of transferring the packet to a transfer destination determined in the transfer route determination step.   A program for causing a route information management device as a computer to execute the route information management method according to claim 7.   A program for causing a router as a computer to execute the route information management method according to claim 8.

Images