JP4377858B2 - Hierarchical distributed routing method and its management device - Google Patents

Description

  The present invention relates to a hierarchical routing method used for performing the shortest path selection in an MPLS (Multi Protocol Label Switching) or GMPLS (Generalized MPLS) network, and an interdomain routing management apparatus for realizing this method.

  In recent years, with the expansion of information communication networks represented by the Internet, more efficient routing methods have been demanded. Particularly, in a network including a plurality of domains, routing between domains is necessary, and various studies have been attempted on this. As an example, an OSPF (Open Shortest Path Fist), a PNNI (Private Network to Network Interface) routing method defined by the ATM (Asynchronous Transfer Mode) forum, and its application to the above technical field will be described.

  First, a method using the OSPF protocol for shortest path selection in a network having a plurality of domains will be described with reference to FIGS. 3 to 5 (see, for example, Non-Patent Document 1). Here, a case will be described as an example where the origin router or an area border router existing in the same area as the origin router serves as a routing management system and determines the shortest route between the origin router and the destination router in another domain.

  First, the network configuration / conditions on which the description is based will be described with reference to FIG. In FIG. 3, 101 to 106 are areas which are domain units of OSPF, and in these areas 101 to 106, there are intra-area routers 201 to 210 and area border routers 301 to 307, respectively. The network also includes backbone routers 401 and 402, and the area routers 301 to 307 are connected to each other by the backbone routers 401 and 402. Reference numerals 601 to 611 denote links in the area A to the area F, and reference numerals 701 to 716 denote links in the backbone area.

  The numbers at both ends of each link indicate the link cost, which is the cost from that end of the link to the other end of the link. The link state information including the link cost is represented by numerical values according to conditions such as delay, throughput, and reliability, but is dynamically reflected according to the link usage frequency. The area border routers 301 to 307 and the backbone routers 401 and 402 share link state information regarding the link in the backbone area connected to the own router with all other area border routers.

  The link state information exchanged between the areas 101 to 106 is notified to all the routers 201 to 210 in the area by the area border routers 301 to 307 in each area. In this way, each router can share link state information in its own area with other routers in its own area, and can also share link state information in its backbone area with all other routers. However, link state information inside other areas is not retained.

  As can be seen from FIG. 3, all the area border routers 301 to 307 need to be connected to the backbone area. Although not shown in this example, the area border routers 301 to 307 are not directly connected to the backbone area, but are called virtual links, using virtual links connecting the area border routers 301 to 307 and the backbone area. Sometimes connected.

  FIG. 4 shows a shortest path tree from the intra-area router (R1) 201 to the backbone router (R2) 402 under the conditions of FIG. This shortest route is obtained as a result of selecting the route having the shortest distance with the link cost shown in FIG. 3 as the distance. Reference numerals 801 to 810 denote distances between the routers. This shortest path tree is used to find the shortest path from the router at the top of the tree to another router.

FIG. 5 shows a flowchart of Dijkstra's SPF (Shortest Path First) algorithm used for obtaining the shortest path tree as shown in the example of FIG. 4 (see, for example, Non-Patent Document 3). .
First, the origin router is selected in step ST5a. Next, in step ST5b, a link coming out of the router selected in step ST5a is selected. When the starting point is the router (R1) 201, the target link is only 601 in FIG. In step ST5c, the router and link permutation from the starting point to the link selected in step ST5b is registered in the shortest path candidate list as a route. When the starting point is the router (R1) 201, only R1 201, link 601 and ABR A 301 are registered as the processing result of step ST5c. Here, if there is no link going out from the router, it is not added to the shortest path candidate list, and the elements of the shortest path candidate list do not change.

  In step ST5d, it is determined whether or not one or more shortest path candidates exist in the elements of the shortest path candidate list. If there is no route, the algorithm ends. On the other hand, when one or more routes exist, the process proceeds to step ST5e. In step ST5e, the shortest path candidate whose total cost of links constituting the route is the minimum in the shortest path candidate list is selected, and the route is excluded from the shortest path candidate list.

  In step ST5e, a process of comparing the selected shortest path candidate with an element of the path list is performed. If a route having the same starting and ending router as the shortest route candidate already exists in the shortest route list, the process returns from step ST5f to step ST5e. If not, the process proceeds to step ST5g. . In step ST5g, a process of adding the shortest path candidate as a new element of the shortest path list is performed, and after this process, the process returns to step ST5a. In step ST5a, the shortest path candidate added in step ST5g is set as the target of the end point router, and the processes in step STb and subsequent steps are performed.

In this example, the destination router is ABR A 301,
(1) R1 201, 1ink 601, ABR A 301, link 701, ABR D 304,
(2) R1 201, link 601, ABR A 301, 1ink 702, Backbone R1 401,
(3) R1 201, link 601, ABR A 301, link 704, ABR B 302
Are registered in the shortest path candidate list.

  As described above, when the algorithm is completed, the shortest paths to all the routers connected to the router are obtained from the origin router selected in the first processing step ST5a. There are the following four scenarios in the multi-domain (multi-area) GMPLS / MPLS routing scenario using this OSPF (see, for example, Non-Patent Document 2).

In the following description, it is assumed that the start point is R1 201 in FIG. 3 and the end point is R10 210.
Scenario 1: The route from the origin router R1 201 to the area border router ABR in the same area is selected by the origin router R1 201, the route from the origin area ABR to the destination area ABR is selected by the origin area ABR, and the destination area ABR The route from the destination router to the destination router is calculated by the destination area ABR.

  In FIG. 3, since ABR A 301 is the only ABR in the same area as the origin router R1 201, the shortest path to the ABR in the same area is obtained as R1 201, link 601 and ABR A 301. Next, the area border router ABR 301 obtains the shortest route to the area F106. Since the distance to ABR F 306 is smaller than the distance to ABR G 307, the path to ABR F 306 is selected as the shortest path. The paths are ABR A 301, 1ink 702, Backbone R1 401, link 708, Backbone R2 402, link 714, and ABR F 306. The area border router ABR F obtains the shortest route to the destination router R10 as ABR F 306, link 610, and R10 210.

  Scenario 2: The origin router requests the representative area border router ABR in the same area to obtain the shortest path from the origin router to the destination area. Thereafter, the origin router makes a shortest route selection request from the ABR to the destination router with respect to the ABR in the determined destination area. The representative ABR for obtaining the shortest route in the starting point area does not necessarily have to exist on the shortest route. In FIG. 3, router R 201 makes a shortest path search request to area border router ABR A 301. Since there is only one ABR in area A101, area border router ABR A301 is necessarily selected as the representative ABR. The area border router ABR A 301 uses the shortest path tree from the origin router R1 201, and the shortest path to the area border router ABR F 306 is R1 201, link 601, ABR A 301, link 702, Backbone R 401, link 708, Backbone R2 402, link 714, ABR F 306. The area border router ABR F 306 obtains the shortest distance to the destination router R10 as ABR F 306, Link 610, and R10 210.

  Scenario 3: The origin router calculates the shortest path of the area border router ABR of the origin area in scenario 2. Others are the same as scenario 2. In FIG. 3, the origin router R1 201 obtains the shortest route to the area F 106 from the shortest route tree from the origin router R1 201, and the area border router ABR F 306 obtains the shortest route to the destination router R10.

  Scenario 4: In scenarios 1 to 3, it is not guaranteed that the shortest route to the end point area selected by the start point router or the representative ABR of the start point area will reach the end point router. In this case, it is necessary to select a route again at the origin router or the representative ABR. In order to solve this problem, the originating router first causes the representative ABR of the end point area to calculate arrival availability information from the plurality of ABRs in the end point area to the end point router. Subsequently, the representative router in the end point area passes the calculated arrival / absence information from the plurality of ABRs to the end point router in the end point area to the representative of the start point area. The representative ABR of the starting point area obtains the shortest path from the starting point router to the ending point area by arriving at an ABR that can reach the ending point area as a constraint.

  In FIG. 3, the origin router R1 201 sends a shortest route selection request to the area border router ABR A 301 in the same area. The area border router ABR A 301 notifies the ABR G 307 (assuming that the ABR G 307 is the representative ABR), which is the representative ABR of the end point area, to the route information in the end point area (information on availability of arrival at the end point router). Request. The area border router ABR G 307 in the end point area reports that it cannot reach from ABR F 306 to R10 210 but is reachable from ABR G 307 (this condition is assumed). The area border router ABR A 301 takes the shortest route from the origin router R1 201 to the area border router ABR G 307 in the destination area, and R1 201, link 601, ABR A 301, link 702, Backbone R2 401, link from the shortest route tree. 708, Backbone R2 402, link 715, ABR G 307.

Next, a method of using a PNNI (Private Network to Network Interface) protocol for selecting the shortest path on a plurality of domains will be described with reference to FIGS. 6 and 7 (see, for example, Non-Patent Document 4).
FIG. 6 is a diagram in which ATM physical nodes arranged in the same manner as FIG. 3 are divided into logical nodes belonging to PG (Peer Group). In FIG. 6, 2101 to 2106 are PGA to PGF, which are PGs in the lowest hierarchy. 2111 to 2113 are PGs one level higher than these six PGs, and are PG1 to PG3. Reference numerals 2201 to 2210 denote N1 to N10 which are logical nodes, but they do not have a link to go out of their own PG. Reference numerals 2301 to 2307 denote border nodes BNA to BNG (BN; Border Node), and these nodes hold links that go out of at least one PG.

  2601 to 2633 are logical links. The PGs 2101 to 2106 are treated as logical group nodes in the upper level (211 to 2113). The logical link connects the logical nodes in the lowermost PG (2101 to 2106), but also connects the logical group nodes in addition to the logical nodes in the upper PG or between the PGs. Multiple links connected to one logical group node are aggregated and treated as one logical link.

  For example, the logical link 2626 substantially handles two links (714 and 715 in FIG. 3) as one logical link. Each PG has one node of PGL (Peer Group Leader). This node is either a logical node or a logical group node. In FIG. 6, PGL of PG A 2101 is N1 2201, PGL of PG B 2102 is N3 2203, PGL of PG C 2103 is N5 2205, PGL of PG D 2104 is N7 2207, PGL of PGE 2105 is N8 2208, PG The PGL of F 2106 is N9 2209. For the PGL of the PG one level higher, the logical group node is PGL.

  For example, the PGL of PG1 2111 is the logical group node corresponding to PG A 2101 is PGL, the PGL of PG2 2112 is the logical group node corresponding to PG B 2102, and the PGL of PG3 2113 is the logic corresponding to PG C 2103 The group node becomes PGL.

  The PGL has two important functions other than the functions of other nodes of the same PG. One is to transmit the link topology between the nodes of the PG to a higher-level PG in an aggregated form. The other is to distribute the link state information obtained in the upper PG to all nodes in the PG. That is, a physical node that plays the role of PGL of a certain PG simultaneously becomes a physical representative node of the logical group node in the upper PG, and aggregated link topology information from the lower PG between the lower PG and the upper PG, Further, the link state information exchanged by the upper PG is exchanged from the upper.

  FIG. 7 shows an example of an algorithm for creating aggregated link topology information in the lower PG. In FIG. 7, 2626 and 2627 are logical links between logical group nodes corresponding to N12 2402 and PG F 2106, logical group nodes corresponding to PG C 2103 and logical group nodes corresponding to PG F 2106, respectively, as in FIG. Indicates the logical link between. Reference numeral 2701 denotes the center of the aggregate link topology information, and 2702 and 2703 denote the output port of the logical link to PG C 2103 and the output port to N12 2402, respectively. Reference numerals 2704 and 2705 are default spokes from the center 2701 to the output ports 2702 and 2703, respectively, and have equal default values from the center.

  However, if the link value between the ports is significantly different from the default value (for example, if a sufficient bandwidth is secured between the ports 2702 and 2703, the link value is reduced), an exception bypass such as 2706 is used. To set the link value. As described above, the outline of the aggregated link topology information is determined by determining the center of the PG and calculating the radius from the center, and exception bypass is applied only in exceptional cases.

  There are two reasons for using aggregate values in this way. One is to disclose the information in one PG to the other PG as much as necessary. The other is that the aggregated information is communicated to be shared among the logical group nodes of the higher-level PG, so that the traffic is reduced. Here, as in the OSPF example, a scenario will be described in which the shortest path selection from the start point N1 2201 to the end point Nl0 2210 is performed based on the PNNI routing protocol.

  First, the starting point N1 2201 corresponds to PG A 2101 with link state information exchanged with other logical group nodes (PG2 2112, PG3 2113) in the same PG (highest PG1) with the logical group node corresponding to PG1 2111. Held in N1 via the logical group node. For this reason, the shortest route is obtained from the link state information. As a result, between PG1 2111 and PG3 2113, it is determined that (PG1 2111, link 2633, PG3 2113) is the shortest path using the logical link 2633 without going through PG2.

  Next, N1 obtains the shortest route to N11 2401 through which the route to PG C 2103 physically passes from the link state information in PG1 2111 obtained from the logical group node corresponding to PG A 2101. As a result, it is determined that the cost directly connected to N11 is lower using the logical link 2622 than via the logical group node corresponding to PGD 2104. As a result, it is determined that (PG A 2101, link 2622, N11 2401) is the shortest path in PG1 2111.

The route from N1 2201 to N11 must pass through BNA 2301. However, the route is only (N1 2201, link 2601, BNA 2301). As a result, the route from PG1 2111 to PG3 2113 is hierarchical,
Bottom layer: (N1 2201, link 2601, BNA 2301)
Second layer: (PG A 2101, link 2622, N11 2401)
Top layer: (PG1 2111, link 2633, PG3 2113)
It becomes.

  As a result, N1 2201, link 2601, BNA 2301, link 2622, N11 2402, link 2633, and N12 2402 are selected as the shortest paths at the logical node level of the lowest layer. N12 is a node connected to N11. As a result, a route from N1 2201 to N12 2402 is secured. In N12, the shortest route to PG F 2106 in PG3 2113 is obtained. N12 2402 is a logical node belonging to PG3 2113. For this reason, route selection is performed using the link states 2625 to 2627, and the route directly connected to the PGF by the logical link 2626 is obtained as the shortest route.

  The logical link 2628 is aggregated and actually consists of two sections of links. One is a link connecting N12 2402 and BN F 2306 and a link connecting N12 2402 and BN G2307. As to which of these is selected, PNNI does not provide a guideline, and it is assumed that it is selected randomly. If BNG 2307 is selected, (N12 2402, link 2626, BNG 2307) is secured as the shortest path. BNG 2307 is a logical node of PG F 2106 and holds link state information in the PGF. However, since the route to N10 is only (BN G 2307, 1ink 2611, N10 2210), this route is selected and secured as the shortest route. Thus, the route from N1 to N10 is selected.

  As a technology to replace OSPF and PNNI, the present inventors distribute PCS (Path Computation Server) or NMS (Network Management System) to each domain, and also distribute it to higher domains including multiple lower domains, and inter-domain routing Has been proposed (see, for example, Non-Patent Document 5). FIG. 8 shows an arrangement example in the case of distributing the PCS, and FIG. 9 shows an algorithm when inter-domain routing is performed by the PCS higher in the hierarchy.

  In the example of FIG. 8, a network having the same configuration as that of FIG. 3 is managed in order to facilitate comparison with OSPF and PNNI. However, the area in OSPF is handled as SN (Sub network) 3001 to 3006. The SN is equivalent to the domain, and the OSPF area and the PNNI PG can also be handled as the SN. Similar to the PNNI PG, the SN can hierarchically hold a plurality of SNs in a lower layer, and can realize a nested structure of a plurality of SNs. The T0PSN 3007 has six SNs at the lower level, and is an SN for managing inter-domain routing between the six SNs. Reference numeral 1011 denotes a PCS distribution server, which distributes PCS software to the PCS installed servers 1012 to 1017 arranged for each SN.

  In the example of FIG. 9, an inter-domain (SN) routing algorithm of a PCS that holds a plurality of SNs in the lower order like the PCS-mounted server 1011 is shown. The PCS selects a lower one SN (or router) in the processing process of step ST9a. The router handles a domain including one SN as an extreme example of the SN.

  For example, in FIG. 8, the PCS 1011 may select SN A 3001 as the starting point SN, or may select BackBone R 1401 as the starting point. However, routers in the lower SN such as R1 to R10 and ARB A to ARB G are not targeted. In the Dijkstra algorithm of FIG. 5, only the router is selected in step ST5a, but the SN can also be selected, which is different from the process in step ST5a. In the processing process of step ST9b, a link that has gone out from the SN (or router) selected in step ST9a is selected.

  When this process is applied in FIG. 8, if the starting point SN is SN A 3001, the links 701, 702, and 704 are outside the SN. The link selected in step ST9b is added to the shortest path candidate list in step ST9c, similar to the process in step ST5c of Dijkstra. However, as shown in the example of step ST9c, in addition to the permutation of the SN and the link between them, the information of the origin router and the destination router is retained. In the condition of FIG. 8, there are three routes registered here. That is, the route from ABR A 301 to ABR D 304, the route from ABR A 301 to Backbone R1 401, and the route from ABR A 301 to ABR B 302.

  In the process of step ST9d, as in the process of step ST5d of Dijkstra, it is determined whether or not there is one or more shortest path candidates in the shortest path candidate list. If there is no shortest path candidate, the algorithm is terminated. On the other hand, if there are one or more, the process proceeds to step ST9e. In the process of step ST9e, as in the process of step ST5e, the one with the smallest sum of the costs of the links constituting the route is selected from the shortest path candidate list. Then, the selected shortest path candidate is excluded from the shortest path candidate list and compared with the elements of the shortest path list. In the example of FIG. 8, the cost comparison of the three routes is performed.

  In the process of step ST9f, it is determined whether or not the same route is in the route of the shortest path list for the origin router and the destination route. If the result of this determination is that there is already the same origin / endpoint router, processing returns to step ST9e. On the other hand, if there is no same origin / endpoint router, the process proceeds to step ST9g, and the route selected in step ST9e is added to the shortest path list. After step ST9g, the process proceeds to step ST9a. At this time, the target domain is the end point domain of the route registered in the shortest path list in the process of step ST9g.

As mentioned above, the difference from Dijkstra's algorithm is that
(1) The domain is handled in the process of step ST9a.
(2) The origin / endpoint router is stored as information in addition to the permutation of the domain and link route components as the route after the processing of step ST9c.
(3) In the process of step ST9f, it is not determined whether or not the same start / end points of the route components are in the shortest path list, but it is determined whether or not the same start / end routers exist. .
It is.

  As can be seen from this example, the algorithm shown in FIG. 9 is for obtaining the shortest path between lower-level SNs directly belonging to lower levels of higher-level SNs or routers. For example, when an optical path is set between lower IP domains in a GMPLS network, it is very effective for obtaining the shortest path between a plurality of IP domains and an optical router.

  If the lower domain wants to hide the internal link topology of its own domain from the upper domain, the upper PCS first finds the shortest path between the domains using the algorithm of FIG. 9, and the internal path is an ABR or the like determined by the upper PCS. It can be set by a method in which each lower PCS whose route passes between the domain border routers or between the origin / endpoint router and the border router is determined.

  For example, when obtaining a route from the starting point R1 201 to the ending point R10 210, the PCS installed in the PCS server 1012 is the shortest route permutation from the SN A 3001 to which the router R1 belongs to the SN F 3006 to which the router R10 belongs. (SN A 3001, link 702, Backbone R 1401, link 708, Backbone R2 402, 1ink 714, SNF), the origin router is ABR A 301, and the destination router is ABR F 306.

  Thereafter, the shortest path between the R1 201 and the ABR A 301 passes through 1ink 601 for the PCS in the PCS server 1012, and the shortest path between the ABR F 306 and the router R10 has a link 610 for the PCS in the PCS server 1017. Both are obtained by applying the Dijkstra algorithm of FIG. 5 within the SN. In this way, it is possible to obtain a route hierarchically.

  In the proposed PCS, the shortest path from all the start points to the end points is regularly updated with the algorithm of FIG. 9 in the SN of the upper layer and the Dijkstra algorithm in the SN of the lower layer. For this reason, the call connection delay is reduced as compared with the method of routing for each call setting request such as OSPR and PNNI. In addition, since link state information can be updated in the PCS, there is no need to exchange link state information between routers / nodes. The algorithm in FIG. 9 is named “independent routing” because the SNs in each layer perform routing independently from the SNs in other layers.

REC 1583, OSPF Varsion2.http: //www.fags.org/rfcs/rfc1583.html.

Multi-area MPLS Traffic Engineering draft-kompella-mpls-multiarea-te-04.txt http://www.watersprings.org/pub/id/draft-kompella-mpls-multiarea-te-04.txt

Aiho / Hotcroft / Ullman, Akihiro Nosaka / Kohei Noshita "Algorithm Design and Analysis I" p188. ISBN: N110569563

ATM Forum, “Private Network-Network Interface Specification Version 1.0 (PNNI 1.0)”, March 1996.

Inter-AS MPLS routing by EJB-based path computation server Hiroshi Matsuura, Yasushi Yamanaka, Tatsuro Murakami, NTT, Japan, Kazumasa Takami, Soka University, 9th IFIP / IEEE International Symposium on Integrated Network Management (IM2005)

  When the shortest route from the origin router to the destination router across a plurality of domains is selected using OSPF, a load is concentrated on the origin router or the representative ABR of the origin area when selecting the shortest route. In particular, scenario 2 and scenario 4 need to calculate the representative ABR of the start area, and scenario 3 needs to calculate the shortest path from the start router to the end area using Dijkstra's SPF in FIG. The amount of calculation increases as the number of MPLS / GMPLS domains and routers increases.

  Scenario 1 in the case of selecting the shortest route from the origin router to the destination router across a plurality of domains using OSPF finds the shortest route from the origin router to the ABR in the origin area using only the information in the origin area. However, the selected ABR does not necessarily arrive at the destination router. Similarly, the arrival of the ABR selected in the end point area is not necessarily guaranteed. In that sense, it is necessary to redo the route selection up to M × N (M and N are the number of ABRs in the start / end points, respectively) times.

  In scenarios 2 and 3, the representative ABR or origin router in the origin area calculates the shortest path to the ABR in the endpoint area. However, arrival of the selected end point area from the ABR to the end point router is not guaranteed. In this case, it is necessary to redo the route selection up to N times.

  In scenario 4, there is no need to redo the route selection. However, after the ABR in the start area acquires route information from the ABR in the end area, the shortest path selection with restrictions is performed. For this reason, a lot of time is spent until the shortest route is selected, and more load is applied to the start point area representative ABR and the end point area representative ABR. Further, only the possibility of arrival at the end point router is considered in the representative ABR of the end point area, and routing from the start point router is not performed in consideration of all link states. That is, the optimum route is not always selected, and a biased route may be selected. Also in PNNI, the routing load of the origin node is large. The reason is that since the origin node holds the link state of the PG of each layer, it is necessary to calculate the shortest path for the number of layers.

  Furthermore, when routing is set, resources such as bandwidth are insufficient, and it is highly necessary to select another route and perform routing again. The reason is that the upper logical group node that has received the link state aggregated through the PGL guides the shortest path using its own PG and the aggregated link state. However, the aggregate link state is not necessarily an accurate link state. For this reason, there is a possibility that routing in the upper hierarchy may not be performed accurately. Another reason is that the logical links between upper level logical group nodes (corresponding to lower PGs) are aggregated, and the actual physical link is extended to two sections like the logical link 2626 of FIG. Is treated as one logical link. For this reason, it is because the PNM protocol cannot obtain the optimum which one is substantially selected.

  Furthermore, the optimal path in the end-to-end is not always selected. The reason is that when the origin node selects a route, it refers to only the link state held by itself. For example, when the route from N1 2201 to N10 2210 is selected in FIG. 6, the origin node N1 can perform routing in consideration of the link state information in PG1 2111. However, the link state information of PG3 2113 in which the end node N10 exists is not considered. N12 2402 performs routing considering the link state in PG3, selects BNG 2307, and BNG performs routing up to N10. In this way, when selecting one route, until the route is optimized Can only be done. In this situation, an optimal solution considering end-to-end cannot be obtained.

  The “independent routing” method of FIG. 9 performed by the PCS deployment architecture of FIG. 8 significantly reduces the load on the origin node, which is a problem of inter-domain routing in OSPE and PNNI. This is because the upper layer PCS performs routing between the lower domains (SN) at once, and the lower layer PCS only needs to perform routing within the own domain (SN) using Dijkstra's algorithm. In addition, since links in different sections between domains (SN) are treated as different links compared to PNNI of the same hierarchical routing method, the optimum link between lower domains (SN) reflects the link state information of each link. It is also possible to do this. For this reason, when setting an optical path between lower IP domains in a GMPLS network, it is optimal for obtaining the shortest path between a plurality of IP domains and an optical router. It is also optimal when the lower domain wants to hide its internal link topology from the upper domain. However, if the lower domain issues a request to the upper domain that the inter-domain routing should be performed in consideration of the link state information in the lower domain, the “independent routing” method in FIG. I can't get it. This problem is the same as the problem where it is difficult for OSPF and PNNI to consider end-to-end link state information.

  The present invention has been made paying attention to the above circumstances. The purpose of the present invention is to select link state information of virtual links between nodes possessed by lower layer domains when selecting the shortest path between lower layer domains. In view of the above, it is an object of the present invention to provide a hierarchical distributed routing method and its management device that can perform optimum shortest path selection.

In order to achieve the above object, a first invention includes a plurality of first domains each including a plurality of nodes and constituting a first hierarchy, and a second domain that includes these first domains and is higher than the first hierarchy. A second domain that constitutes a hierarchy, and each of the plurality of first domains is provided with a first routing management device that manages a shortest path list in the domain, and the second domain includes a first domain In a hierarchical distributed routing method used in a network system provided with a second routing management device for managing a route between domains,
The second routing management device acquires the shortest path cost between the boundary nodes included in the first domain from the intra-domain shortest path list managed by the plurality of first routing management apparatuses. The shortest path cost between the boundary nodes is stored as a virtual link cost between the boundary nodes included in the first domain. The shortest path from the start node to the end node is obtained based on the virtual link cost between the held boundary nodes and the inter-domain link cost managed by the second routing management device. is there.
Therefore, according to the present invention, when selecting the shortest path between lower domains, the link state information of logical links (virtual links) between boundary nodes in the lower domains is taken into account, so that the shortest path can be selected with less processing load. Is possible.

  In the first invention, when the shortest path is obtained, it is determined whether the selected last link is a virtual link between the boundary nodes, and when it is determined that the selected short link is the virtual link, It is determined whether or not there is a link connected to another domain from the end node of the virtual link. When there is a link connecting from the selected end node to another domain, the link is determined to be an inter-domain link and is selected.

  In this way, the following effects are exhibited. That is, the virtual link between the boundary nodes in the lower domain has already been selected in the lower domain, and the virtual link between the boundary nodes in the lower domain does not follow the virtual link between the boundary nodes in the lower domain. For this reason, if the route continues further, it can be determined that this route is an inter-domain link. Therefore, the subsequent link can be appropriately selected based on this determination. That is, it becomes possible to select the shortest route by the dependent routing method depending on the link state information of the virtual link between the boundary nodes in the lower domain.

On the other hand, the second invention includes a plurality of first domains each including a plurality of nodes and constituting a first hierarchy, and a second hierarchy including the first domains and constituting a second hierarchy higher than the first hierarchy. Each of the plurality of first domains is provided with a first routing management device that manages a shortest path list in the domain, and a path between the first domains is provided in the second domain. In a hierarchical distributed routing method used in a network system provided with a second routing management device for management,
The second routing management device acquires the shortest path cost between the boundary nodes included in the first domain from the intra-domain shortest path list managed by the plurality of first routing management apparatuses. The shortest path cost between the boundary nodes is held as a virtual link cost between the boundary nodes included in the first domain. At the same time, the second routing management device determines the shortest path between the internal node and the boundary node included in the first domain from the intra-domain shortest path list managed by the plurality of first routing management devices. The route cost is acquired, and the shortest route cost between the acquired internal node and the boundary node is held as a virtual link cost between the internal node and the boundary node included in the first domain. Based on the virtual link cost between the held boundary nodes, the virtual link cost between the held internal node and the boundary node, and the inter-domain link cost managed by the second routing management device Thus, the shortest path from the start node to the end node is obtained.

  Therefore, according to the present invention, in addition to the virtual link cost between the boundary nodes in the lower domain and the virtual link cost between the internal node and the boundary node in the lower domain, the shortest path from the start node to the end node is determined. Selected. For this reason, the shortest path selection can be performed with higher accuracy.

  In the second aspect of the invention, when the shortest path is obtained, it is first determined whether the selected final link is a virtual link between the boundary nodes, and the final link is a virtual link between the boundary nodes. If it is determined, the presence / absence of a link following the other domain from the end node of the virtual link is determined. When there is a link from the selected end node to another domain, the link is determined as an inter-domain link and selected. On the other hand, it is determined whether or not the selected final link is a virtual link between the internal node and the boundary node. When the final link is determined to be a virtual link between the internal node and the boundary node, the virtual link is determined to be the final link and subsequent route selection is stopped.

  In this way, the following effects are exhibited. In other words, since the virtual link between the internal node and the boundary node in the lower domain is used only at the beginning and end of the route, if the shortest route arrives at the virtual link between the internal node and the boundary node in the lower domain It can be determined that there is no subsequent route extension. Therefore, the selection of the shortest path can be accurately terminated based on this determination result. That is, it becomes possible to select the shortest route by the detail dependent routing method depending on the link state information of the virtual link between the internal node and the boundary node of the lower domain.

  In the first and second inventions, the routing policy of the first domain corresponding to the end point of the selected route is determined when obtaining the shortest route. If the routing policy of the first domain corresponding to the end point is the independent routing method as a result of the determination of the routing policy, the first domain corresponding to the end point is selected. On the other hand, as a result of the determination of the routing policy, if the routing policy of the first domain corresponding to the end point is the dependency type or the detail dependency type routing mode, the virtual route starting from the end point node of the selected route is used. It is also characterized by selecting a link.

  In this way, according to the routing policy of the lower domain, it becomes possible to perform the shortest path selection process by selectively using the independent routing method and the dependent or detailed dependent routing method. For this reason, it is possible to always select the shortest route by the optimum routing method according to the routing policy of the lower domain.

  In short, according to the present invention, when selecting the shortest path between lower layer domains, it is possible to select the optimum shortest path in consideration of the link state information of the virtual link between nodes of the lower layer domain. The hierarchical distributed routing method and its management device can be provided.

Embodiments of the present invention will be described below with reference to the drawings.
FIG. 1 is a diagram showing a configuration of a network system for explaining a hierarchical distributed routing method according to the present invention. In the figure, the same parts as those in FIGS. 3 and 8 are denoted by the same reference numerals.

  A network system (TOP SN) 107 according to this embodiment connects six sub-networks (SN A to SN F) 3001 to 3006 as domains and these sub networks (SN A to SN F) 3001 to 3006. Backbone routers (BB R1, BB R2) 401, 402. The sub-networks (SN A to SN F) 3001 to 3006 include intra-area routers (R1 to R10) 201 to 210 and area border routers (BR A to BRG) 301 to 307. The intra-area routers (R1 to R10) 201 to 210 and the area border routers (BR A to BRG) 301 to 307 are connected via links 601 to 611. The area border routers 301 to 307 and the backbone routers (BB R1, BBR2) 401 and 402 are connected via links 701 to 716.

  Incidentally, PCS (Path Computation Servers) 1012 to 1017 as intra-domain routing management devices are installed in the sub-networks (SN A to SN F) 3001 to 3006, respectively. The PCSs 1012 to 1017 as the intra-domain routing management devices respectively manage all intra-area route lists for each router existing in the sub-networks (SN A to SN F) 3001 to 3006. Then, when a starting point and an ending point are given, the shortest route is selected from the intra-area route list.

  The network system 107 including the sub-networks (SN A to SN F) 3001 to 3006 and the backbone routers (BB R1, BB R2) 401 and 402 is provided with a PCS 1011 as an inter-domain routing management device. . The PCS 1011 as the inter-domain routing management device is positioned above the PCSs 1012 to 1017 as the intra-domain routing management device, and manages the inter-area route list.

  When selecting the shortest route between areas, the shortest route cost (link state information) between border routers included in the area is selected from the intra-area route list managed by the PCS 1012 to 1017 as the intra-domain routing management device. And the obtained shortest path cost between the border routers is regarded as a virtual link cost between border routers included in the area. Then, based on the stored virtual link cost between the boundary nodes and the inter-area route list, a process for obtaining the shortest route from the start point to the end point is performed.

  The PCS 1011 includes a CPU (Central Processing Unit), a memory for storing the inter-area route list and the like, and a communication interface for communicating with the PCS 1012 to 1017 and the like, for obtaining the shortest route. Processing is realized by causing the CPU to execute a program.

Next, a hierarchical distributed routing method according to this embodiment will be described based on the above configuration.
FIG. 1 shows an example of link state information passed from the lower PCS 1012 to 1017 to the upper PCS 1011. As described above, the lower PCSs 1012 to 1017 hold lists of the shortest paths between the routers in the corresponding sub-networks (SN A to SN F) 3001 to 3006, respectively. These shortest path lists are periodically obtained by Dijkstra's algorithm shown in FIG. 5, or are regularly obtained by the algorithm shown in FIG. 2 when the SN hierarchy is not the lowest. Then, link state information is passed to the upper PCS 1011 from the obtained shortest path list. This link state information is named “virtual link information”.

  In this embodiment, since the PCSs 1012 to 1017 are arranged for each of the sub-networks (SN A to SN F) 3001 to 3006, the link state information is exchanged between the lower PCS 1012 to 1017 and the upper PCS 1011. However, when all the upper and lower sub-networks (SN A to SN F) 3001 to 3006 are managed by one PCS 1011, the route update in each sub-network (SN A to SN F) 3001 to 3006 is performed. Management is performed for each of the sub-networks (SN A to SN F) 3001 to 3006 in one PCS 1011. Therefore, in this case, in the PCS 1101, the lower SN A 101 to SN F 106 management object passes the “virtual link” to the TOP SN 107 management object.

At the timing of executing the “dependent routing” method, only the virtual link between the border routers in the lower PCS 1012 to 1017 is passed to the upper PCS 1011. For example, in FIG. 1, only the virtual link BRF-BRG 1042 corresponds to this.
At the timing of executing the “detail-dependent routing” method, the virtual link between the border routers in the lower PCS 1012 to 1017 and the virtual link between the border router and the internal router are passed to the upper PCS 1011. For example, in FIG. 1, a virtual link 1031 between R1 201 and BR A 301 and a virtual link 1032 between R2 202 and BR A 301 are passed from the lower PCS 1012. On the other hand, the route information between R1 201 and R2 202 is not passed because it is not necessary for the upper PCS 1011. Similarly, virtual links 1034 and 1035 are passed from the lower PCS 1013, virtual links 1037 and 1038 are passed from the lower PCS 1014, a virtual link 1040 is sent from the lower PCS 1015, a virtual link 1041 is sent from the lower PCS 1016, and from the lower PCS 1017. Are passed virtual links 1042-1046, respectively.

FIG. 2 shows that the SN management object of the upper PCS 1011 changes the “independent routing” method, the “dependent routing” method, and the “detailed dependent routing” method shown in FIG. It is a flowchart which shows the control procedure and control content in the case of selectively performing according to SN routing policy.
In the figure, the processing process surrounded by a solid line indicates a process common to the three methods, and the processing process surrounded by a dotted line indicates a process specific to the dependent / detail dependent routing method. The underlined portion indicates processing specific to the detail-dependent routing method.

  First, in step ST2a, one lower starting subnetwork SN (which may be a router) is selected. In FIG. 1, in the case of a router, the backballota BB R1 401 or BB R2 402 is selected. Next, in the processing process of step ST2b, a link that goes outside from the subnetwork SN (or router) selected in step ST2a is selected. When this process is applied to the network system of FIG. 1, assuming that the origin subnetwork is SN A 3001, the links exiting from SN A are 701, 702, and 704. Here, if the subnetwork selected in step ST2a adopts the detail dependent routing method, the virtual link from the internal router (IR) of the lower subnetwork to the border router is also selected at the same time. The virtual link exiting from the internal router (IR) simultaneously selected in FIG. 1 is the R1-BRA virtual link 1031.

  Subsequently, in the processing process of step ST2c, for the links other than the virtual link starting from the internal router (IR) among the links selected in step ST2b, the router SN at the end of the link is selected as the starting SN selected in step ST2a. If it is already selected as the shortest route from (possibly a router), the route to that link is deleted. When the origin SN adopts the detail-dependent routing method, the router at the end of the link selected in step ST2b has already been selected as the shortest route from the origin router in the origin SN selected in step ST2a. If so, the route to that link is deleted. In the process of step ST2c, it is checked whether or not a route loops between domains by checking whether or not one route passes through the same domain twice. Then, the looping route is deleted.

  Subsequently, in the process of step ST2d, a process of putting a route that has not been deleted in step ST2c into the shortest path candidate list is performed. In step ST2e, it is determined whether or not one or more shortest path candidates are entered in the shortest path candidate list as a result of the processing up to step ST2d. If the result of this determination is that there is no shortest path candidate in the shortest path candidate list, the algorithm is terminated.

  On the other hand, when there is a shortest path candidate in the shortest path candidate list, a candidate with the smallest sum of link costs constituting the route in the shortest path candidate list is selected in the process of step STe. The selected candidate is stored in the shortest path list and deleted from the shortest path candidate list.

  As described in the control procedure (step ST9c) of the independent routing method shown in FIG. 9, the route management is common to the three routing methods, the sub-network SN (which may be a router), the link permutation, It consists of the router information of the route origin and destination. On the other hand, if the destination router is the internal router (IR) of the subnetwork SN adopting the detail dependent routing method, this means that this internal router (IR) is the last node of the route. For this reason, a cost-minimum route is selected again from the shortest path candidate list.

  On the other hand, if the destination router is not the internal router (IR), it is necessary to follow the route ahead from the destination router, so the route stored in the shortest path list is selected in the process of step ST2f. In step ST2g, it is determined whether or not the final link of the route selected in step ST2f is a virtual link. As a result of this determination, if the link is a virtual link, the virtual link is not continuous. For this reason, the router that is the end of the virtual link is selected in step ST2h, the inter-SN link that is outside the subnetwork SN is selected from the selected router, and then the process returns to step ST2c.

  For example, in FIG. 1, it is assumed that the subnetwork SN C 3003 employs an independent routing scheme and the subnetwork SN F 3006 employs a dependent routing scheme. Further, the route selected in the process of step ST2f is (origin router: BR C 303, domain permutation: SN C 3003, 1ink 716, BR G 307 in SN F 3006, virtual link BR F-BRG 1042, SN F Assume that BRF 306 in 3006, destination router: BRF 306). In this case, in step ST2h, two links 712 and 714 that are outside the SN F 3006 are selected from the destination router BR F 306.

  If the final link of the route selected in step ST2g is not a virtual link, the routing policy of the subnetwork SN at the end point of the selected route is determined in step ST2i. As a result of the determination, if the routing policy of the subnetwork SN at the end point is independent, the end point subnetwork SN is selected by the process at step ST2j, and the process returns to the process at step ST2b.

  When a router (for example, BB R1 401, BB R2 402 in FIG. 1) is selected, the router is handled as an independent subnetwork SN. On the other hand, in the case of the dependent or detailed dependent type subnetwork SN instead of the independent type, the process proceeds to step ST2k. In the process of step ST2k, all the virtual links coming out from the last router of the selected route are selected, and the process returns to the process of step ST2c.

  For example, in FIG. 1, it is assumed that the sub-networks SN C 3003 and SN F 3006 are both independent. Further, it is assumed that the route selected in the process of step ST2f is (origin router: BR C 3003, domain permutation: SN C 3003, link 716, SN F 3006, destination router: BRG 307). In this case, the sub-network SN F 3006 is selected in the processing process of step ST2j.

  In FIG. 1, it is assumed that the subnetwork SN C 3003 is an independent type and the SN F 3006 is a dependent type. Further, it is assumed that the route selected in the process of step ST2f is (origin router: BR C 303, domain permutation: SN C 3003, link 716, SN F 3006, destination router: BRG 307). In this case, three virtual links BR F-BR G 1042, BR G-R9 1045, and BR G-R10 1046 exiting from the border router BRG 307 are selected by the processing process in step ST2k.

  As shown in the flowchart of FIG. 2, even when the routing policy of the lower subnetwork SN is mixed, it follows the three routing policies of the lower subnetwork SN, that is, independent routing, dependent routing, and detailed dependent routing. The proposed algorithm can create a possible optimal route between lower SNs.

As described above, this embodiment provides the following operational effects.
(1) When selecting the shortest route by hierarchical distributed routing, the shortest route cost between the border routers of the lower subnetwork from the shortest route list in the area managed by the lower layer PCS 1012 to 1017 by the upper layer PCS 1011 And this is regarded as the virtual link cost. Then, based on the acquired virtual link cost between the boundary nodes and the link cost between the sub-networks managed by the upper layer PCS 1011, the shortest path from the start point to the end point is obtained.
Therefore, when selecting the shortest path between the lower sub-networks, the shortest path can be selected with a small processing load in consideration of the link state information of the virtual link between the boundary nodes in the lower sub-network.

(2) The virtual link between the border routers of the lower subnetwork has already been selected with the optimum route in the lower subnetwork. The virtual link between the border routers of the lower subnetwork is followed by the border router of the lower subnetwork. The virtual link between them doesn't last. Therefore, if the route continues further, it can be determined that this route is an intersubnetwork link.
Therefore, paying attention to this point, when determining the shortest path, in step ST2g, it is determined whether or not the selected final link is a virtual link between the boundary nodes. In this case, a link connecting from the end node of the virtual link to another domain is determined as an inter-subnetwork link and is selected.
As a result, it is possible to appropriately select the subsequent link, thereby enabling the shortest path selection by the dependent routing method based on the link state information of the virtual link between the border routers of the lower subnetwork.

(3) When selecting the shortest route by hierarchical distributed routing, the shortest route cost between the border routers of the lower subnetwork is selected from the shortest route list in the area managed by the lower layer PCS 1012 to 1017 by the upper layer PCS 1011. , And the shortest path cost between the internal router and the border router is acquired, and these are regarded as virtual link costs. Then, based on the obtained virtual link cost between the border nodes, the shortest path cost between the internal router and the border router, and the link cost between the sub-networks managed by the upper layer PCS 1011, the start point to the end point The shortest route to is obtained.
Accordingly, in addition to the virtual link cost between the border routers of the lower subnetwork, the shortest path from the start point to the end point is selected in consideration of the virtual link cost between the internal router and the border router of the lower subnetwork. For this reason, it becomes possible to perform the shortest path selection with higher accuracy.

(4) Since the virtual link between the internal router and the border router of the lower subnetwork is used only at the beginning and end of the route, the shortest path is provided for the virtual link between the internal router and the border router of the lower subnetwork. When it arrives, it can be determined that there is no subsequent route extension.
Therefore, paying attention to this point, when selecting the shortest path, it is first determined whether or not the selected final link is a virtual link between the border routers, and the final link is a virtual link between the border routers. When it is determined, the presence / absence of a link following the other subnetwork from the destination router of the virtual link is determined. When there is a link from the selected end router to another subnetwork, the link is determined as an intersubnetwork link and selected. Meanwhile, it is determined whether the selected final link is a virtual link between the internal router and the border router, and the final link is determined to be a virtual link between the internal router and the border router. In this case, the virtual link is determined to be the final link, and subsequent route selection is stopped.
As a result, the selection of the shortest route can be accurately terminated based on the above judgment result, and the detailed dependency type can be selected according to the link state information of the virtual link between the internal router and the border router of the lower subnetwork. The shortest route selection by the routing method becomes possible.

(5) When obtaining the shortest route, the routing policy of the sub-layer in the lower layer corresponding to the end point of the selected route is determined. If the routing policy of the lower layer subnetwork corresponding to the end point is the independent routing method as a result of the determination of the routing policy, the lower layer subnetwork corresponding to the end point is selected. On the other hand, if the routing policy of the lower-layer subnetwork corresponding to the end point is the dependency type or the detailed dependency type routing mode as a result of the determination of the routing policy, the end point node of the selected route is set as the starting point. A virtual link is selected.
In this way, the shortest path selection process can be performed by selectively using the independent routing method and the dependent or detailed dependent routing method according to the routing policy of the sub-layer of the lower layer. For this reason, the shortest path selection by the optimal routing method according to the routing policy of the sub-network domain of the lower layer is possible.

  The present invention is not limited to the above embodiment. For example, in the above-described embodiment, the case where the domain is configured in two layers and the PCS as the routing management device is arranged in each of the layers has been described as an example. However, the domain hierarchy may be set to three or more layers, and a PCS as a routing management device may be arranged for each of these layers. At that time, it is also possible for a plurality of PCSs distributed in a hierarchy to manage a route list across a plurality of hierarchies.

Further, the weights for the lower-domain route may be adjusted by variably setting the coefficients of the lower-domain route cost acquired from the lower domain and the inter-domain route cost managed by the upper domain.
In addition, the configuration of the network system, the configuration of the sub-layer of the lower layer, the configuration and function of the routing management device provided in each layer, the procedure and contents of the hierarchical distributed routing control, etc., do not depart from the scope of the present invention Various modifications can be made.

  In short, the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage. Further, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, you may combine suitably the component covering different embodiment.

The figure which shows the structure of the network system which implement | achieves one Embodiment of the hierarchy distributed routing method concerning this invention. The flowchart which shows the procedure and content of the shortest path | route selection control which used 3 systems together by the routing management apparatus (PCS) provided between the systems shown in FIG. The figure which shows the structure of the network system for demonstrating the conventional shortest path selection system which employ | adopted the OSPF protocol. The figure which shows an example of the shortest path | route tree obtained by the conventional shortest path | route selection system in the system shown in FIG. The flowchart for demonstrating the algorithm of Dijkstra which is the conventional shortest path selection system which employ | adopted the OSPF protocol. The figure which shows the structure of the network system for demonstrating the conventional shortest path | route selection system which employ | adopted the PNNI protocol. The figure which shows an example of the aggregate link topology information of PG which PGL produces in the system shown in FIG. The figure which shows the structure of the network system for demonstrating the conventional hierarchical interdomain routing method. The flowchart which shows the procedure and content of the shortest path selection control by the independent routing system employ | adopted in the system shown in FIG.

Explanation of symbols

  101-106 ... OSPF area, 201-210 ... internal router, 301-307 ... area border router, 401-402 ... backbone router, 601-611 ... intra-area link, 701-716 ... inter-area link, 801-810 ... distance (Link cost), 2101 to 2106 ... lowest hierarchy PG, (logical group node in the 2nd hierarchy PG), 2101 to 2103 ... PGL in each PG of the 2nd hierarchy, 2111 to 2113 ... 2nd hierarchy PG (most Logical group nodes in the upper hierarchy PG), 2201 to 2210 ... internal nodes in the lower PG of the PNNI, 2201, 2032, 2205, 2207, 2208, 2209 ... PGL in each lowest PG, 2301 to 2307 ... lower in the PNNI Boundary nodes in PG, 2401 to 2402 ... directly to PG1 and PG3 Internal node to which it belongs, 2601 to 2611 ... logical link in the lowest hierarchy PG, 2621 to 2627 ... logical link in the second hierarchy PG, 2631 to 2633 ... logical link in the highest hierarchy PG, 2701 ... aggregated link topology of PGF Information center, 2702, 2703 ... Output port from aggregate link topology information to node outside PGF, 2704, 2705 ... Default spoke from aggregate link topology information to each port, 2706 ... Exception bypass in aggregate link topology information, 1011 1017 ... PCS arrangement example for routing management between multiple domains, 1018 ... PCS distribution server, 3001 to 3006 ... Subnetwork (SN), 1021 to 1023 ... Route list elements held by upper layer PCS, 1031 to 1047 ... Lower layer PCS Holding route list.

Claims (10)

A plurality of first domains each including a plurality of nodes and constituting a first hierarchy; and a second domain comprising these first domains and constituting a second hierarchy higher than the first hierarchy, A second routing management device that provides a first routing management device that manages a route list in a domain in each of a plurality of first domains, and that manages a route list between the first domains in the second domain A hierarchical distributed routing method used in a network system comprising:
The second routing management device acquires a shortest route cost between boundary nodes included in the first domain from a route list in a domain managed by the plurality of first routing management devices;
The second routing management device holds the acquired shortest path cost between the boundary nodes as a virtual link cost between the boundary nodes included in the first domain;
The second routing management device obtains the shortest path from the origin node to the destination node based on the virtual link cost between the held boundary nodes and the inter-domain link cost managed by the own device. A hierarchical distributed routing method comprising: The process of obtaining the shortest path is:
Determining whether the selected final link is a virtual link between the border nodes;
When it is determined that the final link is a virtual link between the boundary nodes, a process of determining the presence or absence of a link connected to another domain from the end node of the virtual link;
2. The hierarchical distributed type according to claim 1, further comprising: a step of determining, when there is a link connected from the selected end node to another domain, that the link is an inter-domain link and selecting the link. Routing method. A plurality of first domains each including a plurality of nodes and constituting a first hierarchy; and a second domain comprising these first domains and constituting a second hierarchy higher than the first hierarchy, A first routing management device that manages a route list in a domain is provided in each of the plurality of first domains, and a second routing management device that manages a route between the first domains is provided in the second domain. A hierarchical distributed routing method used in a network system provided,
The second routing management device acquires a shortest route cost between boundary nodes included in the first domain from a route list in a domain managed by the plurality of first routing management devices;
The second routing management device holding the acquired shortest path cost between the boundary nodes as a virtual link cost between the boundary nodes included in the first domain;
The second routing management device obtains the shortest route cost between the internal node and the boundary node included in the first domain from the route list in the domain managed by the plurality of first routing management devices. The process of
The second routing management device holds the acquired shortest path cost between the internal node and the boundary node as a virtual link cost between the internal node and the boundary node included in the first domain. Process,
The second routing management device includes a virtual link cost between the held boundary nodes, a virtual link cost between the held internal node and the boundary node, and an inter-domain link cost managed by the own device. And a step of obtaining a shortest path from the start node to the end node based on the hierarchical distributed routing method. The process of obtaining the shortest path is:
Determining whether the selected final link is a virtual link between the border nodes;
When it is determined that the final link is a virtual link between the boundary nodes, a process of determining the presence or absence of a link that continues from the end node of the virtual link to another domain;
When there is a link that continues from the selected end node to another domain, the link is determined as an inter-domain link and is selected.
Determining whether the selected final link is a virtual link between the internal node and a border node;
A step of determining a virtual link as a final link and stopping subsequent route selection when the final link is determined to be a virtual link between the internal node and the boundary node. The hierarchical distributed routing method according to claim 3. The process of obtaining the shortest path is:
Determining a routing policy of a first domain corresponding to an end point of the selected route;
If the routing policy of the first domain corresponding to the end point is the independent routing mode as a result of the determination of the routing policy, the step of selecting the first domain corresponding to the end point;
As a result of the determination of the routing policy, if the routing policy of the first domain corresponding to the end point is the dependency type or the detail dependency type routing mode, a virtual link starting from the end point node of the selected route is selected. 5. The hierarchical distributed routing method according to claim 2, further comprising: a process. A plurality of first domains each including a plurality of nodes and constituting a first hierarchy; and a second domain comprising these first domains and constituting a second hierarchy higher than the first hierarchy, An intra-domain routing management device for managing a route list in the domain is provided in each of the plurality of first domains, and an inter-domain routing management device for managing a route between the first domains is provided in the second domain. In the inter-domain routing management device used in the network system,
Means for obtaining a shortest route cost between boundary nodes included in the first domain from a route list in a domain managed by the plurality of intra-domain routing management devices;
Means for recognizing the shortest path cost between the acquired boundary nodes as a virtual link cost between the boundary nodes included in the first domain;
Means for obtaining a shortest path from a start node to an end node based on the recognized virtual link cost between the boundary nodes and the inter-domain link cost managed by the inter-domain routing management device. Inter-domain routing management device. The means for obtaining the shortest path is:
Means for determining whether the selected final link is a virtual link between the border nodes;
When it is determined that the final link is a virtual link between the boundary nodes, means for determining the presence or absence of a link connected to another domain from the end point node of the virtual link;
The inter-domain device according to claim 6, further comprising means for determining that the link is an inter-domain link when there is a link connecting from the selected end node to another domain. Routing management device. A plurality of first domains each including a plurality of nodes and constituting a first hierarchy; and a second domain comprising these first domains and constituting a second hierarchy higher than the first hierarchy, An intra-domain routing management device for managing a route list in the domain is provided in each of the plurality of first domains, and an inter-domain routing management device for managing a route between the first domains is provided in the second domain. In the inter-domain routing management device used in the network system,
Means for obtaining a shortest path cost between boundary nodes included in the first domain from a path list in the domain managed by the plurality of intra-domain routing management devices;
Means for holding the acquired shortest path cost between boundary nodes as a virtual link cost between boundary nodes included in the first domain;
Means for obtaining a shortest path cost between an internal node and a boundary node included in the first domain from a path list in a domain managed by the plurality of intra-domain routing management devices;
Means for holding the obtained shortest path cost between the internal node and the boundary node as a virtual link cost between the internal node and the boundary node included in the first domain;
Based on the virtual link cost between the retained boundary nodes, the virtual link cost between the retained internal node and the boundary node, and the inter-domain link cost managed by the own device, the source node to the end node An inter-domain routing management device, comprising: The means for obtaining the shortest path is:
Means for determining whether the selected final link is a virtual link between the border nodes;
Means for determining the presence or absence of a link connected to another domain from the end node of the virtual link when the final link is determined as a virtual link between the boundary nodes;
Means for selecting and selecting the link as an inter-domain link when there is a link from the selected end node to another domain;
Means for determining whether the selected final link is a virtual link between the internal node and a border node;
And a means for determining that the virtual link is the final link when the final link is a virtual link between the internal node and the boundary node, and stopping subsequent route selection. The inter-domain routing management device according to claim 6. The means for obtaining the shortest path is:
Means for determining a routing policy of the first domain corresponding to the end point of the selected route;
If the routing policy of the first domain corresponding to the end point is the independent routing mode as a result of the determination of the routing policy, means for selecting the first domain corresponding to the end point;
As a result of the determination of the routing policy, if the routing policy of the first domain corresponding to the end point is the dependency type or the detail dependency type routing mode, a virtual link starting from the end point node of the selected route is selected. The inter-domain routing management apparatus according to claim 7 or 9, further comprising: means.

Images