Efficiency-improving device, efficiency-improving method, and efficiency-improving program

Abstract

In a storage unit (21B) of a leaf (21), a route table (T11) for registering a destination address and an NO table (T12) for registering next hop information are managed in association with each other. A processing unit (21A) of the leaf (21): receives a command which is issued when a failure occurs and which is for deleting a table related to a plurality of next hop information items of an ECMP, and a command which is for creating a table related to next hop information items that remain without becoming unavailable; and deletes next hop information items that become unavailable from the NO table (T12) in which the plurality of next hop information items of the ECMP are registered.

Description

Efficiency improvement device, efficiency improvement method, and efficiency improvement program

[0001] The present invention relates to an efficiency improvement device, an efficiency improvement method, and an efficiency improvement program.

[0002] A leaf-and-spine network is a network consisting of two layers of nodes, leaf and spine, and is used in data centers. The leaf on the access side is not directly connected to other leafs, but is connected to other leafs via the spine on the core side. Then, traffic load balancing can be achieved by forwarding traffic from the leaf to multiple spines using ECMP (Equal Cost Multi Path), which distributes traffic among multiple equal-cost routes.

[0003] FIG. 9 is a configuration diagram of a network system 910 in which ECMP is configured. In the following network diagrams, such as FIG. 9, the IP address of each node is also listed, such as the IP address of leaf 21z = [1.1.1.3]. In the network system 910 when the path is normal, ECMP is configured between leaves 21z-23z and spines 11z and 12z, and traffic load is distributed. For example, ECMP is configured between leaf 21z and spines 11z and 12z, and between leaf 22z and spines 11z and 12z (shown by bold lines). Traffic from leaf 21z to leaf 22z is load-balanced between a first path passing through leaf 21z → spine 11z → leaf 22z and a second path passing through leaf 21z → spine 12z → leaf 22z. Meanwhile, the path from leaf 21z to spine 12z is configured as a single path, not an ECMP.

[0004] When a path failure occurs, a state transition 901 occurs from a network system 910 in an ECMP environment to a network system 920 in a non-ECMP environment. In the network system 920 during the path failure, a failure 921 occurs between the leaf 22z and the spine 11z, so traffic from the leaf 21z to the leaf 22z cannot be forwarded via the spine 11z. Therefore, traffic is forwarded only via the second path passing through the leaf 21z, the spine 12z, and then the leaf 22z (i.e., via a single path, not ECMP). When the path is restored, a state transition 902 occurs from the network system 920 to the network system 910. This allows ECMP to function again.

[0005] FIG. 10 is an explanatory diagram showing the data structure of the tables used by each node in FIG. 9. Each node manages routing-related information in the following two types of databases: - Routing Information Base (RIB) is a database of IP routes. RIB registers routes for each destination address obtained using routing protocols such as Border Gateway Protocol (BGP) and Open Shortest Path First (OSPF) as nexthop objects (NOs) in two tables (ROUTE_TABLE, NEXTHOP_GROUP_TABLE). - Forwarding Information Base (FIB) is a database of IP forwarding. Of the registered RIB routes, the route that is determined to be optimal and used for forwarding is registered in FIB as a route object (RO).

[0006] The FIBs of leaves 21z-23z are managed by separating NO and RO using NG (Nexthop Group) as shown in Figure 10. NG employs a hierarchical data structure, such as the first layer "route group nexthop" → second layer "nexthop group nexthop[N]" → third layer "nexthop dev gateway". This provides the following benefits: - There is no need to store nexthop information for each route, and nexthop information can be aggregated, which leads to reduced memory consumption. - Because NO and RO are separated, only the object to be updated needs to be updated, which leads to shorter update times.

[0007] C. Hopps NextHop Technologies, "Analysis of an Equal-Cost Multi-Path Algorithm," [online], [Retrieved August 23, 2023], Internet <URL: https: / / datatracker.ietf.org / doc / html / rfc2992>, November 2000

[0008] Below, we will explain the problem of packet loss that occurs when a route fails (state transition 901 in FIG. 9) and when a route is restored (state transition 902 in FIG. 9) by referring to each table in leaf 21z (FIG. 9) shown in FIGS. 11 to 13.

[0009] 11 is an explanatory diagram showing the table state of the network system 910 when the route is normal. The leaf 21z manages the combination of "route table T11z → NO table T12z" and the combination of "route table T21z → NO table T22z." The arrow from the route table T11z to the NO table T12z indicates that the route table T11z and the NO table T12z are linked by the same ID (=ID97). The route table T11z describes how the next hop (next forwarding destination) toward the destination address "1.1.1.4" of the leaf 22z is obtained from the NO table T12z with "Nexthop_group=ID97." The NO table T12z registers an ECMP consisting of two routes, the interface toward the spine 11z being 1.1.1.1 (Ethernet0) and the interface toward the spine 12z being 1.1.1.2 (Ethernet1), as the next hop corresponding to "Nexthop_group=ID97." The route table T21z is described as obtaining the next hop toward the destination address "1.1.1.2" of the spine 12z from the NO table T22z of "Nexthop_group=ID98." The NO table T22z registers one route, the interface toward the spine 12z being 1.1.1.2 (Ethernet1), as the next hop corresponding to "Nexthop_group=ID98."

[0010] 12 is an explanatory diagram showing the table state of the network system 920 when a route failure occurs. The leaf 21z updates each table in FIG. 11 as follows using the routing algorithm of the routing protocol: Deletes the NO table T12z. Updates the route table T11z (ID97) to the route table T11x (ID98), switching the linked destination to the NO table T22z.

[0011] 13 is an explanatory diagram showing the table states of the network system 910 when a route is restored. The leaf 21z updates each table in FIG. 12 as follows using the routing algorithm of the routing protocol: - Adds a NO table T12y (ID 103) that has the same entries as the NO table T12z. - Updates the route table T11x (ID 98) to the route table T11y (ID 103), switching the linked destination to the NO table T12y.

[0012] FIG. 14 is a sequence diagram illustrating a path failure. This sequence diagram starts from the state illustrated in FIG. 11 and ends when the state illustrated in FIG. 12 is reached. A controller (not illustrated) includes a processing unit for the RIB 51z. The leaf 21z includes a RIB setting unit 52z that sets information instructed by the RIB 51z, a RIB storage unit 53z that stores the RIB 51z, a FIB storage unit 54z that stores the FIB 55z, and a processing unit for the FIB 55z. The RIB 51z transmits a NO deletion instruction (S211) to the RIB setting unit 52z to delete the NO table T12z. The RIB setting unit 52z executes data deletion (NO deletion) in accordance with the NO deletion instruction (S211) in each unit (the RIB storage unit 53z, the FIB storage unit 54z, and the FIB 55z) within the leaf 21z (S212, S213, S214).

[0013] The RIB 51z sends an RO deletion instruction (S221) to the RIB setting unit 52z to delete the route table T11z (ID97). The RIB setting unit 52z executes data deletion (RO deletion) in accordance with the RO deletion instruction (S221) in each unit within the leaf 21z (S222, S223, S224). The RIB 51z sends an RO addition instruction (S231) to the RIB setting unit 52z to add the route table T11x (ID98). The RIB setting unit 52z executes data addition (RO addition) in accordance with the RO addition instruction (S231) in each unit within the leaf 21z (S232, S233, S234). The above-described RO deletion process (S221-S224) and RO addition process (S231-S234) are repeatedly executed the number of times equal to the number of routes associated with NO.

[0014] 15 is a sequence diagram for path recovery. This sequence diagram starts from the state shown in FIG. 12 and ends when the state shown in FIG. 13 is reached. The RIB 51z sends a NO addition instruction (S311) to the RIB setting unit 52z to add a NO table T12y (ID103). The RIB setting unit 52z adds data (NO addition) to each section in the leaf 21z in accordance with the NO addition instruction (S311) (S312, S313, S314).

[0015] The RIB 51z sends an RO deletion instruction (S321) to the RIB setting unit 52z to delete the route table T11x (ID98). The RIB setting unit 52z executes data deletion (RO deletion) in accordance with the RO deletion instruction (S321) in each unit within the leaf 21z (S322, S323, S324). The RIB 51z sends an RO addition instruction (S331) to the RIB setting unit 52z to add the route table T11y (ID103). The RIB setting unit 52z executes data addition (RO addition) in accordance with the RO addition instruction (S331) in each unit within the leaf 21z (S332, S333, S334).

[0016] Here, the period from the completion of RO deletion (S224) to the completion of RO addition (S234) in Fig. 14 is a blank period in which there is no route table for information directed to the destination address "1.1.1.4". Similarly, the period from the completion of RO deletion (S324) to the completion of RO addition (S334) in Fig. 15 is also a blank period for the destination address "1.1.1.4". In this way, during the transition period between an ECMP environment such as the network system 910 and a non-ECMP environment such as the network system 920, packet loss occurs due to the blank period in the table supporting the destination address "1.1.1.4" in the leaf 21z.

[0017] Therefore, a main object of the present invention is to reduce packet loss during a transition period between an ECMP environment and a non-ECMP environment.

[0018] In order to solve the above problems, the efficiency improvement device of the present invention has the following features: The present invention is characterized by comprising: a storage unit that manages, in association with each other, a first table that registers a destination address and a second table that registers multiple pieces of next hop information as ECMPs, for a route to a destination address via a next hop indicated by the next hop information; and a processing unit that receives an instruction to delete the table related to the multiple pieces of next hop information of the ECMP, which is issued when some of the multiple pieces of next hop information of the ECMP become unavailable due to a failure, and an instruction to create a table related to the remaining next hop information that has not become unavailable, and deletes the next hop information that has become unavailable from the second table in which the multiple next hop information of the ECMP is registered.

[0019] According to the present invention, packet loss during a transition period between an ECMP environment and a non-ECMP environment can be reduced.

[0020] 10 is a configuration diagram of a network system in which ECMP is set according to the present embodiment. FIG. 11 is a configuration diagram of a leaf in FIG. 1 according to the present embodiment. FIG. 12 is a hardware configuration diagram of each device in the network system according to the present embodiment. FIG. 13 is a configuration diagram showing a set of tables stored in a RIB storage unit according to the present embodiment. FIG. 14 is a sequence diagram at the time of a route failure according to the present embodiment. FIG. 15 is an explanatory diagram showing the contents of messages exchanged in the sequence diagram of FIG. 5 according to the present embodiment. FIG. 16 is a sequence diagram at the time of route recovery according to the present embodiment. FIG. 17 is an explanatory diagram showing the contents of messages exchanged in the sequence diagram of FIG. 18 according to the present embodiment. FIG. 19 is a configuration diagram of a network system in which ECMP is set. FIG. 10 is an explanatory diagram showing the data structure of tables used by each node in FIG. 10. FIG. 11 is an explanatory diagram showing the table state of a network system when a route is normal. FIG. 12 is an explanatory diagram showing the table state of a network system when a route has failed. FIG. 13 is an explanatory diagram showing the table state of a network system when a route has been restored. FIG. 14 is a sequence diagram at the time of a route failure. FIG. 15 is a sequence diagram at the time of route recovery.

[0021] Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings.

[0022] FIG. 1 is a configuration diagram of a network system 110 in which ECMP is configured. As in FIG. 9, in network diagrams such as FIG. 1, the IP address of each node is also listed, such as the IP address of leaf (efficiency device) 21 = [1.1.1.3]. When the route is normal, the network system 110 has ECMP configured between leaves 21-23 and spines 11 and 12, and traffic is load-balanced. For example, ECMP is configured between leaf 21 and spines 11 and 12, and between leaf 22 and spines 11 and 12 (shown by bold lines). Then, traffic from leaf 21 to leaf 22 is load-balanced between a first route passing through leaf 21 → spine 11 → leaf 22 and a second route passing through leaf 21 → spine 12 → leaf 22. On the other hand, the route from leaf 21 to spine 12 is configured as a single route, not an ECMP.

[0023] When a path failure occurs, a state transition 101 occurs from the network system 110 in the ECMP environment to the network system 120 in the non-ECMP environment. In the network system 120 when the path failure occurs, a failure 121 occurs between the leaf 22 and the spine 11, so traffic from the leaf 21 to the leaf 22 cannot be forwarded via the spine 11. Therefore, traffic is forwarded only via the second path that passes through the leaf 21, the spine 12, and the leaf 22 (i.e., via a single path, not ECMP). When the path is restored, a state transition 102 occurs from the network system 120 to the network system 110. This allows ECMP to function again.

[0024] 2 is a configuration diagram of the leaf 21 in FIG. 1. The controller 30 of the network system 110 has a processing unit for the RIB 51 in order to manage each node of the network system 110. The controller 30 grasps the latest network status, such as the fault 121 in FIG. 1, and controls the state transitions 101 and 102. The leaf 21 has a processing unit 21A and a storage unit 21B. The processing unit 21A includes an RIB setting unit 52 that sets information instructed by the RIB 51, and an FIB 55. The storage unit 21B includes an RIB storage unit 53 that stores the RIB 51, and an FIB storage unit 54 that stores the FIB 55. The RIB 51 may be included in the processing unit 21A of the leaf 21.

[0025] 2, the components of leaf 21 have been described as a representative, but the other nodes, leaves 22 and 23 and spines 11 and 12, have the same configuration. Unlike the conventional technology shown in FIGS. 9 to 15, which deletes the entire table related to failures in RIB storage unit 53, this embodiment responds to environmental changes (state transitions 101 and 102) at the time of route failure and route recovery by changing the values ​​in the table related to failures in RIB storage unit 53 (see FIG. 4 for details). As a result, there is no change in the NO associated with the RO, reducing the time for packet loss.

[0026] FIG. 3 is a hardware configuration diagram of each device in the network system 110. Each node (leaf 21-23 and spine 11, 12) and controller 30 in the network system 110 is configured as a computer 800 having a CPU 801, RAM 802, ROM 803, HDD 804, communication I / F 805, input / output I / F 806, and media I / F 807. The communication I / F 805 is connected to an external communication device 815. The input / output I / F 806 is connected to an input / output device 816. The media I / F 807 reads and writes data from a recording medium 817. Furthermore, the CPU 801 controls each unit by executing a program (efficiency program) loaded into RAM 802. This program (also called an application, or an app for short) can be distributed via a communication line or recorded on a recording medium 817 such as a CD-ROM.

[0027] FIG. 4 is a configuration diagram showing a set of tables stored in the RIB storage unit 53. The table set 110T stored in the RIB storage unit 53 of the leaf 21 corresponds to the network system 110 of FIG. 1 and is for when the route is normal (before a failure occurs and after recovery from the failure). The table set 120T stored in the RIB storage unit 53 of the leaf 21 corresponds to the network system 120 of FIG. 1 and is for when a route failure occurs. State transitions 101 and 102 are as shown in FIG. 1. The table set 110T for when the route is normal is, as in FIG. 11, a combination of "route table (first table) T11 → NO table (second table) T12" for the destination address "1.1.1.4" of the leaf 22 and a combination of "route table T21 → NO table T22" for the destination address "1.1.1.2" of the spine 12. In the NO table T12, when the route is normal, two routes are registered as ECMPs: the interface to the spine 11 is 1.1.1.1 (Ethernet0), and the interface to the spine 12 is 1.1.1.2 (Ethernet1).

[0028] In the table set 120T at the time of a route failure, the processing unit 21A (RIB setting unit 52) ​​of the leaf 21 does not delete an existing table or add a new table to the table set 110T at the time of a normal route, unlike in Fig. 12. Instead, in the event of a route failure, the RIB setting unit 52 updates the data to delete 1.1.1.1 (Ethernet0) from the NO table T12 so that one route (non-ECMP) is registered, the interface being 1.1.1.2 (Ethernet1) toward the spine 12 that can bypass the failure.

[0029] 5 is a sequence diagram for when a path failure occurs. This sequence diagram starts from the state of table set 110T in FIG. 4 and ends when the state reaches table set 120T. The RIB 51 sends a NO deletion instruction (S111A) to the RIB setting unit 52 of the leaf 21 to delete the NO table T12. The RIB setting unit 52 does not expand the NO deletion instruction (S111A) to each unit in the leaf 21 (the RIB storage unit 53, the FIB storage unit 54, and the FIB 55), and leaves the NO table T12. The RIB 51 sends an RO deletion instruction (S121A) to the RIB setting unit 52 of the leaf 21 to delete the route table T11 linked to the NO table T12 that was selected for deletion in the NO deletion instruction (S111A). The RIB setting unit 52 does not expand the RO deletion instruction (S121A) to each unit in the leaf 21 either, and leaves the route table T11.

[0030] The RIB 51 transmits an RO addition instruction (S131A) to the RIB configuration unit 52 of the leaf 21 to add a new route table to be used as a switching destination for the route table T11 that was targeted for deletion in the RO deletion instruction (S121A). The RIB configuration unit 52 receives the RO addition instruction (S131A) and performs the following processing instead of adding a new route table. The RIB configuration unit 52 acquires the source RO (route table T11) and its ID (=ID97) specified in the RO deletion instruction (S121A) from the RIB storage unit 53 (S141A). The RIB configuration unit 52 acquires the NO table T22 linked to the new route table to be used as a switching destination specified in the RO addition instruction (S131A) and its ID (=ID98) from the RIB storage unit 53 (S142A). The RIB setting unit 52 instructs each unit in the leaf 21 to change the data content of the NO table T12 linked to the route table T11 at the switching source (S151A, S152A, S153A). This instruction changes the data content of the NO table T12 from the content at the switching source (NO table T12 of table set 110T, two paths) to the content at the switching destination (NO table T12 of table set 120T, one path, which is the same as the NO table T22 acquired in S142A).

[0031] As described above with reference to FIG. 5 , the memory unit 21B (RIB storage unit 53) of the leaf 21 manages a route table T11 that registers destination addresses for routes to destination addresses via next hops indicated by the next hop information, and a NO table T12 that registers multiple next hop information as ECMPs (Equal Cost Multi Paths). The processing unit 21A (RIB setting unit 52) ​​of the leaf 21 receives from the RIB 51 commands (S111A, S121A) to delete a table related to multiple ECMP next hop information, which are issued when some of the ECMP next hop information becomes unavailable due to a failure, and a command (S131A) to create a table related to the remaining next hop information that has not become unavailable. The processing unit 21A deletes the unavailable next hop information from the NO table T12 in which the ECMP next hop information is registered (S151A).

[0032] As a result, each table in the RIB setting unit 52 of the leaf 21 becomes a table set 120T. In this way, even if a data rewrite instruction is issued three times from the RIB 51 (S111A, S121A, S131A), the RIB setting unit 52 consolidates the three instructions into a single data rewrite instruction (S151A) so that the three instructions are reflected, instead of executing them as they are. Therefore, even in the event of a route failure, there is no blank period in the tables (route table T11, NO table T12) that support the destination address "1.1.1.4" in the leaf 21, so packet loss can be prevented.

[0033] FIG. 6 is an explanatory diagram showing the contents of messages exchanged in the sequence diagram of FIG. 5. Message 211 is an example of an NO deletion instruction (S111A) and describes the contents (two routes) of the NO table T12 to be deleted. Message 212 is an example of an RO addition instruction (S131A) and describes the ID (ID98) of the new route table to be switched to. Message 213 is a message for acquiring the existing (source) route table T11 from the RIB configuration unit 52 (S141A). Message 214 is a message for acquiring the NO table T22 and its ID for S142A from the RIB configuration unit 52. Message 215 is a message for the RIB configuration unit 52 to instruct each component in the leaf 21 to change the contents of the NO acquired in message 214 (one route) without changing the NO ID (ID97) (S151A).

[0034] FIG. 7 is a sequence diagram for path recovery. This sequence diagram starts from the state of table set 120T in FIG. 4 and ends when the state of table set 110T is reached. The RIB 51 sends a NO addition instruction (S111B) to the RIB setting unit 52 of the leaf 21 to add a NO table T12. The RIB setting unit 52 does not deploy the NO addition instruction (S111B) to each unit in the leaf 21 (the RIB storage unit 53, the FIB storage unit 54, and the FIB 55), but instead reuses the NO table T12 that was not deleted by the NO deletion instruction (S111A) in FIG. 5. The RIB 51 sends an RO deletion instruction (S121B) to the RIB setting unit 52 of the leaf 21 to delete the route table specified in the RO addition instruction (S131A) in FIG. 5. The RIB setting unit 52 does not expand the RO deletion instruction (S121B) to each unit in the leaf 21 either, and leaves the route table T11.

[0035] The RIB 51 transmits an RO addition instruction (S131B) to the RIB configuration unit 52 of the leaf 21 to add a new route table to be used as a switching destination for the route table T11 that was selected for deletion in the RO deletion instruction (S121B). The RIB configuration unit 52 receives the RO addition instruction (S131B) and, instead of adding a new route table, performs the following processing: The RIB configuration unit 52 acquires the source RO (route table T11) and its ID (=ID97) specified in the RO deletion instruction (S121B) from the RIB storage unit 53 (S141B). The RIB configuration unit 52 acquires the NO table T12 linked to the route table T11 acquired in S141B and its ID (=ID97) from the RIB storage unit 53 (S142B). The RIB setting unit 52 instructs each unit in the leaf 21 to change the data content of the NO table T12 in S142B (S151B, S152B, S153B). By this instruction, the data content of the NO table T12 is changed to the content of the message 221 (FIG. 8) specified in the NO addition instruction (S111B).

[0036] As described above with reference to FIG. 7 , the memory unit 21B (RIB storage unit 53) of the leaf 21 manages a route table T11 that registers destination addresses for routes to destination addresses via next hops indicated by the next hop information, and a NO table T12 that registers some next hop information for ECMPs, in association with each other. The processing unit 21A (RIB setting unit 52) ​​of the leaf 21 receives from the RIB 51 commands to delete a table related to some next hop information for ECMPs (S111B, S121B), which are issued when some of the next hop information for ECMPs becomes available due to failure recovery, and a command to create a table related to multiple next hop information available as ECMPs (S131B). The processing unit 21A adds the available next hop information to the NO table T12 in which the part of the next hop information for ECMPs is registered (S151B).

[0037] As a result, each table in the RIB setting unit 52 of the leaf 21 returns to the state of the table set 110T. In this way, even if a data rewrite instruction has been issued three times from the RIB 51 (S111B, S121B, S131B), the RIB setting unit 52 consolidates the three instructions into a single data rewrite instruction (S151B) so that the three instructions are reflected, instead of executing them as they are. Therefore, even when the route is restored, there is no blank period in the tables (route table T11, NO table T12) that support the destination address "1.1.1.4" in the leaf 21, so packet loss can be prevented.

[0038] FIG. 8 is an explanatory diagram showing the contents of messages exchanged in the sequence diagram of FIG. 7. Message 221 is an example of an NO addition instruction (S111B) and describes the contents (two routes) of the NO table T12 to be added. Message 222 is an example of an RO addition instruction (S131B) and describes the ID (ID100) of the new route table to be switched to. Message 223 is a message for obtaining the existing (source) route table T11 from the RIB configuration unit 52 (S141B). Message 224 is a message for obtaining the NO table T12 and its ID for S142B from the RIB configuration unit 52. Message 225 is a message for the RIB configuration unit 52 to instruct each component in the leaf 21 to change the contents (two routes) of the NO specified in message 221 without changing the NO ID (ID97) (S151B).

[0039] [Effect] The leaf 21 of the present invention is characterized by having a storage unit 21B that manages, in association with each other, a route table T11 that registers a destination address and a NO table T12 that registers multiple pieces of next hop information as ECMP (Equal Cost Multi Path) for a route to a destination address via a next hop indicated by the next hop information, and a processing unit 21A that receives an instruction to delete the table related to multiple pieces of ECMP next hop information, which is issued when some of the multiple pieces of ECMP next hop information becomes unavailable due to a failure, and an instruction to create a table related to the remaining next hop information that does not become unavailable, and deletes the next hop information that has become unavailable from the NO table T12 in which the multiple pieces of ECMP next hop information are registered.

[0040] As a result, the distribution device 10 can omit the process of deleting or adding the entire table (route table T11, NO table T12) when an ECMP route failure occurs, thereby eliminating blank periods in the table and reducing the occurrence of packet loss.

[0041] The leaf 21 of the present invention is characterized in that the processing unit 21A receives an instruction to delete a table relating to the next hop information that remains available and is issued when some of the multiple next hop information of the ECMP becomes available due to failure recovery, and an instruction to create a table relating to the multiple next hop information that can be used as ECMP, and adds the next hop information that has become available to a second table in which the next hop information that remains available and is registered.

[0042] As a result, when an ECMP route is restored, the distribution device 10 can omit the process of deleting or adding the entire table (route table T11, NO table T12), thereby eliminating blank periods in the table and reducing the occurrence of packet loss.

[0043] 11, 12 Spine 21-23 Leaf (efficiency device) 30 Controller 51 RIB 52 RIB setting unit 53 RIB storage unit 54 FIB storage unit 55 FIB 21A Processing unit 21B Storage unit T11, T21 Route table (first table) T12, T22 NO table (second table) 110, 120 Network system

Claims

1. An efficiency improvement device comprising: a memory unit that links and manages a first table that registers a destination address and a second table that registers multiple next hop information as ECMP (Equal Cost Multi Path) for a route to a destination address via a next hop indicated by the next hop information; and a processing unit that receives an instruction to delete the table related to the multiple next hop information of the ECMP, which is issued when some of the multiple next hop information of the ECMP becomes unavailable due to a failure, and an instruction to create a table related to the remaining next hop information that is not unavailable, and deletes the next hop information that has become unavailable from the second table in which the multiple next hop information of the ECMP is registered.

2. The efficiency device of claim 1, wherein the processing unit receives an instruction to delete a table relating to the remaining next hop information of the ECMP that has not become unavailable, which is issued when some of the multiple next hop information of the ECMP becomes available due to failure recovery, and an instruction to create a table relating to the multiple next hop information that is available as the ECMP, and adds the next hop information that has become available to the second table in which the remaining next hop information that has not become unavailable is registered.

3. An efficiency improvement method comprising the steps of: an efficiency improvement device having a memory unit and a processing unit, wherein the memory unit manages a first table that registers a destination address and a second table that registers multiple next hop information as ECMP (Equal Cost Multi Path) for a route to a destination address via a next hop indicated by the next hop information; and wherein the processing unit receives an instruction to delete the table related to the multiple next hop information of the ECMP, which is issued when some of the multiple next hop information of the ECMP becomes unavailable due to a failure, and an instruction to create a table related to the remaining next hop information that has not become unavailable, and deletes the next hop information that has become unavailable from the second table in which the multiple next hop information of the ECMP is registered.

4. An efficiency improvement program for causing a computer to function as the efficiency improvement device according to claim 1 or 2.

Images