KR20060025471A - Apparatus and method for transmitting forwarding information base in distributed structure - Google Patents

Abstract

In the present invention, the distributed structure forwarding information base delivery apparatus and method, the network service module (NSM) selects the least cost route from the routing information base table including all possible routes of the network, and stores in the form of forwarding information base table, At least one routing cache module (RCM) arranged in a distributed structure receives forwarding information base entry information from the network service module and transmits it to a forwarding plane.

Description

Apparatus and Method for Transmitting Forwarding Information Base in Distributed Structure}             

1 is a centralized RCM configuration diagram.

2 is a block diagram of a system to which the present invention is applied.

3 is a distributed RCM configuration table.

4 is a block diagram of a distributed RCM.

5 is a flow chart of an RCM configuration procedure.

6 is a distributed RCM structure diagram after some NSM initialization processes are performed.

7 is a preferred embodiment of the RCM configuration flow.

8 is a preferred embodiment of a distributed RCM structure after the RCM configuration procedure is completed.

9 is a preferred embodiment of an RCM configuration table of RCM0;

10 illustrates one preferred embodiment of an RCM configuration table of RCM1.

11 is a preferred embodiment of an RCM configuration table of RCM4.

* Description of the symbols for the main parts of the drawings *                 

100: Route Cache Module (RCM)

110: NSM (Network Service Module)

200: interface blade 210: routing module

220: Switch 31: RCM ID

32: IP Address 33: Master Field

34: Slave field 35: Hello field

36: State field

The present invention relates to a distributed structure forwarding information base delivery apparatus and method for introducing a distributed structure to deliver a forwarding information base entry to a forwarding plane in a device equipped with a routing function.

In general, routing refers to setting up a route at a router. The router has a cable with the address of the network it is connected to and always keeps the correct information. When a packet arrives at the router, it calculates the optimal route to the destination by looking at the table address and transmits the packet accordingly. That is, the router must examine the destination address and subnet mask of the forwarded packet, compare the source IP address with the subnet mask, and decide whether to forward the packet to the internal or external network.

Types of routing depend on how the router builds a routing table, which uses static routing and routing protocols in which administrators enter routes for the router to use for routing. Dynamic Routing to build

The routing algorithm can be roughly divided into a distance vector algorithm (DVA) and a link state algorithm (LSA).

The distance vector algorithm calculates the minimum number of hops by the routing table. Whenever a node is added or removed from the network, the distance vector algorithm sends the changed routing information of the neighboring network to all routers in the network, which may cause unnecessary traffic. have. Suitable for small networks.

A representative example of the distance vector algorithm is RIP (Routing Information Protocol). RIP is a protocol proposed to determine the optimal network path and was developed for use in XNS (Xerox Network System). It has the characteristics of keeping the entire routing table and standardized to RFC1085.

The link state algorithm is an algorithm that controls the routing process in real time. When the network changes, the link state algorithm transmits the changed information only to the router of the related network, thereby reducing network overhead. It is ideal for routing environments that are complex to carry large amounts of information such as hop count, line speed, and traffic.                         

The link state algorithm is Open Shortest Path First (OSPF). OSPF is a protocol for determining routing paths in TCP / IP networks and was developed to compensate for the shortcomings of RIP. OSPF has the advantage of delivering minimal changes between routers and is standardized in RFC 1583.

As discussed, the route for delivering traffic can be obtained in a number of ways. The Routing Information Base contains all routes identified from each routing protocol. The Forwarding Information Base (FIB) selects an optimal route among several routes included in the routing information base and is used to forward traffic. In other words, unlike the routing information base including all routing information, the forwarding information base includes only a unique path and does not include a lane path.

The configuration of the system in which the route selection process as described above is performed will be described.

1 illustrates a centralized RCM configuration.

The forwarding element (FE) is radiated to the center of the central routing cache module (RCM), and the RCM is connected to the network service module (NSM).

The FE may be expressed as a forwarding engine or a forwarding plane, and serves as a subject of packet transmission.

The Network Service Module (NSM) is a module that manages RIB tables and selects routes that require the least cost. The Route Cache Module (RCM) linked with the NSM maintains the FIB entry information selected from the RIB table and delivers the information to the FE.

In the structure as shown in FIG. 1, since the route selected by the NSM is delivered to all the FEs through only one RCM, an unreliable forwarding information is configured in the forwarding plane because the RCM at the center is abnormal or fails, or the RCM transitions to an abnormal state. If this is the case, there will be a problem in forwarding traffic.

In other words, when FIB entries are forwarded to multiple forwarding planes that forward traffic through a centralized RCM, the overall system performance depends on the processing power of the RCM, so the RCM becomes a bottleneck of the entire system. .

The present invention provides a distributed structure forwarding information base delivery apparatus and method for introducing a distributed RCM configuration rather than a centralized RCM configuration to deliver the forwarding information base entry to the forwarding plane stably and effectively. The purpose.

In a distributed structure forwarding information base delivery method according to an aspect of the present invention, a network service module (NSM) selects a route of least cost from a routing information base table including all possible routes of a corresponding network, and stores the minimum cost route in a forwarding information base table format. And receiving, by the at least one routing cache module (RCM) arranged in a distributed structure, forwarding information base entry information from the network service module and transmitting the forwarding information base entry information to the forwarding plane.

The distributed structure of the routing cache module may include at least one other routing cache module connected to the top-level master routing cache module as a first-stage dependent module, starting from a top-level master routing cache module, and each of the first-stage dependent modules. Is connected with at least one second stage routing cache module to serve as a master module, and the upper subordinate module stage interconnects as a master module to the lower dependent module stage until the lowest dependent module stage is configured. A repeating tree structure is formed, but one dependent routing cache module can set only one routing cache module as a master.

In addition, the number of subordinate routing cache modules connected to the one master routing cache module may be set and changed by an administrator in consideration of the traffic distribution degree and the transmission speed.

Each routing cache module includes one RCM configuration table including at least one item of a Routing Cache Module (RCM) ID, a master field, a dependent field, a hello field, and a status field.

The master field indicates whether the routing cache module of the corresponding RCM ID is the master RCM for the RCM having the RCM configuration table, and the dependent field indicates whether the routing cache module of the corresponding RCM ID is the dependent RCM for the RCM having the RCM configuration table. Characterized in that represents.

In the distributed structure forwarding information base delivery method according to another aspect of the present invention, a routing cache module of one of the routing cache modules sends an activation message to all remaining routing cache modules, and each routing that receives the activation message is provided. Transmitting, by the cache module, a registration request message to the dispatch routing cache module, and the dispatch routing cache module that receives the registration request message corresponds to a registration limit number of the at least one registration request message in the order in which the registration request message is received. And sending a registration success response message to the registration request routing cache module and transmitting a registration failure response message to the registration request routing cache module exceeding the registration limit. Master routing If the route cache module that does not have a break module, by repeating the registration response message transmitting step and the registration request message transmitting step to form the dispersion structure of the routing cache module.

The dispatch routing cache module of the registration response message transmitting step may set a registration request routing cache module that transmits the registration success response message as its subordinate routing cache module.

The registration request routing cache module receiving the registration success response message may further include setting the dispatch routing cache module as its master routing cache module.

The distributed structure forwarding information base delivery device according to another aspect of the present invention is a network service module (NSM) that selects a route of the least cost from a routing information base table including all possible routes of the corresponding network and stores it in the form of a forwarding information base table (NSM). And at least one routing cache module (RCM) that receives forwarding information base entry information from the network service module and transmits it to a forwarding plane, wherein the routing cache module is arranged in a distributed structure.

Hereinafter, a preferred embodiment according to the present invention will be described in detail with reference to the drawings.

2 shows a configuration of a system to which the present invention is applied.

The system to which the present invention is applied will be basically any device or system equipped with a routing function, and in the case of FIG. 2, a routing module 210, an interface blade 200, and a switch ( A system consisting of a switch 220 is shown. Both the routing module 210 and the plurality of interface blades 200 are connected to the switch.

The routing module 210 represents a processor board in which a routing protocol is mounted and operated. The routing module has a built-in Network Service Module (NSM) 110 that manages a Routing Information Base (RIB) table and interworks with a plurality of RCMs 100. This configuration distinguishes the NSM from communicating with one RCM in a centralized RCM configuration.                     

Each interface blade 200 includes an RCM 100, which is a forwarding entry management block for maintaining and delivering FIB entry information. The interface blade 200 may be understood as a board that forwards packets.

The RIB table managed by the NSM is stored as a forwarding information base in the form of a look-up table for performing actual forwarding at the interface blade.

3 shows a distributed RCM configuration table.

The RCM ID 31 is an RCM Identifier uniquely assigned to the corresponding RCM in the system. The RCM having an RCM ID 31 value of 0 receives FIB entry information from the NSM. The Internet Protocol (IP) address 32 is the address assigned to the processor of the board on which RCM is enabled. It is used as a destination address for Interprocess Communication (IPC) communication between a Master RCM and a Slave RCM.

The Master (33) field indicates whether the corresponding RCM is a master RCM or a dependent RCM. The master RCM means the RCM from which the FIB entry is downloaded. The slave 34 field indicates whether the RCM of the corresponding RCM ID is the slave RCM with respect to the self RCM. Dependent RCM means RCM to which FIB entry should be downloaded.

The Hello 35 field indicates the test period in seconds of the Hello message to test connectivity. The State 36 field indicates status information of the corresponding RCM. The "SELF" indicates that the RCM of the RCM ID field is itself, and the "SYNC" indicates that the RCM is in sync with another RCM. FULL "indicates that synchronization is complete. "NC" indicates Not Concerned.

The table of FIG. 3 is configured in all RCMs, and maintains RCM information for transmitting and receiving FIB entries with itself, and a logical distributed tree (Tree) such as a master RCM and a dependent RCM configuration for stable and efficient FIB entry transmission and reception. ) Construct the structure.

4 shows the configuration of the distributed RCM.

In the distributed structure of FIG. 4, unlike the centralized RCM configuration of FIG. 1, it can be seen that the RCM is distributed to form a tree structure.

Each RCM can be a dependent RCM or a master RCM for other RCMs. This is determined according to the connection structure and shape, and as shown in FIG. 4, three subordinate RCMs need not be connected to one master RCM. The number of subordinate RCMs connected to one master RCM can be set differently by the administrator according to the characteristics of the network and the flow of traffic.

4 illustrates a state in which all connections between RCMs are made, and the connection process between these RCMs can be seen through FIG. 5.

5 shows an RCM configuration procedure flow.

First, the RCM0 having the RCM ID 0 transmits an activation message (INIT) to all RCMs configured in the system at initialization, and each RCM that receives the activation message transmits a registration request message (JOIN) to the RCM0 (S51). Upon receiving the registration request message from the plurality of RCMs, the RCM0 registers the registration request RCM as the dependent RCM in the RCM table according to the order in which the registration request message is received by a preset number of RCM registration restrictions. In this case, the maximum number of dependent RCMs that can be accepted by one RCM may be changed by the user, and may be set in consideration of the RCM distribution degree, processing capacity, and the like.

The RCM0, which is the master RCM, transmits a registration success response message (REG_OK) to the RCM registered as the dependent RCM in the RCM table (S52), and transmits a registration failure response message (REG_FAIL) to the RCM that is not registered in the RCM table. (S53). The RCM that receives the registration success response message sets the master field of RCM0 to 1 in the RCM table and indicates that RCM0 is the master RCM.

The distributed RCM structure after the NSM initialization procedure is performed is shown in FIG. 6.

6 shows a distributed RCM structure after some NSM initialization procedure.

Only the RCM that receives the registration response message from the NSM is connected to the master RCM and configures the same FIB table as the FIB table configured in the master RCM.

In the embodiment shown in FIG. 6, three subordinate RCMs are connected using RCM0 as the master RCM. The remaining RCMs of the distributed architecture are RCMs that are not connected to the RCM0 because they do not receive a registration success response message from the RCM0, which is the master RCM, during the RCM configuration as shown in the embodiment of FIG. 5. These RCMs are not connected to any other RCMs other than RCM0, and RCMs that do not have a master field of their own RCM configuration table.

RCMs that are not connected make certain attempts to establish a connection with other RCMs, which is well illustrated in FIG. 7.

Figure 7 illustrates one preferred embodiment of the RCM configuration flow.

RCMs in which the master field of the RCM configuration table is not transmitted, transmit a registration request message (JOIN) to all RCMs present in the system configuration information (S71). In FIG. 7, the case of RCM5 is taken as an example.

In the case of the RCM in which the registration limit number is not exceeded among the plurality of RCMs receiving the registration request (JOIN) message from the RCM5, the registration response message REG_OK is transmitted to the RCM5 (S72). In the case of Figure 7 RCM1 and RCMn is shown to be RCM does not exceed the registration limit number. RCM1 and RCMn also set the dependent field of RCM5 to "YES" in their own RCM configuration table to indicate that RCM5 is its dependent RCM.

On the other hand, in the case of the RCM exceeding the RCM registration limit number of its own, the registration failure (REG_FAIL) response message is transmitted to the RCM5 that has sent the registration request message (S73). In FIG. 7, it can be seen that the RCMm exceeds the RCM registration limit number. The number of RCM registration limits can be arbitrarily set by the administrator, and has already been determined in consideration of load distribution according to the RCM distribution degree and traffic delay time due to the increase of elapsed nodes on the path.

Receiving a registration success (REG_OK) response message from one RCM, RCM5 sets the master field of the RCM to "YES" in its own RCM configuration table, and subsequently receives a registration success (REG_OK) response message. The LEAVE message is transmitted to the RCM (S74).

That is, RCM5 registers itself as its master RCM only for the RCM that sent the first successful registration (REG_OK) response message, and sends a LEAVE message for the remaining RCMs that have sent the second and subsequent messages. It is. Here, it can be seen that each RCM can register only one RCM as its master RCM.

In the embodiment of FIG. 7, only the RCM n of the RCM 1 and the RCM n having transmitted the registration success (REG_OK) response message to the RCM5 is transmitted. This means that a registration success (REG_OK) response message transmitted from RCM1 arrives first in RCM5, and it can be seen that RCM1 is selected as the master RCM of RCM5. Therefore, RCM5 has no choice but to send a LEAVE message for RCMn.

The RCMn that has received the LEAVE message from the RCM5 should release the dependent field setting for the RCM5 of the RCM configuration table that was performed when the registration success (REG_OK) message was transmitted to the RCM5.

FIG. 8 illustrates a preferred embodiment of a distributed RCM structure after the RCM configuration procedure of FIG. 7 as described above is completed.

Many RCMs form a tree structure around RCM0. RCM0 has RCM1, RCM2, and RCM3 as dependent RCMs, and RCM1, which is a dependent RCM for RCM0, has a status as a master RCM for RCM4, RCM5, and RCM6. As such, each RCM has the status of a subordinate RCM for one RCM and a master RCM for another RCM, forming a continuous configuration of RCMs.

In the example of FIG. 8, three subordinate RCMs are connected to one master RCM, but the number of subordinate RCMs does not need to be three, which can be easily determined by those skilled in the art. It can be said that the level.

9, 10, and 11 show examples of the FIB table held by each RCM in the RCM configuration. Configuration tables such as the tables of FIGS. 9 to 11 are held by each RCM, and will be described in detail in what type of configuration table each RCM has.

9 illustrates one preferred embodiment of an RCM configuration table of RCM0.

The configuration table held by each RCM contains information about all the RCMs that make up the RCM structure. That is, all RCMs are represented in the RCM configuration table as RCM distinguishing factors represented by the RCM ID 31.

RCM0 has a plurality of subordinate RCMs, but does not have a master RCM connected to the subordinate RCM, so there is no RCM ID in which the master 33 field is set to “YES”. The dependent 34 field is set to "YES" for RCM1, RCM2, and RCM3, indicating that RCM0 has RCM1, RCM2, and RCM3 as dependent RCMs.

It is recalled that the 'Hello' 35 field is a cycle in which the corresponding RCM tests connectivity with the counterpart RCM. Since RCM0 is itself, no connectivity test is performed, so the 'Hello' (35) field is set to '0'. It can be seen that the connectivity test is performed at intervals of 3 seconds with RCM1, RCM2, and RCM3, which are dependent RAMs of RCM0. On the other hand, the status 36 field for subordinate RCMs is "FULL", indicating that all synchronization between the RCM0 and subordinate RCMs is completed.

Since the connection with the RCM4 is not established, both the master 33 field and the subordinate 34 field are set to "NO". Since the connection state is not checked, the 'Hello' 35 field is also set to "0". State 36 of RCM4 is represented by " NC " which translates to 'Not Concerned' that there is no association with RCM0.

Figure 10 illustrates one preferred embodiment of the RCM configuration table of RCM1.

In FIG. 10, since the state 36 field of the RCM1 is set to "SELF", it can be seen that the RCM configuration table held by the RCM1.

RCM1, unlike RCM0, has both a master RCM and a dependent RCM. Referring to FIG. 10, it can be seen that the master RCM of RCM1 is RCM0 and the dependent RCMs are RCM4, RCM5, and RCM6.

In addition, it can be seen that RCM1 communicates with its master RCM and subordinate RCM, is not connected to the remaining RCMs, and of course, no connectivity test is performed.

Figure 11 shows one preferred embodiment of the RCM configuration table of RCM4.

11 shows that the state 36 field of the RCM4 is set to "SELF", so that it is an RCM configuration table held by the RCM4.

RCM4 has RCM1 with master 33 field set to "YES" as master RCM and RCM7, RCM8 and RCM9 with subordinate 34 field set to "YES" as subordinate RCM. The remaining RCMs in the RCM configuration table are shown to have no association with RCM4.

You can see that the 'Hello' field of RCM4 and slave RCM are set to 3 and connection test is performed every 3 seconds. Their status 36 is "Full" and the synchronization with RCM4 is completed. have.

As described above, the FIB table held by each RCM has a constant form but the contents are all different, which is distinguished from a routing information table (RIB) containing the same contents and the same contents.

The present invention improves the existing centralized FIB delivery method and provides an effect of traffic load distribution and stability in FIB delivery by using an RCM distribution structure.

Claims (12)

In the traffic forwarding method of a system equipped with a routing function, The network service module (NSM) selecting a least cost route from a routing information base table including all possible routes of the network and storing the route in the form of a forwarding information base table; Wow At least one routing cache module (RCM) arranged in a distributed structure, receiving forwarding information base entry information from the network service module and transmitting the forwarding information base entry information to a forwarding plane. The method of claim 1, Distributed structure of the routing cache module, Starting from a top-level master routing cache module, at least one other routing cache module connects to the top-level master routing cache module as a first-stage dependent module, Each of the first stage subordinate modules is connected to at least one or more second stage routing cache modules to serve as a master module. Until the lowest dependent module stage is configured, the upper dependent module stage forms a tree structure that repeats the process of interconnecting as the master module to the lower dependent module stage. The distributed routing forwarding information base delivery method, wherein one subordinate routing cache module can set only one routing cache module as a master. The method of claim 1, The number of subordinate routing cache modules connected to the one master routing cache module may be set and changed by an administrator in consideration of a traffic distribution degree and a transmission speed. The method of claim 1, Each routing cache module, And a RCM configuration table including at least one of an RCM ID, a master field, a dependent field, a hello field, and a status field. The method of claim 4, wherein The master field indicates whether the routing cache module of the corresponding RCM ID is a master routing cache module for the routing cache module having the RCM configuration table. And wherein the dependent field indicates whether the routing cache module of the corresponding RCM ID is the dependent routing cache module with respect to the routing cache module having the RCM configuration table. The method of claim 4, wherein And wherein the hello field indicates a test period of a message for testing connectivity between two routing cache modules. A traffic forwarding method of a system having at least one routing cache module (RCM) for receiving and forwarding forwarding information base entry information including an optimal route in a network to a forwarding plane, the method comprising: Sending, by one routing cache module of the routing cache module, an activation message for all remaining routing cache modules; Each routing cache module receiving the activation message sends a registration request message to the dispatch routing cache module; And The dispatch routing cache module that has received the registration request message transmits a registration success response message to the corresponding registration request routing cache module in the order in which the registration request message is received by the limited number of registration request messages. The registration request routing cache module for exceeding the registration limit number includes a registration response message transmission step of transmitting a registration failure response message, If there is a routing cache module that does not have a master routing cache module even after the above three steps, distributed structure forwarding information forming a distributed structure of the routing cache module by repeating the registration request message transmission step and the registration response message transmission step. Bass delivery method. The method of claim 7, wherein The dispatch routing cache module of the registration response message transmission step, And a registration request routing cache module for transmitting the registration success response message as its subordinate routing cache module. The method of claim 7, wherein The registration request routing cache module that receives the registration success response message, Setting the dispatch routing cache module as its master routing cache module. In the traffic forwarding device of a system equipped with a routing function, A network service module (NSM) which selects the least cost route from the routing information base table including all possible routes of the network and stores the route of the least cost in the form of a forwarding information base table; and At least one routing cache module (RCM) for receiving the forwarding information base entry information from the network service module and transmits to the forwarding plane, The routing cache module is distributed structure forwarding information base delivery device, characterized in that arranged in a distributed structure. The method of claim 10, Distributed structure of the routing cache module, Starting from a top-level master routing cache module, at least one other routing cache module connects to the top-level master routing cache module as a first-stage dependent module, Each of the first stage subordinate modules is connected to at least one or more second stage routing cache modules to serve as a master module. Until the lowest dependent module stage is configured, the upper dependent module stage forms a tree structure which repeats the process of interconnecting as the master module to the lower dependent module stage. A distributed routing forwarding information base delivery device, characterized in that one subordinate routing cache module can set only one routing cache module as a master. The method of claim 10, The number of subordinate routing cache modules connected to the one master routing cache module may be set and changed by an administrator in consideration of a traffic distribution degree and a transmission speed.

Images