CN1744567A - A Method of Smooth Restart of Router Realizing Fast Routing Convergence - Google Patents

Abstract

Before being restarted smoothly, router generates an Opaque LSA 'flooding' in own domain to inform other router that self is ready to restart smoothly. After receiving the Opaque LSA, remote routers mark links relevant to the smooth restarted router as status indication of 'potential unavailable', but no operation is executed. Till receiving Opaque LSA indicating that smooth restarting state is ended from the router, remote routers deletes the said status indication. But, if in restarting, synchronizing or ending phases, change of link status happens in network, then remote routers will switches status indications of links relevant to the smooth restarted router from 'potential unavailable' to 'unavailable' as well as updates information base of link status and recalculates routes immediately, and remote routers makes the said corresponding action unnecessary waiting for a unavailable LSA of relevant links sent from router adjacent to the smooth restarting router.

Description

A Method of Smooth Restart of Router Realizing Fast Routing Convergence

technical field

The invention relates to a method for stably restarting a router for realizing rapid route convergence, and belongs to the technical field of data communication.

Background technique

When the link state changes, the convergence time of the traditional routing protocol on the IP network is generally more than ten seconds. During the convergence period, routing loops and black holes will occur, resulting in incorrect delivery of data packets. With the development of technology, IP networks are carrying more and more real-time services such as voice and video. These services are obviously different from the previous services such as e-mail and web browsing. The quality of service (QoS) requirements and the tolerable service interruption time It is usually expected to be within 1 to 2 seconds, so the convergence time of the IP network must be accelerated. The ways to achieve fast routing convergence mainly include: shortening the fault detection time, adjusting the default timer parameters of routing protocols, optimizing routing calculation algorithms, etc. Under the current technical level, the fast convergence time of routing protocols in a typical network can reach about 1 second .

The purpose of routing protocol graceful restart (Graceful Restart) is to enhance the stability of IP network. The current mainstream routers have realized the separation of the control module and the data forwarding module. When the software upgrade needs to be restarted, the normal forwarding of the data can not be affected, but the support of the corresponding routing protocol is required. IETF has formulated some related standards.

Currently, commonly used routing protocols include: OSPF (OpenShortest Path First IGP), IS-IS (Intermediate System-to-intermediate System), and BGP (Border Gateway Protocol). There are corresponding smooth restart mechanisms, and the working principles are generally similar. Here OSPF is taken as an example (see Figure 1) for illustration.

Step (1) Handshake phase: Router X that needs to restart smoothly and the adjacent router Y send handshake information to each other, X tells Y: it will enter the process of smooth restart, and the length of time required for restart;

Step (2) Restart stage: restart router X smoothly and enter the control software restart stage. In this stage, although adjacent router Y does not receive any OSPF protocol packets to restart router X, it still considers X to be normal and sends information to the network normally. Refresh and restart the relevant status of router X in other routers;

Step (3) synchronization stage: the control software of restarting router X returns to normal, and requests link state information from adjacent router Y, and Y sends all link state advertisement LSAs in its own state information database to restarting router X, so as to realize the two The link state information databases of routers X and Y are synchronized with each other;

Step (4) end stage: restart router X after receiving the LSA sent by adjacent router Y, through comparison, after finding that it is synchronized with the state information database of adjacent router Y, tell adjacent router Y: the smooth restart process ends normally;

Step (5) Normal stage: restart router X and adjacent router Y have established adjacency relationship, respectively enter the standard OSPF interactive state.

The above is the normal smooth restart process. If the link status somewhere in the entire network changes during this period, the adjacent router Y will immediately exit the auxiliary state of smooth restart and immediately announce the link between itself and the restarting router X. The link is unavailable. The purpose of this is to prevent routing loops and black holes in the network.

Refer to Figure 2, which introduces the routing convergence process when the link state changes in the current smooth restart mechanism:

First, after router B detects the failure of the link on the B-C segment, it issues a link status update, floods the LSA (B-C fail) to the corresponding network area in a "flooding" manner, and notifies that the link on the B-C segment is unavailable. The so-called "flooding" is a standardized processing process stipulated in the OSPF protocol, which is used for publishing and synchronizing link state information databases between routers. After "flooding", except for the restarting router X that happened to be in the process of restarting at that time, all other routers in the network area have generated a completely consistent new link state information base, and use the new information base to recalculate Routing and forwarding table. Because the restart router X was in the process of restarting at that time, its routing process had stopped working, so it could not understand that the link in the B-C segment had failed, that is: X still retained the previous link state information base and corresponding routing forwarding table. In this way, each router in the network area has an asynchronous link state information database. According to the working principle of the link state routing protocol, this situation may cause routing loops and black holes.

In order to avoid the generation of routing loops and black holes, the standard stipulates that after receiving the LSA (B-C fail), the adjacent router Y connected to the restarting router X needs to immediately generate an LSA (X-Y fail) in addition to "flooding" it normally. ), and "flood" it out, declaring that the X-Y segment link is unavailable. In this way, other routers can remove the relevant link of restarting router X from the link state information database, thereby realizing the synchronization of all active routers in the network, and avoiding the occurrence of loops and black holes.

However, the result of this processing is that other routers in the network will receive two consecutive link state updates in a short period of time. Take the restart of router A as an example: because the adjacent router Y only generates LSA (B-Cfail) after receiving the LSA (B-Cfail). LSA (X-Y fail), so in most cases, restarting router A will receive LSA (B-C fail) first. In the fast route convergence mechanism, restarting router A will immediately update the link state information database and recalculate the routing table However, during the calculation process or just after the calculation, it will receive LSA ((X-Y fail), so that the restart router A will need to update the link state information base and recalculate the routing table again.

If multiple routers in the network are in the process of graceful restart at the same time, a single link status change somewhere will cause each router to receive multiple different link status messages consecutively in a short period of time. In the case of route convergence, these routers may update their link state information databases and recalculate routing tables many times, which affects the speed of route convergence. This issue has aroused the attention and attention of people in the industry.

Contents of the invention

In view of this, the purpose of the present invention is to provide a method for smooth restart of a router that realizes rapid route convergence. In the prior art, routers update their link state information databases and recalculate routing tables multiple times, which affects the routing convergence speed.

In order to achieve the above-mentioned purpose, the present invention provides a kind of method that realizes the router steady restart of routing fast convergence, comprises following operation stages: handshake, restart, synchronization, end and return to normal; It is characterized in that: before the restart of described process, Add an announcement stage: before the router restarts smoothly, it first generates an opaque link state announcement Opaque LSA that is "flooded" in the area, and notifies other routers in the network that it is ready to restart smoothly; the remote router receives After the Opaque LSA, the relevant link of the graceful restart router is marked as "potentially unavailable", but no other related operations are performed; the remote router will not delete until the remote router receives the OpaqueLSA that the smooth restart router has ended the graceful restart state A "potentially unavailable" state flag for the graceful restart router link; however, if a link state change occurs on the network during the restart, synchronization, or end phases of the graceful restart process, the remote router Mark the relevant links of the graceful restart router from the "potentially unavailable" state to the "unavailable" state, and update the link state information base and recalculate the route without waiting for the received Corresponding processing will be performed only after the relevant link is unavailable LSA.

The "flooding" of the smooth restart router into the local area is an opaque link state advertisement Opaque LSA. The Opaque LSA is an optional LSA type defined by the IETF standardization organization in the standard RFC2370, and is used to carry the shortest path Prioritize the relevant information required by the optional functions of the interior gateway protocol OSPF; since the relevant information required by the optional functions is only carried in the Opaque LSA, changes to the standard LSA are avoided to maximize the protection of the standard OSPF routing protocol functions stability.

The link state type value of the opaque link state announcement Opaque LSA is 10, and its effective "flooding" range is the network area (Area-Local), that is, the Opaque LSA information is only between routers in the network area Delivery and diffusion, but not delivery to other network areas.

Including the following specific steps:

(1) Handshake phase: each router sends handshake information to each other, and the smooth restart router tells the adjacent router that it will enter the smooth restart process;

(2) Announcement stage: the smooth restart router generates an opaque link state announcement Opaque LSA that "floods" in the local area, and notifies other routers in the network that it is ready to perform a smooth restart; the remote router receives the opaque LSA After the opaque link state advertises Opaque LSA, the relevant link of the smooth restart router is marked as "potentially unavailable", but the follow-up operations of updating the link state information base and recalculating the route are not performed;

(3) Restart stage: The graceful restart router enters the control software restart stage. At this stage, the adjacent routers have not received any data packets of the smooth restart router’s Shortest Path First Interior Gateway Protocol OSPF, but still think that the smooth restart router is working Normal, and refresh the relevant standard link state announcement Standard LSA information of the graceful restart router to other routers in the network normally, so as to be compatible with the standard graceful restart protocol;

(4) Synchronization phase: the control software of this smooth restart router returns to normal, and requests link state information from adjacent routers, and adjacent routers send all standard link state announcement Standard LSA information in their own state information databases to this smooth restart router , to realize the synchronization of the link state information base of the smooth restart router and the adjacent router;

(5) End stage: after the smooth restart router receives the LSA sent by the adjacent router, after comparing and finding that it has synchronized with the state information database of the adjacent router, it tells the adjacent router that the smooth restart state has ended normally; at the same time, The smooth restart router will also generate an opaque link state announcement Opaque LSA that "floods" in the area, and announces to other routers in the network that it has ended the smooth restart state; the remote router deletes the Opaque LSA after receiving the Opaque LSA A previously established status flag of "potentially unavailable" for the graceful restart router link;

(6) Return to normal stage: the smooth restart router establishes an adjacency relationship with the adjacent router, and the remote router also learns that the smooth restart router has returned to normal, and the whole network enters the standard shortest path first interior gateway protocol OSPF interactive state.

The step (3) restart stage, (4) synchronization stage, or (5) end stage further includes the following operations: if a link state change occurs in the network, the remote router will immediately reset the relevant link of the smooth restart router. The link state changes from "potentially unavailable" to "unavailable", and the link state information base is updated and the route is recalculated.

The method allows traditional routers that can only recognize and process standard OSPF routing protocols and routers that support the Opaque LSA to form a hybrid network.

Because the traditional router does not support the present invention, there may be multiple routing calculations; and it discards after receiving the Opaque LSA, which may affect the "flooding" of the Opaque LSA that restarts the router smoothly, so that the Opaque LSA cannot be "flooded". ” to all routers in this area, causing some routers that support the Opaque LSA to fail to start the corresponding extension mechanism because they cannot receive the corresponding Opaque LSA, and their behavior is equivalent to that of traditional routers; therefore, the method suggested that all routers Can support Opaque LSA.

The method is applicable to the Shortest Path First Interior Gateway Protocol OSPF, Intermediate System Routing Protocol IS-IS, or Border Gateway Protocol BGP.

The present invention is a method for smooth restart of routers that realizes rapid route convergence. The method is realized by extending the existing smooth restart mechanism. The end router also knows that a certain router in the network is in the process of restarting smoothly, and marks the relevant links of the restarting router as "potentially unavailable", but does not perform other related operations; if a failure occurs somewhere in the network during this period, Each router immediately marks all relevant links of the restarting router as unavailable, updates the link state information base, and recomputes routes without waiting to receive relevant link information from the gracefully restarting router's neighbor routers. Corresponding processing is performed only after the unavailable LSA is routed; thus, only one routing status update and recalculation is required, which speeds up the convergence speed of the network.

The present invention has the advantages of: avoiding the adverse effect of smooth restart on the rapid convergence of routing protocols, and improving the routing convergence speed; by using the Opaque LSA to announce the start and end of the smooth restart state, avoiding the modification of the stable standard routing protocol, And the existing protocol and restart method are only slightly expanded, simple and easy to implement, and will not adversely affect the processing capability of the router device; the backward compatibility is good, and the present invention will neither affect the normal operation of the original mechanism, Also do not require that all routers in the same network must support the extension method of the present invention, allowing the routers and traditional routers supporting the present invention to be mixed and networked, which is convenient for popularization and application; moreover, the method of the present invention is not only applicable to the OSPF routing protocol, but also Applies to IS-IS and BGP routing protocols. Therefore, the present invention is an extension and improvement of the existing smooth restart method, can coordinate well with the fast route convergence mechanism, and has a good application prospect.

Description of drawings

FIG. 1 is a schematic diagram of an existing graceful restart method.

Fig. 2 is a schematic diagram of link state changes between routers during a graceful restart.

Fig. 3 is a flow block diagram of a method for realizing smooth restart of routers for fast route convergence in the present invention.

FIG. 4 is a schematic diagram of an embodiment of a router smooth restart process for realizing rapid route convergence in the present invention.

Detailed ways

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings.

In the current graceful restart mechanism, only the routers adjacent to the restarting router know that the router is performing a smooth restart, while the remote router does not know this information and thinks that router is still working normally. Thus, when a link state change occurs somewhere in the network during graceful restart, neighbors of the router that is gracefully restarting must declare it unavailable, in order to avoid possible routing loops and black holes caused by out-of-sync link state information bases , for the remote router, another link state change occurs, that is, two different link state change messages (LSAs) arrive at the remote router successively. When the fast route convergence is enabled on the network, since the router will start recalculating the route immediately after receiving the LSA (only about 1 millisecond delay), when a link state change somewhere in the network during the smooth restart period, it will cause multiple failures. The update and recalculation of the secondary routing status affect the convergence speed of the network.

In order to reduce the number of calculations in the above cases, two methods are generally adopted:

(1) Set the waiting time for route recalculation: When the router receives an LSA, it does not start routing calculation immediately, but waits for a period of time before starting calculation. If the waiting time is set reasonably, the router will wait All relevant LSAs are received before the time expires, so that only one route calculation is performed. However, this method has obvious disadvantages: it affects the convergence speed of the network, which is contrary to the purpose of fast convergence.

(2) By extending the smooth restart mechanism, the remote router also knows that a smooth restart is taking place somewhere in the network, and marks all the links of the restarted router as "potentially unavailable", temporarily considering that these links are still working normally, However, when a fault occurs somewhere in the network (related LSA is received), the state of these links will be changed to "unavailable" immediately, and the next link state information database update and route recalculation operation will be performed immediately to achieve fast route convergence . The advantage of this approach is that it does not slow down fast convergence.

The present invention adopts the above-mentioned second method, which adds a "notification stage" to the existing traditional smooth restart process, and performs corresponding expansion processing on the operations of other stages.

Referring to Fig. 3 and Fig. 4, each operation step of the present invention is specifically introduced:

(1) Handshake phase: each router sends handshake information to each other, wherein the smooth restart router X tells the adjacent router Y: it will enter the smooth restart process;

(2) Announcement phase: The smooth restart router X generates an opaque link state announcement Opaque LSA that "floods" to the local area, and notifies other routers in the network that it is ready to perform a smooth restart; the remote router Z receives After receiving the opaque link state advertisement Opaque LSA, mark the relevant link of the smooth restart router X as "potentially unavailable", but do not perform the follow-up operations of updating the link state information base and recalculating the route;

(3) Restart stage: the gracefully restarting router X enters the control software restart stage, at this stage, the adjacent router Y does not receive any data packets of the shortest path first interior gateway protocol OSPF of the gracefully restarting router X, but still considers that the smooth restarting router X Restarting router X works normally, and refreshes the relevant standard link state advertisement StandardLSA information of this smoothly restarting router X to other routers in the network, so as to achieve backward compatibility;

(4) Synchronization stage: the control software of this smooth restart router X returns to normal, and requests link state information from adjacent router Y, and adjacent router Y sends all standard link state advertisement Standard LSA information in its own state information database to this Restart router X smoothly, so as to realize the link state information base synchronization between router X and adjacent router Y;

(5) End stage: After receiving the LSA sent by the adjacent router Y, the smooth restart router X finds that the state information database of the adjacent router Y has been synchronized through comparison, and then tells the adjacent router Y that the smooth restart state is normal At the same time, the smooth restart router X will also generate an Opaque LSA that "floods" to the local area, and announces to other routers in the network that it has ended the smooth restart state; after the remote router Z receives the Opaque LSA, Delete the previously established "potentially unavailable" status flag for the link associated with this graceful restart router X;

(6) Restoration to normal stage: the smooth restart router X establishes an adjacency relationship with the adjacent router Y, and the remote router Z also learns that the smooth restart router X has returned to normal, and the whole network enters the standard shortest path first interior gateway protocol OSPF interactive state.

The above steps (3) restart stage, (4) synchronization stage, or (5) end stage all further include the following operations: if a link state change occurs in the network, the remote router Z will immediately reset the smooth restart router X The state of the associated link changes from "potentially unavailable" to "unavailable", and the link state information base is updated and the route is recalculated.

The present invention has been implemented and tested, and the method adopted is mainly to add the sub-module of the present invention's smooth restart expansion processing in the protocol processing software of the router. The result of the test is successful, and the purpose of the invention is realized.

Claims (8)

1, a kind of router of realizing fast route convergence method of steadily restarting comprises the following operational phase: shake hands, restart, synchronously, finish and recover normal; It is characterized in that: before the restarting of described flow process, increase an advertisement phase: router is before steadily restarting, produce earlier an opaque LSA Opaque LSA of " inundation " in the one's respective area, other router advertisement in network: own preparation is steadily restarted; After far-end router is received this Opaque LSA, this relevant link of steadily restarting router is labeled as " potential unavailable ", but does not carry out other associative operation; Receive that up to far-end router this steadily restarts after router finished the Opaque LSA of steady rebooting status, just relevant this of deletion steadily restarted the status indication of " potential unavailable " of router link; But, if restarting in described steady restarting process, synchronously or in the ending phase, during network generation link-state change, far-end router will be this relevant link of steadily restarting router " unavailable " state from " potential unavailable " status indication immediately, and upgrade the link-state information storehouse and recomputate route, and needn't wait for and just handling accordingly after receiving the disabled LSA of relevant link that sends by this neighbor router of steadily restarting router.

2, the router according to claim 1 method of steadily restarting, it is characterized in that: described steadily restart router in the one's respective area " inundation " be opaque LSA Opaque LSA, described Opaque LSA is that ietf standardization is organized in a kind of optional LSA type that defines among the standard RFC2370, is used to carry the needed relevant information of SPF Interior Gateway Protocol OSPF optional function.

3, the router according to claim 2 method of steadily restarting, it is characterized in that: the link-state type value of described opaque LSA Opaque LSA is 10, its effective " inundation " scope is the present networks zone, promptly this Opaque LSA information is only transmitted between the router in present networks zone and diffusion, does not go and be not delivered in other the network area.

4, the router according to claim 1 method of steadily restarting is characterized in that: comprise following concrete operations step:

(1) handshake phase: each router sends handshaking information each other, wherein steadily restarts router and tells neighbor router: it will enter steadily restarts process;

(2) advertisement phase: this is steadily restarted router and produces an opaque LSA Opaque LSA of " inundation " in the one's respective area, and other router advertisement in network: own preparation is steadily restarted; Far-end router is labeled as " potential unavailable " with this relevant link of steadily restarting router after receiving this opaque LSA Opaque LSA, but does not carry out the subsequent operation of upgrading the link-state information storehouse and recomputating route;

(3) restart the stage: this is steadily restarted router and enters Control Software and restart the stage, in this stage, neighbor router does not receive that this steadily restarts any packet of the SPF Interior Gateway Protocol OSPF of router, but still it is working properly to think that this steadily restarts router, and in network other router normal refresh this steadily restart the relevant standard LSA Standard LSA information of router realize compatible so that steadily restart agreement with standard;

(4) synchronous phase: this Control Software of steadily restarting router recovers normal, to neighbor router request link state information, neighbor router all sends to this with all the standard LSA Standard LSA information in the own state information storehouse and steadily restarts router, and is synchronous to realize this link-state information storehouse of steadily restarting router and neighbor router;

(5) ending phase: after this steadily restarts the LSA that router receives that neighbor router sends, by contrast, find that its state information storehouse with neighbor router has been realized synchronously after, tell neighbor router: steady rebooting status normal termination; Simultaneously, this steadily restarts router also will produce an opaque LSA Opaque LSA of " inundation " in the one's respective area, and other router advertisement in network: oneself has finished steady rebooting status; After far-end router was received this Opaque LSA, relevant this of setting up before the deletion steadily restarted the status indication of " potential unavailable " of router link;

(6) recover normal phase: this steadily restarts router and neighbor router is set up syntople, and far-end router learns that also this steadily restarts router and recovered normally, and the whole network enters the SPF Interior Gateway Protocol OSPF interaction mode of standard.

5, the router according to claim 3 method of steadily restarting, it is characterized in that: described step (3) is restarted stage, (4) synchronous phase or (5) ending phase and is all further comprised following operation: if the Link State change has taken place in the network, far-end router will change this relevant link state of steadily restarting router into " unavailable " from " potential unavailable " immediately, and upgrades the link-state information storehouse and recomputate route.

6, the router according to claim 1 method of steadily restarting is characterized in that: described method allows to use the conventional router that can only discern and handle standard OSPF Routing Protocol and the router mixed networking of supporting described OpaqueLSA.

7, the router according to claim 6 method of steadily restarting is characterized in that: described method suggestion all-router can both be supported described Opaque LSA.

8, the router according to claim 1 method of steadily restarting, it is characterized in that: described method is applicable to SPF Interior Gateway Protocol OSPF, intermediate system Routing Protocol IS-IS or Border Gateway Protocol (BGP).

Images