Overview:
At present, domestic telecom operators use IS-IS protocol extensively in their IP backbone networks. In such a large-scale network, IS-IS convergence speed is very important. If the network fails, IS-IS protocol If the convergence speed is too slow, it will directly affect the response speed of the service. Therefore, IS-IS adopts the following methods to improve the convergence speed:
- Incremental SPF algorithm (I-SPF).
- Partial routing calculation (PRC).
- Smart timer.
- LSP spreads rapidly.
- Convergence by priority
1. Incremental SPF algorithm (I-SPF):
It is the same as the I-SPF used by OSPF. For the traditional Dijkstra SPF algorithm, the more complex the network topology, the greater the resource overhead of the SPF algorithm. If the network topology changes, for example, a new router is added to the network, or a link fails, these events will cause all routers in the network to perform a complete SPF calculation process, that is, it is necessary to calculate the SPT ( Shortest path tree). If the network fails frequently, the SPF calculation will affect the performance of all routers and impact the stability of the entire network.
In view of the above situation, the incremental SPF algorithm improves the traditional SPF algorithm. All nodes are only performing a complete SPF calculation for the first time, calculating and maintaining the SPT with themselves as the root. Later, if the network changes, only the incremental SPF algorithm is needed according to the situation, thus avoiding the repeated SPF calculation of nodes and consuming too many network resources, including the following situations.
Figure 1: SPF calculation
Figure 2: I-SPF calculation
-
In the first case
, the right side of Figure 1 shows that if there is a topology change during the operation of the SPT network formed by R1 after the first SPF calculation, such as adding or deleting a new router to the network, the nodes in the network just A new router is added to the existing SPT as a leaf node, only a very simple SPF calculation (similar to the calculation of the distance vector routing protocol) is required, and it is not necessary for all nodes to perform a complete SPF calculation again. As
shown in Figure 2 , when R7 is connected to the network as a new node, R1 does not need to recalculate the SPT of other nodes, but just add R7 as a leaf to its own SPT. -
In the second case,
sometimes a link failure will cause the node to perform SPF calculation again, but if the failed link is not on the SPT, it will not affect the node. For example, in Figure 1, if the link between R3-R4 is The link fails. Since this link is not on the SPT of R1, R1 does not need SPF calculation. If the cost of the link changes, it may cause the router to change the optimal path to other target networks. At this time, SPF calculation is required. -
The third case The
last case is when the path on the SPT fails, the path to the downstream node needs to be recalculated. Those nodes that are not affected do not need to perform SPF calculation. For example, in the topology of Figure 1, If the path between R2 and R4 fails, then R2 only needs to recalculate the shortest path to R4 and R5, and the shortest path to other nodes does not need to be recalculated.
2. Partial routing technology (PRC):
When a stub link (Stub Link) is added to the network, or a subnet segment in the network changes, these conditions do not lead to a change in the entire topology, so the nodes in the network do not need to perform SPF calculations, but only Like the distance vector routing protocol, the routing and metric information of the new link pair application can be added to the routing table, thereby saving the cost of the equipment, which is part of the routing calculation.
Part of the routing calculation feature was originally defined in IS-IS, and OSPFv2 later introduced this feature, but it is still somewhat different from the IS-IS partial routing calculation feature, which will be introduced later in this section. In the previous article, we learned that the LSP of IS-IS is based on the TLV structure, and different information is transmitted through different TLVs. For example, for the neighbor and network prefix information of a router, IS Reachability TLV and IP Reachability TLV are used respectively. To carry and deliver. When a router receives an LSP with only IP Areachability information, it can install the network prefix information in it as a route into the routing table without recalculating the SPF.
At first, because of IS-IS's partial route calculation features, the scalability of IS-IS protocol in a single area was better than OSPFv2, because OSPF puts both network reachability information and topology information in Type-1 LSA The OSPFv2 LSA is not based on the TLV structure, so these two types of information cannot be placed and transmitted separately. In this case, no matter the node or link changes in the network, or a new stub link is added, OSPFv2 nodes will perform a complete SPF calculation. In OSPFv2, only LSA3/4/5/7 will perform partial route calculation.
Note:
IS-IS and OSPF only calculate the surrounding topology of the affected node after using Full SPF in the initial route calculation. This is iSPF (incremental SPF); as for the change of the leaf route on the node on the iSPF tree, Only the PRC calculation is required, that is, only the leaf (route) is calculated, which greatly reduces the calculation load of the CPU. However, OSPF will trigger iSPF calculation when any routing changes in the area (new LSA1 and LSA2 need to be diffused). If the inter-area or external routing (LSA3/4/5/7) changes, only PRC is required, while IS-IS Only PRC is performed when any routing changes.
3. Smart timer:
In an unstable network, when the network changes frequently, the IS-IS protocol will frequently generate new LSPs and perform SPF calculations, which will consume a large amount of CPU and memory resources of the device, thereby affecting the service forwarding of the device. Smart timer is to reduce the impact on the performance and stability of the network when the network changes frequently. It includes the following two types.
1. LSP intelligent timer
LSP generation has a minimum interval limitation. Generally, the default minimum interval is 5s. After this limitation, if the network changes, because of this time limitation, it cannot quickly spread new If the network changes frequently, the continuous generation of new LSPs will take up a lot of system resources. The LSP intelligent timer ensures that the LSP generation interval tends to be reasonable according to the stability of the network, while meeting the requirements of fast convergence and low CPU usage. After using the LSP intelligent timer, the routing device can quickly respond to emergencies (such as the interface Up/Down) and speed up the network convergence. At the same time, when the network changes frequently, the interval of the smart timer will be automatically extended to avoid excessive use of CPU resources.
2. SPF smart timer By
default, IS-IS protocol uses a fixed SPF calculation interval, the default maximum value of this interval is 5s. When the network changes frequently, you can reduce the calculation interval to speed up the convergence speed, but this will cause SPF to consume a lot of CPU resources in the system; if you increase the calculation interval, it will affect the convergence speed. SPF intelligent timer can make SPF calculation better adapt to the needs of the network.
After using the SPF smart timer, when the router just starts the SPF calculation, the interval between the two calculations is small to ensure the convergence speed of IS-IS routing. Later, as the oscillation frequency of the entire IS-IS network accelerates, the interval between two SPF calculations can be appropriately extended, thereby reducing unnecessary resource consumption.
4. The rapid spread of LSP:
Under normal circumstances, after a router receives a new LSP from an interface, it will copy it into the database and perform SPF calculations, and then flood the new LSP from other IS-IS interfaces. IS-IS protocol generally uses periodic flooding of LSPs. By default, the minimum interval for sending LSPs on the interface is 50ms, and a maximum of 10 LSPs are sent each time. When the network changes frequently, it is necessary to wait for the timer to expire. New LSPs can be flooded out in the future, which slows down the network convergence speed. The rapid diffusion of LSP improves this method. When a new LSP is received and SPF is calculated, there is no need to wait until the cycle time expires. These LSPs are directly flooded out according to the specified number, which speeds up the link state. The synchronization process of the database improves the convergence speed of the network.
5. Convergence by priority:
IS-IS Priority Convergence refers to a technology that allows certain specific routes (for example, routes matching a specified IP prefix) to converge preferentially under a large number of routes. Therefore, users can configure routes related to key services to a relatively high priority, so that these routes converge faster, thereby reducing the impact on key services. By configuring different convergence priorities for different routes, important routes converge first and improve network reliability.
Sorting out the source: Huawei hedex document, "HCIE Routing and Exchange Learning Guide"