IS-IS Scaling Issues-Network Stability and Convergence

IS-IS Scaling Issues ”Network Stability and Convergence

As with any protocol, the capability to scale is important because it allows a network to grow in a controlled and stable manner. Many factors determine whether a network will successfully scale. If a network is inherently unstable, it will be difficult to scale. The instabilities might be because of flapping links, which frequently are not under the control of the network operator. However, the network operator can impose certain changes to contain the instabilities.

Scalability

The main factors that contribute to network scalability are as follows :

• Number of routers

• Number of links

• Number of internal and external routes

• Stability of links

• Flooding

• Memory

• Processing capacity (CPU)

The flooding of LSPs is probably the most influential of these factors when looking at scaling. The less information that routers in the network have to handle, the more the network can scale. In other words, you can scale by controlling the amount of flooding throughout the network. This also profoundly affects the amount of processing that the routers have to perform. The route processors in the routers don't need to work as hard when there are only few LSPs and not much routing information in each LSP to work with. Therefore, the amount of background flooding can be significantly reduced to enhance scalability. The two main flooding- related timers ”Maxage and LSP refresh interval ”can be increased to the region of their maximum values, approximately 18.7 hours, to help reduce the frequency of periodic flooding.

The LSP refresh interval and Maxage can be adjusted, as shown in the following configuration

 snippet:router isis     lsp-refresh-interval 65000     max-lsp-lifetime 65535 

IOS jitters these timers within some limits from the configured values to ensure that routers do not refresh their LSPs at the same time to prevent overloading the routers.

NOTE

The lsp-refresh-interval is always configured with a value less than the max-lsp-lifetime to ensure that existing LSPs are refreshed before they reach maxage and expire.

Other timers can be modified to reduce the amount of flooding and increase scalability. For example, because excessive flooding over low-bandwidth links might cause problems, Cisco IOS Software uses a pacing mechanism that allows flooding to consume only 50 percent of the configured bandwidth, on low-bandwidth links at T1 rates and below. The interface command isis lsp-interval also can be used in conjunction with this to control the interval between successive LSP retransmissions.

Another scenario in which flooding needs to be controlled to assist scalability is when running Integrated IS-IS over NBMA networks. IS-IS does not support the point-to-multipoint network type, and, therefore, it is recommended to use point-to-point subinterfaces for NBMA media. For highly meshed NBMA environments, mesh groups can be used to control the flooding over the point-to-point links. Mesh groups selectively block flooding on a per-subinterface basis. Flooding must be facilitated over the best- and most-robust paths only. The mesh- group feature must be used with care because misconfiguration coupled with link failures can result in partitioned flooding. See Chapter 8 for further discussions on IS-IS mesh-group applications. The operation of IS-IS mesh groups has been recently standardized in the IEFT and published in RFC 2973.

Stability

By using a hierarchical network design with multiple areas, you can control the amount of flooding in the network. Also, by implementing a structured IP addressing scheme and taking advantage of summarization, you can reduce the amount of IP information that the routers have to cope with. So, by using a combination of areas, sound IP addressing, and summarization, the network can be optimized to scale to a large number of nodes.

This combination of areas, well-planned IP addressing scheme, and summarization also contributes largely to the stability of the network. As discussed previously, by using summarization, instabilities can be hidden from the rest of the network. This contributes to the overall stability of the network and must be implemented where possible.

A stable network provides an excellent base for growth and needs to be considered with equal or greater importance than the introduction of new features. It is extremely difficult to introduce new features if the network is not stable; therefore, stability must be given priority attention.