Fault-Tolerant IP and MPLS Networks

C HAPTER

5

Intradomain IP Control Plane:

Restarting IS-IS Gracefully

This chapter describes two mechanisms that enable you to reduce the negative effects of the

Intermediate System-to-Intermediate System (IS-IS) restart. The chapter begins with an

overview of IS-IS routing focusing primarily on concepts necessary for a better

understanding of the proposed restart mechanisms. Two mechanisms are described: IETF

and Cisco IS-IS restart. Both mechanisms help to reduce the detrimental effects caused by

the IS-IS restart. In this regard, the strengths and weaknesses of each approach are also

discussed. In addition, the key concepts underlying each approach are illustrated through

operational examples and deployment scenarios. The chapter concludes with a chapter

summary.

Understanding the Detrimental Effects of the IS-IS

Restart

Similar to the Open Shortest Path First (OSPF) protocol, the IS-IS protocol is a link-state

protocol for intradomain routing. (For more details on OSPF, see Chapter 4, “Intradomain

IP Control Plane: Restarting OSPF Gracefully.”) In a link-state protocol, each router in the

routing domain maintains a synchronized topology database that describes the

interconnection of routers within an area or domain and reachable destinations. To build a

synchronized topology database, each router initially originates and subsequently re-

advertises its local piece of topology information periodically, which is propagated

throughout the area or routing domain through flooding. Based on the topology database,

each router computes the shortest paths for all known destinations. (A path is a sequence

of connected routers and links between a source and a destination.) The calculation of the

shortest paths is a processing-intensive task because it involves running the Dijkstra

algorithm (also referred to as Shortest Path First [SPF] algorithm) to construct a tree of

shortest paths.

Original IS-IS Restart Behavior

To ensure loop-free forwarding, any change in the network topology typically necessitates

a recomputation of the shortest paths. In general, network topology changes result from

hardware and software failures of links and nodes. For instance, when an IS-IS router