Network routing ii 6 routing around failures

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: nt. Once a node has its own correct link state, it takes a time proportional to the diameter of the network (the number of hops in the longest shortest-path in the network) for that CHAPTER 19. NETWORK ROUTING - II 6 ROUTING AROUND FAILURES advertisement to propagate to all the other nodes, assuming no packet loss. If there are losses, then notice that each node receives as many copies of the advertisement as there are neighbors, because each neighbor sends the advertisement once along each of its links. This flooding mechanism provides a built-in reliability at the cost of increased bandwidth consumption. Even if a node does not get another node’s LSA, it will eventually get some LSA from that node given enough time, because the links have a high-enough packet delivery probability. 3. At a time roughly ADVERT INTERVAL/2 after receiving every other node’s correct link-state, a node will compute the correct routing table. Thus, one can see that under good packet delivery conditions, a link-state protocol can converge to the correct routing state as soon as each node has advertised its own linkstate advertisement, and each advertisement is received at least once by every other node. Thus, starting from some initial state, because each node sends an advertisement within time ADVERT INTERVAL on average, the convergence time is expected to be at least this amount. We should also add a time of roughly ADVERT INTERVAL/2 seconds to this quantity to account for the delay before the node actually computes the routing table. This time could be higher, if the routes are recomputed less often on average, or lower, if they are recomputed more often. Ignoring when a node recomputes its routes, we can say that if each node gets at least one copy of each link-state advertisement, then the expected convergence time of the protocol is one advertisement interval plus the amount of time it takes for an LSA message to traverse the diameter of the network. Because the advertisement interval is many orders of magnitude larger than the message propagation time, the first term is dominant. Link-state protocols are not free from routing loops, however, because packet losses could cause problems. For example, if a node A discovers that one of its links has failed, it may recompute a route to a destination via some other neighboring node, B . If B does not receive a copy of A’s LSA, and if B were using the link to A as its route to the destination, then a routing loop would ensue, at least until the point when B learned about the failed link. In general, link-state protocols are a good way to achieve fast convergence. ￿ 19.6 Distance-Vector Protocol Under Failure and Churn Unlike in the link-state protocol where the flooding was distributed but the route computation was centralized at each node, the distance-vector protocol distributes the computation too. As a result, its convergence properties are far more subtle. Consider for instance a simple “chain” topology with three nodes, A, B , a...
View Full Document

Ask a homework question - tutors are online