This preview shows page 1. Sign up to view the full content.
Unformatted text preview: lows routing protocols to adapt to dynamic network conditions is periodic routing advertisements and the integration step that follows each such advertisement. This method applies to both distance-vector and link-state protocols. Each node
sends an advertisement every ADVERT INTERVAL seconds to its neighbors. In response,
in a distance-vector protocol, each receiving node runs the integration step; in the linkstate protocol each receiving node rebroadcasts the advertisement to its neighbors if it has
not done so already for this advertisement. Then, every ADVERT INTERVAL seconds, offset from the time of its own advertisement by ADVERT INTERVAL/2 seconds, each node
in the link-state protocol runs its integration step. That is, if a node sends its advertisements at times t1 , t2 , t3 , . . ., where the mean value of ti+1 − ti =ADVERT INTERVAL, then
the integration step runs at times (t1 + t2 )/2, (t2 + t3 )/2, . . .. Note that one could implement a distance-vector protocol by running the integration step at such offsets, but we
don’t need to because the integration in that protocol is easy to run incrementally as soon
as an advertisement arrives.
It is important to note that in practice the advertisements at the different nodes are
unsynchronized. That is, each node has its own sequence of times at which it will send its
advertisements. In a link-state protocol, this means that in general the time at which a
node rebroadcasts an advertisement it hears from a neighbor (which originated at either
the neighbor or some other node) is not the same as the time at which it originates its own
advertisement. Similarly, in a distance-vector protocol, each node sends its advertisement
asynchronously relative to every other node, and integrates advertisements coming from
neighbors asynchronously as well. 19.5 Link-State Protocol Under Failure and Churn We now argue that a link-state protocol will eventually converge (with high probability)
given an arbitrary initial state at t = 0 and a sequence of changes to the topology that all
occur within time (0, τ ), assuming that each working link has a “high enough” probability
of delivering a packet. To see why, observe that:
1. There exists some ﬁnite time t1 > τ at which each node will correctly know, with
high probability, which of its links and corresponding neighboring nodes are up and
which have failed. Because we have assumed that there are no changes after τ and
that all packets are delivered with high-enough probability, the HELLO protocol running at each node will correctly enable the neighbors to infer its liveness. The arrival
of the ﬁrst HELLO packet from a neighbor will provide evidence for liveness, and if
the delivery probability is high enough that the chances of k successive HELLO packets to be lost before the correct link state propagates to all the nodes in the network
is small, then such a time t1 exists.
2. There exists some ﬁnite time t2 > t1 at which all the nodes have received, with high
probability, at least one copy of every other node’s link-state advertiseme...
View Full Document
This document was uploaded on 02/26/2014 for the course CS 6.02 at MIT.
- Fall '13