Z. Mao, R. Govindan, G. Varghese, and R.~H. Katz, "Route Flap Damping Exacerbates Internet Routing Convergence", in ACM SIGCOMM, Aug 2002.
|Route Flap Damping Exacerbates Internet Routing Convergence|
R. H. Katz
|Published:||ACM SIGCOMM, 2002|
|Abstract:||Route flap damping is considered to be a widely deployed mechanism in core routers that limits the widespread propagation of unstable BGP routing information. Originally designed to suppress route changes caused by link flaps, flap damping attempts to distinguish persistently unstable routes from routes that occasionally fail. It is considered to be a major contributor to the stability of the Internet routing system.|
In this paper, we analyze a previously not well-studied
interaction between BGP's route withdrawal process and its route flap
damping mechanism for ensuring the overall stability of the
Internet routing system. This interaction can, depending upon the
topology, suppress up to one hour the propagation of a route that
has been withdrawn once and re-announced. We have shown that
this interaction has a number of subtle features. For instance, we
found that in the pyramid topology increasing the size of the
topology actually improved the rate of convergence.
We have proposed a simple fix to this withdrawal triggered suppression called selective flap damping. It relies on being able to weed out secondary flaps using a monotonicity condition which selectively avoids penalizing such secondary flaps. Our selective flap damping mechanism successfully eliminates withdrawal triggered suppression in all the topologies that we have analyzed.
We leave for further work the problem of accurately characterizing the network topologies and sizes which will induce withdrawal triggered suppression. A theoretical analysis of the properties of selective flap damping would also be desirable. Despite this, our paper together with [7, 8] makes it clear that faster convergence does require modifying BGP. This could be done by either fixing the withdrawal path exploration phenomenon (the direction followed in ) or by deploying a mechanism similar in spirit to selective flap damping (as in our paper). Either way, such BGP modifications could move us closer to the Holy Grail: an inter-domain routing protocol that is stable and yet reroutes traffic extremely fast after failure.
|Entry TODO:||Paraphrase results.|