On 8 Nov 2011, at 21:37, "Leo Bicknell" <bicknell@ufp.org> wrote:
In a message written on Tue, Nov 08, 2011 at 04:22:48PM -0500, Christopher Morrow wrote:
I think actually it wouldn't have caused more validation requests, the routers have (in some form of the plan) a cache from their local cache, they use this for origin validation... there's not a requirement to refresh up the entire chain. (I think).
I kinda think everyone is wrong here, but Chris is closer to accurate. :P
When a router goes boom, the rest of the routers recalculate around it. Generally speaking all of the routers will have already had a route with the same origin, and thus have hopefully cached a lookup of the origin. However, that lookup might have been done days/weeks/months ago, in a stable network.
While I'm not familar with the nitty gritty details here, caches expire for various reasons. The mere act of the route changing paths, if it moved to a device with a stale cache, would trigger a new lookup, right?
Basically I would expect any routing change to generate a set of new lookups proportial to the cache expiration rules.
Which may very well fail because all the routing is hosed. I'm not all that familiar with the potential implementation issues, but I would think that network-local caches would be in order. Even with local caches, I would expect a high incidence of change to trigger something sensible to mitigate this kind of craziness from happening. I am sure enough people have had incorrectly scaled RADIUS farms blow up when a load of DSLAMS vanish and come back again not to repeat such storms. -- Leigh Porter ______________________________________________________________________ This email has been scanned by the MessageLabs Email Security System. For more information please visit http://www.messagelabs.com/email ______________________________________________________________________