On 02/10/2023 19:24, Matthew Petach wrote:
The problem with this approach is you now have
non-deterministic routing.
Depending on the state of FIB compression, packets *may*
flow out interfaces that are not what the RIB thinks they
will be.
This can be a good recipe for routing micro-loops that
come and go as your FIB compression size ebbs and flows.
Had NOT considered the looping - that's what you get for writing
in public without thinking it all the way through *blush*.
Thanks for poking holes appropriately,
Tim.
No worries--if this were easy, we would have been doing it decades ago without thinking twice.
To William's point to Tom--we are perhaps using the term "compression" in incompatible ways at times during this conversation.
There is a difference between what the papers Williams cited are doing, which is finding more optimal ways of storing the full structure in memory with what I think the general thread here is talking about, which is 'proxy-aggregation' of a form--reducing the actual number of entries in the forwarding table, regardless of the *method* of storage.
"FIB compression" of the form talked about in the papers William cited is already being done; we don't store whole string representations of the routing table in memory, and look them up sequentially, we store them in binary tries, which are faster and take up less space (e, compressed), but they still encode and represent the *whole* set of prefixes in the forwarding table.
"FIB-count-reduction" would be a more accurate term for what we're tossing about here, and that's where dragons lie, because that's where your FIB and RIB no longer represent the same set of information. And while Jon is right, it can help struggling ISPs stave off expensive upgrades, it does so at the cost of potentially increased troubleshooting nightmares when packets stop going where the RIB expects them to go, and network engineers are left scratching their heads trying to figure out why. ^_^;
As Mark just said--sane ISPs push their vendor for a knob to disable it, so that they can return back the land of deterministic lookups for the sanity of their engineers. ;)
Thanks!
Matt