hi Brian, On Mon, Dec 6, 2021 at 9:23 PM Brian E Carpenter <brian.e.carpenter@xxxxxxxxx> wrote: > Generally we expect that ASAs will run at a much lower frequency than > any "production" workload in the node, so CPU load should not be a big > issue, but memory footprint in a constrained node is certainly a > concern. We tend to assume that ASAs will be mainly installed in > non-constrained devices, or that if they are in a constrained device, > they'll have a subset of functionality. Officially, we punted on that > issue - RFC8993 says "At a later stage, the ANIMA Working Group may > define a scope for constrained nodes with a reduced ANI and well- > defined minimal functionality." Super, thanks! If you could add it to the text I'm sure it'd help the readers to better understand the assumptions under which these recommendations have been written. > I think the answer depends on the resource. For the one that we fully > defined (IP address prefixes, RFC8992) there certainly needs to be a > solid logging and recovery mechanism, as there is for traditional APAM > systems. Since GRASP operations are not intrinsically idempotent, that > must be done by the ASAs. > I don't think it can be a single global map, because it has to survive > network partition and reconnection. The global map could be > constructed if necessary from the log in each ASA. On the other hand, > if the resource being shared is upstream network capacity from a given > router, which is shared among many downstream routers, there is no > need for a global map. Thanks for taking the time to explain this to me, and again, I reckon distilling this kind of wisdom in the document could help the implementers - and the reviewers too :-) cheers, -- Thomas -- last-call mailing list last-call@xxxxxxxx https://www.ietf.org/mailman/listinfo/last-call
