Andrey - There is current work being done to do exactly what you are proposing w/ zebra. Any work along this line would of course be appreciated, but we do have a generalized approach that we are attempting to make here. If you interested let me or Mark know and we'll fill you in on the details. donald On Fri, May 11, 2018 at 1:59 PM, Andrey Korolyov <andrey@xdel.ru> wrote:
Hi,
as frr implements protobuf communication for convenient route side-handling for some time, are there any cases around where OS kernel is not needed anymore for route handling, while bgp reflector/serializer still needs to be used? For example, one may imagine full-view or something like that being pushed to the every compute node to virtual switch which handles actual traffic overlay - in this case, route redistributor such as frr need not to maintain the same full view somewhere outside the reflector application.
Would be upstreaming of this case useable for anyone or I`m just projecting special needs for specially-designed software as a generic solution? If not, are there any reasons behind current lack of implementation of no-op zebra?
Thanks.
[please CC be back, I`m not a persistent frog@ subscriber]
_______________________________________________ frog mailing list frog@lists.frrouting.org https://lists.frrouting.org/listinfo/frog