This use-case is a little too specific. I can see that having a ³new nexthop type² might assist with dynamic flow manipulations without causing packet drops.
We just have to be careful about adding this to the main draft, since it opens up to more and more special cases being added after WG LC. On 10/2/16, 3:29 PM, "i2rs on behalf of Joel M. Halpern" <[email protected] on behalf of [email protected]> wrote: >To add to that, I am wondering about ading this now. Unless I am >confused, and we have also made a serious mistake, the model is quite >extensible. It is highly likely that we will add additional next hop >types in the future. If we do not yet know of any use for this >particularly one, why add it now? > >Yours, >Joel > >On 10/2/16 6:20 PM, Jeff Tantsura wrote: >> Russ, >> >> What would be the semantics of such NH? >> Don't use for forwarding at all? Don't use for forwarding within ECMP >>bunde? Don't use in ECMP but use otherwise? >> >> Regards, >> Jeff >> >>> On Oct 1, 2016, at 7:59 PM, Russ White <[email protected]> wrote: >>> >>> Y'all -- >>> >>> I would like to suggest we add one more next hop type in section 2.4 of >>> draft-ietf-i2rs-rib-data-model, and sections 2.4 & 7.2 of >>> draft-ietf-i2rs-rib-info-model to include a nexthop type that tells the >>> forwarding device to not use a particular next hop without sending >>>traffic >>> to dev/null. The specific case is this -- assume I have an ECMP group >>>with >>> 32 or 64 entries. In this group, I want to pin one flow to a >>>particular set >>> of links, while making certain some other set of flows do _not_ use >>>those >>> links. The only way to accomplish this today would be to replace every >>>route >>> with the same ECMP group to provide a different set of next hops for >>>each >>> one. It would be cleaner to have a construction that says, "take this >>>next >>> hop out of all ECMP groups, so it is only used when specified by this >>> client," or some such. >>> >>> Maybe something like this in section 7.2 of the rib data model might >>>work -- >>> >>> == >>> 7.4. Remove from ECMP next hop >>> >>> If the a particular route is marked "remove from ECMP," then any routes >>> which recurse onto this next hop as part of an ECMP group will remove >>>any >>> paths using this route as a next hop from the ECMP group. This allows >>>the >>> I2RS controller to remove ECMP traffic from a particular next hop to >>>attain >>> finer grained control over the use of specific links to which >>>particular >>> elephant flows may be pinned, or which may be otherwise congested, etc. >>> == >>> >>> I don't know of any implementation that could do this right now, but >>>this >>> seems like a useful addition to the set of available next hops (?). >>> >>> :-) >>> >>> Russ >>> >>> >>> _______________________________________________ >>> i2rs mailing list >>> [email protected] >>> https://www.ietf.org/mailman/listinfo/i2rs >> >> _______________________________________________ >> i2rs mailing list >> [email protected] >> https://www.ietf.org/mailman/listinfo/i2rs >> > >_______________________________________________ >i2rs mailing list >[email protected] >https://www.ietf.org/mailman/listinfo/i2rs _______________________________________________ i2rs mailing list [email protected] https://www.ietf.org/mailman/listinfo/i2rs
