in line (all IMHO): On 01/03/2018 04:51, Toerless Eckert wrote: > On Wed, Feb 28, 2018 at 10:40:55AM +1300, Brian E Carpenter wrote: >> (a) Undoubtedly we will have changes to make after IETF Last Call, >> so we can put this on ice until then. >> (b) Yes, I think the text here and in RFC7575 is fine. Maybe it's >> the WG charter which is wrong :-). As you know, the point was to >> avoid any clash with HOMENET. But "traditionally managed" is a better >> phrase than "professionally managed", I think. > > Interesting aside! > > Is a network with an SDN controller traditionally managed ?
If the SDN controller is configured by a NOC or an NMS database managed by the NOC, yes. I think we are talking about configured by a human in the NOC when we say "traditional" or "professional". > If the SDN controller uses the ACP, does this change the result ? No. But if the SDN controller is configured by an ASA, not by a human, it's autonomic. > Long term NMS'ler would say controller/orchestrator are just fancy new > words for mostly the provisioning part of NMS, but: > > Good controllers/orchestrator would also be intent based > only that the rendering of the intent would not necessarily be > autonomic. But then we have never IMHO well described what criteria > are required to call a particular method of intent rendering autonomic. One reason why "intent" is still a buzzword left for future study. > Some for example would make distribution the key criteria while i > would just look at the observable resilience of a function. > > To use a militaristic explanation: When doing PIM-SM, defense customers > asked us what happens when the enemy shoots down the RP. And then shoot > down its replacement. An so on. Now replace RP with SDN controller. > > Aka: resilient = able to automatically restart in one of enough places to > make the solution survive all covered incidents. > > The "ability to shoot down anything" was btw. the original 1st requirement > by DoD against Arpanet, so i would be careful to call autonomic networks > non-traditional (managed). I would not know what would be more traditional > than meeting this expectation for an IP network. Right. That's why the GRASP draft used to contain a riff about how routing algorithms were the original autonomic mechanisms. If you remember, we had to remove that to satisfy the Routing Area people. If ANIMA succeeds, we will have simply extended that original DARPA requirement. Brian > > Cheers > Toerless > >> Brian >>> >>> Michael >>> >>> >>> >>> On 26/02/18 22:21, Brian E Carpenter wrote: >>>> While looking at Pascal's ACP review, I noticed that although ANIMA >>>> scope is limited by charter to "professionally managed" networks, >>>> we do not mention that scope in draft-ietf-anima-reference-model. >>>> It seems like something to be added to the Introduction. >>>> >>>> Comments? >>>> >>>> Regards >>>> Brian Carpenter >>>> >>>> >>>> _______________________________________________ >>>> Anima mailing list >>>> [email protected] >>>> https://www.ietf.org/mailman/listinfo/anima >>> >>> _______________________________________________ >>> Anima mailing list >>> [email protected] >>> https://www.ietf.org/mailman/listinfo/anima >>> >> >> _______________________________________________ >> Anima mailing list >> [email protected] >> https://www.ietf.org/mailman/listinfo/anima > _______________________________________________ Anima mailing list [email protected] https://www.ietf.org/mailman/listinfo/anima
