Hi Aijun, Your draft introduces the Source Router ID which is, by itself, an useful protocol extension. However, the use-case on inter-as topology retrieval has issues which has been shared by many of us at the mike, offline and on the list.
Could you consider de-coupling the two? Also, if the proposal for learning inter-AS as described by you works for your own specific network design (and you don't think any of the points made affect that decision), then please go ahead. However, I do not see the point of trying to get that as an IETF document? Thanks, Ketan -----Original Message----- From: Peter Psenak (ppsenak) Sent: 24 July 2018 04:22 To: Aijun Wang <wangai...@tsinghua.org.cn>; 'Rob Shakir' <r...@rob.sh> Cc: 'Dongjie (Jimmy)' <jie.d...@huawei.com>; cho...@chopps.org; Ketan Talaulikar (ketant) <ket...@cisco.com>; firstname.lastname@example.org; Acee Lindem (acee) <a...@cisco.com> Subject: Re: 答复: [Lsr] 答复: 答复: Regarding OSPF extension for inter-area topology retrieval Hi Aijun, On 24/07/18 05:37 , Aijun Wang wrote: > _Hi, Peter:_ > > For point-to-point interface, as described in OSPFv2(RFC2328 18.104.22.168. > Describing point-to-point interfaces) > <https://tools.ietf.org/html/rfc2328#page-130>, the router LSA will include > two links description for each interface, within which the “type 3 link”(stub > network) will describe the subnet mask of the point-to-point interface. > > For broadcast/NBMA interface, the DR will be elected and it will > generate the network LSA which will include also the subnet mask of > the connected interface. > > For unnumbered and virtual link, if you consider we should include > them also for all possible scenarios even if we seldom use them in > large network, we can consider expand the summary LSA to cover it, as > done by this draft. there is no way to address unnumbered p2p case your way, because there is no Summary LSA generated to other area in such case. Anyway, reconstructing a topology of a remote area based on the prefix announcements that come from it is a broken concept. I have given you several examples where your proposal does not work. thanks, Peter > > For Anycast prefixes situation that you described(although we seldom > plan our network in such way), the PCE controller will not deduce the > wrong information from the reported information------Different router > advertise the same prefix, why can’t they be connected in logically? > > On summary, the ABR can know and report the originator of the > connected interface’s prefixes, and also the connected information for > the unnumbered/virtual link from the traditional router LSA/network > LSA message, thus can transfer them to the router that run BGP-LS, > then to the PCE controller to retrieval the topology. > > _To Rob: _ > > BGP-LS is one prominent method to get the underlay network topology > and has more flexibility to control the topology export as described > in RFC > 7752 <https://tools.ietf.org/html/rfc7752#section-1>. > > Solution 1) that you proposed is possible, but we will not run two > different methods to get the topology. > > Solution 2) is also one possible deployment, but it eliminates the > advantage of the BGP-LS, in which it needs several interaction points > with the underlay network. and such deployment is not belong to > redundancy category as information got from different areaes are different. > > Solution 3)--Streaming telemetry technology should be used mainly for > the monitor of network devices, it requires the PCE controller to > contact with every router in the network, is also not the optimal > solution when compared with BGP-LS. > > We can update the current draft to include all possible scenarios that > we are not aiming at beginning for integrity consideration. The > proposed extension does not add to complexity of IGP. What we > discussed here is the complexity of IGP protocol itself. > > Best Regards. > > Aijun Wang > > Network R&D and Operation Support Department > > China Telecom Corporation Limited Beijing Research Institute,Beijing, China. > > *发件人:*Rob Shakir [mailto:r...@rob.sh] > *发送时间:*2018年7月24日7:04 > *收件人:*Peter Psenak > *抄送:*Dongjie (Jimmy); cho...@chopps.org; Ketan Talaulikar (ketant); > Aijun Wang; email@example.com; Acee Lindem (acee) > *主题:*Re: [Lsr] 答复: 答复: Regarding OSPF extension for inter-area > topology retrieval > > +1 to Peter. We should not define fragile solutions within the IETF. > > There are also a multitude of other solutions here already: > > 1) IGP adjacency with one router in each area (at a minimum - probably > two for a robust system) over a tunnel. Deployed in many networks for > years. > 2) BGP-LS to one router (ditto robustness comment) in each area. > 3) streaming telemetry of the LSDB contents via gNMI. > > All these solutions exist in shipping implementations - please let’s > not add to the system complexity of the IGP here. > > r. > > On Mon, Jul 23, 2018 at 12:30 Peter Psenak > <ppsenak=40cisco....@dmarc.ietf.org > <mailto:40cisco....@dmarc.ietf.org>> > wrote: > > Hi Aijun, > > On 23/07/18 13:07 , Aijun Wang wrote: > > Hi, Peter: > > > > For routers that connected by LAN, the router will establish adjacent > > neighbor with DR, but not only DR advertises the connected prefixes. > > only the Network LSA includes the network mask, other routers would > include the interface address only. > > > > DR and > > DRother will all advertise type 1 and type 2 LSA with each other, then > the > > process and proposal described in this draft will still work. > > We seldom use the ip unnumbered features in our network, can we ignore > it > > then? Or other persons has some idea on such situation? > > the fact that you do not use unnumbered is not really relevant. IETF > defines solutions that MUST work for all possible scenarios, not only > for a specific one. > > > Anycast prefixes are often /32 long, this can be easily filtered by SDN > > controller because the link prefixes between two routers will be no > larger > > than /32 for IPv4 network. > > protocol allows to advertise the same prefix with an arbitrary mask from > multiple routers without these routers being directly connected. Your > proposal is based on the assumptions that are incorrect. > > thanks, > Peter > > > > > Best Regards. > > > > Aijun Wang > > Network R&D and Operation Support Department > > China Telecom Corporation Limited Beijing Research Institute,Beijing, > China. > > > > -----邮件原件----- > >发件人: Peter Psenak [mailto:ppsenak > <mailto:ppsenak>=40cisco....@dmarc.ietf.org > <mailto:40cisco....@dmarc.ietf.org>] > >发送时间: 2018年7月23日18:20 > >收件人: Aijun Wang; 'Peter Psenak'; cho...@chopps.org > <mailto:cho...@chopps.org> > >抄送: firstname.lastname@example.org <mailto:email@example.com>; 'Ketan Talaulikar > (ketant)'; 'Acee Lindem (acee)'; > > 'Dongjie (Jimmy)' > >主题: Re: [Lsr] 答复: Regarding OSPF extension for inter-area topology > > retrieval > > > > Hi Aijun, > > > > On 23/07/18 11:16 , Aijun Wang wrote: > >> Hi, Peter: > >> > >> Actually, I consider mainly the point-to-point connection and the > >> numbered interface, which are normal in our network. > >> For the two situations that you mentioned, I will investigation the > >> possible solutions and feedback you later. > >> > >> For the point-to-point and numbered interface, do you have other > >> questions then? > > > > the fact that two routers announce the same subnet, does not mean they > are > > connected together by p2p link. Anycast prefix is an example. > > > > The idea you are proposing simply does not work. > > > > thanks, > > Peter > > > > > >> > >> Best Regards. > >> > >> Aijun Wang > >> Network R&D and Operation Support Department China Telecom Corporation > >> Limited Beijing Research Institute,Beijing, China. > >> > >> -----邮件原件----- > >>发件人: Peter Psenak [mailto:ppsenak > <mailto:ppsenak>=40cisco....@dmarc.ietf.org > <mailto:40cisco....@dmarc.ietf.org>] > >>发送时间: 2018年7月23日16:15 > >>收件人: Aijun Wang; cho...@chopps.org <mailto:cho...@chopps.org> > >>抄送: firstname.lastname@example.org <mailto:email@example.com>; 'Ketan Talaulikar > (ketant)'; 'Acee Lindem (acee)'; > >> 'Dongjie (Jimmy)' > >>主题: Re: [Lsr] Regarding OSPF extension for inter-area topology > >> retrieval > >> > >> Hi Aijun, > >> > >> you are trying to reconstruct the topology of the remote area based on > >> the fact that two routers are connected to the same subnet. It does > >> not work > >> because: > >> > >> 1. connections between routers can be unnumbered 2. routers can be > >> connected via LAN, in which case only DR announces the prefix. > >> > >> In summary, what you propose does not work. > >> > >> thanks, > >> Peter > >> > >> > >> > >> On 23/07/18 09:49 , Aijun Wang wrote: > >>> (Sorry for my previous mail sent wrongly to the IDR mail list, please > >>> reply on this thread within the LSR wg) > >>> > >>> Hi, Peter: > >>> > >>> I am Aijun Wang from China Telecom, the author of draft about “OSPF > >>> extension for inter-area topology retrieval” > >>> <https://datatracker.ietf.org/doc/draft-wang-lsr-ospf-inter-area-topo > >>> l ogy-ext/>, which is presented by Mr.Jie Dong during the IETF102 > >>> meeting. > >>> > >>> Thanks for your comments on the presentation material > >>> > >> <https://datatracker.ietf.org/meeting/102/materials/slides-102-lsr-osp > >> f-inte > >> r-area-topology-retrieval-00>. > >>> > >>> Below are my explanation that regarding to the question about “how it > >>> retrievals the inter-area topology based on the extension > information”: > >>> > >>> Let’s see the graph that illustrates in Fig.1 at section 3 > >>> <https://tools.ietf.org/html/draft-wang-lsr-ospf-inter-area-topology- > >>> e xt-00#section-3> of the draft(I copy it also below for your > >>> conveniences ): > >>> > >>> Assuming we want to rebuild the connection between router S1 and > >>> router > >>> S2 that locates in area 1: > >>> > >>> 1)Normally, router S1 will advertise prefix N1 within its router LSA > >>> > >>> 2)When this router LSA reaches the ABR router R1, it will convert it > >>> into summary LSA, add the “Source Router Information”, which is > >>> router id of S1 in this example, as proposed in this draft. > >>> > >>> 3)R1 then floods this extension summary LSA to R0, which is running > >>> BGP-LS protocol with IP SDN Controller. The controller then knows the > >>> prefixes of N1 is from S1. > >>> > >>> 4)Router S2 will do the similar process, and the controller will also > >>> knows the prefixes N1 is also from S2 > >>> > >>> 5)Then it can reconstruct the connection between S1 and S2, which > >>> prefix is N1. The topology within Area 1 can then be recovered > >> accordingly. > >>> > >>> Does the above explanation can answer your question. if so, I can add > >>> it into the context of this draft in updated version. > >>> > >>> Best Regards. > >>> > >>> Aijun Wang > >>> > >>> Network R&D and Operation Support Department > >>> > >>> China Telecom Corporation Limited Beijing Research Institute,Beijing, > >> China. > >>> > >> > >> _______________________________________________ > >> Lsr mailing list > >>Lsr@ietf.org <mailto:Lsr@ietf.org> > >>https://www.ietf.org/mailman/listinfo/lsr > >> > >> _______________________________________________ > >> Lsr mailing list > >>Lsr@ietf.org <mailto:Lsr@ietf.org> > >>https://www.ietf.org/mailman/listinfo/lsr > >> > > > > _______________________________________________ > > Lsr mailing list > >Lsr@ietf.org <mailto:Lsr@ietf.org> > >https://www.ietf.org/mailman/listinfo/lsr > > > > . > > > > _______________________________________________ > Lsr mailing list > Lsr@ietf.org <mailto:Lsr@ietf.org> > https://www.ietf.org/mailman/listinfo/lsr > _______________________________________________ Lsr mailing list Lsr@ietf.org https://www.ietf.org/mailman/listinfo/lsr