Hello Thomas, Many thx for your comments and review. See inline ..
On Sun, Jul 10, 2022 at 7:13 AM <thomas.g...@swisscom.com> wrote: > Dear Robert, > > > > I reviewed draft-raszuk-lsr-imp-00 and have some firsts comments and > suggestions. > > > > First of all, speaking as a network operator who is using BMP to gain > visibility into the BGP control-plane, seeing the real benefits in > operation every day, I was looking very forward at IETF seeing a similar > debugging kind of approach being applied to IGP. You addressed that aspect > very well. Thank you very much. > > > > I would like to understand if data type 3-9 described in section 5 > exporting the initial LSDB state when the transport session is being > established and once fully exported, only the subsequential updates? > Comparable to route-monitoring in BMP. > Yes this is exactly the intention. I will make it more clear in the text. > In section 5 you are describing data type 7, 8 and 9. Exporting LSDB as > structured data in YANG. I like the idea in general but doubt that vendors > will fancy implementations due to the additional I/O impact on the IGP > process. > Well that addition was actually inspired by NOKIA claiming existing support for it already. I had some discussions with Gunter off line on this. Furthermore I think telemetry with YANG model is reality. Architecture wise conversion to YANG does not need to happen in the IGP process itself. There can be a standalone encoder process which takes feed from IGP process in any internal format as prepared and packs it into YANG envelops. Many modern vendors keep all protocol data in a database. The database can on changes just push "raw" data to such YANG export engine. > However I think it is very valuable to have the LSDB modelled in YANG for > data correlation purposes. I suggest that the YANG modelling can happen on > the producer with data type 7, 8 and 9 or on the receiver side by > converting data type 3-6 and 10-13 into JSON/XML with the YANG data model. > Indeed - that very well can be the case if YANG model for IGPs will be up to speed with all extensions. And I wanted to try to stay out of specifying anything how Receiver(s) can consume and modify the data. The only exception I made here was a statement that DATA Types 1 & 2 are 100% compatible by design with today's BGP-LS SAFI 71 & 72. > I suspect that the YANG model itself for modelling the LSDB itself is not > part of the document, therefore a reference to existing drafts such as > draft-ietf-ospf-yang would help to better understand that context. > Indeed. Will add it ! In section 5 you are describing in figure 2 the data message header. Here I > suggest to add besides the "Router Identifier" also the "Process > Identifier" to have the proper device context if more than one process is > running. > Well This is exactly why I have Section 8 :) Do you think in the light of that section we still need to carry process ID ? > Lessons learned from BMP is that knowing the control-plane state alone > isn't enough for proper data correlation for IPFIX Flow Aggregation as > described in RFC 7015. We need also to understand how (active, passive, > ecmp, not considered etc.) the path is being installed into the RIB. At BMP > we are addressing this with draft-cppy-grow-bmp-path-marking-tlv. I > suggest to include this RIB aspect into IMP from day 1. If that makes sense > to you as well, I gladly make a proposal. > Very interesting comment ! I was thinking for a while if I should add anything in respect to direct peer's state. I decided not to as part of that is reflected in LSDB. You are suggesting the RIB side. I agree on the importance of it but I think we have two challenges: A) The RIB interactions with local subsystems are almost always proprietary. There is no standard there. So may be a bit hard to spec. B) I think this is very important and to do it well perhaps a separate document would be a better idea ? RIB Monitoring Protocol (RMP) ? To your analogy with draft-cppy-grow-bmp-path-marking-tlv it is mainly about best path selection criteria. Not sure if this is applicable to link state protocols where protocol's local RIB is a product of SPF algo and insertion to main RIB is subject to admin distance comparison. > Regarding the subscription aspect described in section 3. Here I would > prefer a similar approach as draft-cptb-grow-bmp-yang, being done through a > NETCONF/RESTCONF configured subscription. A YANG model gives the > possibility not only to define but also to monitor the subscription which > is fundamental for proper data collection. > draft-arokiarajseda-ipfix-data-export-yang-model > does the same for IPFIX. For YANG push this is defined in RFC 8641. > Please observe that the PUB-SUB subscription only happens at the DATA Types. That IMO is an IMP property. I agree that what specific elements are sent can be done with the NETCONF/RESTCONF mechanism. For now I refrained from adding this to the spec. Only added top level TLV filters. Said this I am very open to follow your above recommendation for YANG DATA Types and need for more granular subscriptions. But I think I would rather see this applicable to Producer's Proxies not Producers directly. Just trying to keep Producers as light as possible as not all RP/RE can handle too much of such load. > Regarding the terminology of "Producer" and "Receiver". I suggest to align > the wording with existing Network Telemetry (RFC 9232) protocols. > Unfortunately they aren't aligned among at all even though they are doing > more or less all the same. Exporting monitoring data to a data collection. > My personal favorite is Exporting and Collecting Process. > > > > IPFIX: Exporting Process, Collecting Process > > BMP: Management Station, Monitoring Station > > YANG push: Publisher, Receiver > > IMP: Producer, Receiver > I used the Producer term as this is how some implementations call its components. The receivers are Clients. But if we think that "Publisher" term is a better fit I have no issue renaming it such. Receiver could stay as is or could be called Client too. Many thx and thank you again for your review and excellent comments. Robert > > > Best wishes > > Thomas > > > > *From:* GROW <grow-boun...@ietf.org> *On Behalf Of *Robert Raszuk > *Sent:* Saturday, July 9, 2022 9:49 PM > *To:* Jeff Tantsura <jefftant.i...@gmail.com> > *Cc:* lsr <lsr@ietf.org>; g...@ietf.org g...@ietf.org <g...@ietf.org>; > idr@ietf. org <i...@ietf.org> > *Subject:* Re: [GROW] [Idr] [Lsr] IGP Monitoring Protocol > > > > Thx Jeff ! > > > > Also I welcome more reviews and suggestions for additions or deletions of > parts of it. For now I tried to keep it very simple for routers - > essentially setup new p2p TCP or QUIC sessions and send over exactly what > you put in BGP today. In the same time I see use cases beyond that so added > few optional more DATA Types. > > > > With basic DATA Types 1 or 2 there is zero changes needed on the receivers > - some folks told me this is huge advantage. > > > > Then two optional messages REQUEST and FILTER provide ability for trimming > excessive data either on the Producer or Producer's Proxy. > > > > Many thx, > > Robert > > > > > > On Sat, Jul 9, 2022 at 9:39 PM Jeff Tantsura <jefftant.i...@gmail.com> > wrote: > > Speaking as RTGWG chair: > > > > Robert - I don’t think we’d have enough time to accommodate a good > discussion during IETF114 (we got only 1 slot), however would be happy to > provide a platform for an interim. > > The topic is important and personally (being a very large BGP-LS user) I’d > like to see it progressing. > > Cheers, > > Jeff > > > > On Jul 8, 2022, at 14:44, Robert Raszuk <rob...@raszuk.net> wrote: > > > > Hi Acee, > > > > Yes, by all means input from the operator's community is needed. It can be > collected through LSR WG, IDR WG or GROW WG. RTGWG could also contribute. > We have already seen input from some operators and their opinion on adding > and distributing more and more link state protocol and topology data in > BGP. More such input is very welcome. > > > > And to your point about RFC9086 - I see nothing wrong in keeping BGP > information in BGP. So IGP Monitoring Protocol does not target to shut down > BGP-LS. It only aims to remove 100% of non BGP sourced information from it. > > > > Controllers which today listen to BGP-LS need a number of information > sources and that spread will only keep increasing as more inputs are > becoming necessary for its computations. > > > > Regards, > > Robert. > > > > > > On Fri, Jul 8, 2022 at 11:32 PM Acee Lindem (acee) <a...@cisco.com> wrote: > > Hi Robert, > > > > *From: *Robert Raszuk <rob...@raszuk.net> > *Date: *Friday, July 8, 2022 at 4:36 PM > *To: *Acee Lindem <a...@cisco.com> > *Cc: *lsr <lsr@ietf.org>, IDR List <i...@ietf.org>, Susan Hares < > sha...@ndzh.com> > *Subject: *Re: [Lsr] IGP Monitoring Protocol > > > > Hi Acee, > > > > Thank you. I was not planning to present it in the upcoming IETF. > > > > > Let’s see how many stakeholders actually want to this protocol - then we > can talk about a WG home. > > > > An alternative approach could be to see how many stakeholders do not want > to further (for no good reason) to trash BGP. That to me would be in this > specific case a much better gauge. > > > > In that case, it seems to me that this discussion should be relegated to > IDR. Note that there is already non-IGP information transported in BGP-LS, > e.g., Egress Peer Engineering (https://datatracker.ietf.org/doc/rfc9086/ > <https://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Frfc9086%2F&data=05%7C01%7CThomas.Graf%40swisscom.com%7C5ab3e0eb45c041f362c808da61e426fc%7C364e5b87c1c7420d9beec35d19b557a1%7C0%7C0%7C637929929779223085%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=JAytgVAYRCrfCPWpfAHJMLk5fH67aEFtoyCHbo8s0Ps%3D&reserved=0>). > I implemented this on our data center routers (NXOS) years and it is solely > BGP specific. > > > > Thanks, > > Acee > > > > Kind regards, > > Robert > > > > > > On Fri, Jul 8, 2022 at 9:54 PM Acee Lindem (acee) <a...@cisco.com> wrote: > > Speaking as WG chair: > > > > *From: *Lsr <lsr-boun...@ietf.org> on behalf of Robert Raszuk < > rob...@raszuk.net> > *Date: *Friday, July 8, 2022 at 3:21 PM > *To: *lsr <lsr@ietf.org> > *Cc: *IDR List <i...@ietf.org>, Susan Hares <sha...@ndzh.com> > *Subject: *[Lsr] IGP Monitoring Protocol > > > > Dear LSR WG, > > > > Based on ongoing discussion in respect to the future of BGP-LS I > committed myself to put together an alternate proposal. > > > > The main goal is not to just publish a -00 version of the draft using > different encapsulation. The goal is to make a useful tool which can help > to export link state information from network elements as well as assist in > network observability. > > > > The IGP Monitoring Protocol (IMP) draft has been posted and should be > available at: > > > > https://datatracker.ietf.org/doc/draft-raszuk-lsr-imp/ > <https://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fdraft-raszuk-lsr-imp%2F&data=05%7C01%7CThomas.Graf%40swisscom.com%7C5ab3e0eb45c041f362c808da61e426fc%7C364e5b87c1c7420d9beec35d19b557a1%7C0%7C0%7C637929929779223085%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=rju%2FzMhLKC2gbVjN8uk2CZqolzCxAmHazPuiqHKOpJQ%3D&reserved=0> > > > > One of the key points I wanted to accomplish was full backwards > compatibility with TLVs defined for BGP-LS. In parallel other formats > (optional) are also supported. > > > > The PUB-SUB nature or FILTERING capabilities are in the spec however as > noted in the deployment section there is no expectation that this should be > supported directly on routers. Concept of Producer's Proxies has been > introduced to support this added functionality as well as provide fan-out > (analogy to BGP route reflectors). > > > > I encourage everyone interested to take a look and provide comments. At > this point this document is nothing more than my individual submission. > Offline I have had few conversations with both operators and vendors > expressing some level of interest in this work. How we proceed further (if > at all :) depends on WG feedback. > > > > Kind regards, > > Robert. > > > > PS, I do believe this work belongs in LSR WG pretty squerly. > > > > Let’s see how many stakeholders actually want to this protocol - then we > can talk about a WG home. By stakeholders, I mean operators and vendors > who are committed to implementing and deploying it - not simply those who > you are able to enlist as co-authors. Note that our IETF 114 LSR agenda is > full (with multiple agenda items not making the cut). > > > > Thanks, > > Acee > > > > > > > > _______________________________________________ > Idr mailing list > i...@ietf.org > https://www.ietf.org/mailman/listinfo/idr > <https://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.ietf.org%2Fmailman%2Flistinfo%2Fidr&data=05%7C01%7CThomas.Graf%40swisscom.com%7C5ab3e0eb45c041f362c808da61e426fc%7C364e5b87c1c7420d9beec35d19b557a1%7C0%7C0%7C637929929779223085%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=2Zr2Do3M4%2FKuXYgHuaA6Q5vAqfZAiB9DyemclasStEc%3D&reserved=0> > >
_______________________________________________ Lsr mailing list Lsr@ietf.org https://www.ietf.org/mailman/listinfo/lsr