Combined OPSAWG and OpsArea minutes Agenda Administrivia - scribes / minutes, etc. Warren / Tianran 5 minutes 3 active wg documents: - capwap-alt-tunnel is in IESG reviewA D.Romascanu: pls expect comments from IEEE - mud is now a wg accepted as WG document ------------ MUD in Your Eye Brian Weis Draft: https://datatracker.ietf.org/doc/draft-ietf-opsawg-mud/ 10 minutes Manufacturer Usage iescription Brian provides summary of what it meansA And described the istructure/content of a MUD file. variages changes since adoption by the WG early IANA assigments of namespaces requested there is some early code implemenation available on github - for controller - for mud file generator (mudmaker) - and soon some code for DNS-based ACLs a related draft is draft-weis-radext-mud-00 more eyes needed on the YANG data model more experience using mud and more code needed Adrian Farrel: I support this, but worries about privacy implications SO he welcomes more attention to this aspect Henning Schulzrining: why can we not do a signed file on a mud server that device can download.? he worries about billions of light bulbs Elliot should answer this one: Some files can be big, so that is why we chose a server approach, but they can be cached. 2ND QUESTION: HOW DO YOU PREVENT A DVICE ASSUMING AN IDENTITY FROM ANOTHER DEVICE ELLIOT 802.1AR CERTEFICate is used as dicussed in draft --------- Service Models Explained Adrian Farrel Draft: https://datatracker.ietf.org/doc/draft-wu-opsawg-service-model-explained/ 10 minutes 10 minutes explanes objectives of the draft there seem to be 2 sets of service models that is talked about So we want to present the differences. was first introduced in Berlin reulted in varioiuos updates and clarifications Customer service model and a Service Delivery model (operator side) questions: - DO we have it right that there are these 2 types of models? - DO we have the names right?Is this obvious or does it make sense to publish an RFC. Our experience (in L3SM) is that it is not obvious - Have we got correlation with MEF LSOs correct? Did we get it right? Should we ask MEF to help? Mehmet Ersu: maybe use another word for "service" Dan Romascanu: Important, maybe rename it to IETF views on .... Scott Mansfield: working with MEF would be fabulous. Hairs: humming for adoption: seems yes wins Balaz Lengel: there is already a yanda models classification document --------- Composite VPN Service Delivery Model Richard Chen Draft: https://datatracker.ietf.org/doc/draft-chen-opsawg-composite-vpn-dm/ 15 minutes - this doc is follow on on the requirements document from Hui Deng - explains background with the help of figures on a slide figure(s) are also in draft in the form of ascii art - motivation: Operator (Delivery) service model - OAM has needs on the composite network (another slide with a figure) so it would be meaningful to standardize - Another figure on slide to show how it coordinates wuth L3SM/L2SM - Another slide shows the Model requirements (they are also in the -dm draft -call for more operator involvement - FAQs: - is this a new VPN service. No it is not. - If model is not use among multiple operators, why standardize It is to help operators with the complex systems in this space. - ... see slide - SLide with diagram fro initial view of the model - SLide with basic framework for YANG data model Summary - we propose to work on composite service data model for VPN service in a multidomain, multi technology network - Scope ... see slide - Moticve: --- see slide Someobne from NOKIA: q on Initial view of Model Can there be different layer VPNs (i.e. l2, l3,lw etc) Yes we can work on that Benoit Claise: We start too get many of those service models. We've put our thoughts together and I will present that in the OPSAREA part of this meeting. Reza Rokui (Nokia person): we must have a framework that is flexible.A Hum on question: is this work interesting Chairs: it seems so based on the hum --------- In Situ/Band OAM Frank Brockners Drafts: https://tools.ietf.org/html/draft-brockners-inband-oam-requirements-02.txt https://tools.ietf.org/html/draft-brockners-inband-oam-data-02.txt https://tools.ietf.org/html/draft-brockners-inband-oam-transport-02.txt https://tools.ietf.org/html/draft-brockners-proof-of-transit-02.txt 20 minutes In Situ OAM - Update - many more new authors - we got good feedback from a few people - A brief recap: what is it: Gather telemetry dat by embedding OAM info in the data packet. So not passive, but not really active measurement either - Claim: no extra probe traffic (as with pping and traceroute) - Transport options - IPv6 Native HbyH extension hjeader or double encap - vxlan-gpe: Embeddewd telemetry protocol header ... - Deployment - doing an open source implementation to prove concept - Updates in -02 revision ... quite q few, see slide - also changes to the Data Records again, see slides (colored to show changes) - POT (Prove of TYransit) Compose an union (see slides) Could be processor inensive - There will a demo at (actually 3) at the Bits and Bytes on Thursday Open question as where to land this work (OPS or TSV) ?? Joel: the actual OAM methods vary on what the transport is. BW: there seem to be various telemetry ideas goi presented at variour places (WGs, RGs) should we get a summary of what is being discussed before we adaopt a particular approach Daniel King: yes we have seen some in SDNRG Benoit: how do you get the data off the devices That is still to be decided CRhHairs: humm for: do we i(OPSAWG) want to work on telemetry conclusion: yes we do Benoit: there is a pub/subscribe which is related DanRomascanu: As with service models, we should relaize that this is YANG oriented Chairs: Hum on do we want to work on these specific documents? conclusion: yes ---------- Export BGP Community Information in IPFIX Zhenqiang Li Draft: https://datatracker.ietf.org/doc/draft-li-opsawg-ipfix-bgp-community/ 10 minutes Proble: IPfix exports flow info - TE/TS(traffic Steering)/load balancing - ipfix is too small - This document introduces new IEs - See slide. - Application scenario: 2 architectures. -- first one with a mediator between exporter and collector, -- second one direct from exporter to collector -- see figures on slide - slide with figurs (layout) of new IEs - slides with encoding examples - IANA requirements: some assignments Chairs: hum for or against adoption of this document The room hummed in favor --------- Client Defined Private Networks Laid over Thin CPEs Linda Dunbar Drafts: https://datatracker.ietf.org/doc/draft-dunbar-opsawg-private-networks-over-thin-cpe/ 10 minutes - also called AD-WAN, but we prefer to name if private networks ... - slide with figure/opicture to show concept - What is SD-WAN - basically a multiplexer - must support MPLS, Internet,m LTE, etc see slide - Key dfifferences from L3VPN/L2VPN see slide - Simple Operations of SD-WAN (OVerlay VPN) slide with figure - Why do we need standardization of operation model? - to enable operators to deploy solutions from multiple vendors - to stimulate more usage of IETFs standarrdized YANG models - we sollicit input and comments. we also look for co-authors Brian Weis: thinks it will be very difficult to standardize this answer: we think to odo a generic datamodel and internal there are translations Brian thinks there are so many underlying tunneling models that it will be very difficult. End of OpsAWG -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= Start of OpsArea Minutes: SIDROps SIDR Work nearing publication Mostly through LC / in IESG, or nearly done. This is now chartered. Will meet at the next meeting. L3SM-> L2SM L3SM closed now L2SM. Adrian’s drafts, and others make it more clear what a service is. Chairs doing a good job in L3SM, conscripted into L2SM. QIN: L2SM new WG - will meet Thursday. Will kickoff then. L2VPN Service Model. Liaison from MEF - will discuss relationship with them. Have initial version of draft. DAN: What will be relationship with SUPA? A: Clearly different scope, mainly policy model can be used at different layers. Daniel (chair of Supra): We’ve been working on models, still haven’t identified killed app, possibly some of the SUPA can be applied here. Benoit: Charter says metric for success will be when this is used, but not holding up work. It will be a good exercise to see if we can reuse the SUPA constructs.. or not… Open Mic: Benoit on services — brainstorming. I’ve been ranting about the tsunami of YANG models. We’ve been wanting to focus on network layers, but we have lots of others to… We’ve been getting more requests for service models - and last time we had one. When discussing this with authors, they want to do the right thing, and keep them simple — but they grow with policies, and get, er, complex. Often domain specific. … so, what to do? When we have an operator coming, we want to make them welcome. But we end up with lots of models which are implementation specific. We don’t do models just for models sake. Concerned that we get lots of vendor specific ones - cannot accept all of them. Joel: We solved the AD overload issue by dedicating Benoit to this — but this doesn’t means that we should standardize everything… Bert: We had something similar in IRTF, and they didn’t create a document. It is nice that there is now a language (YANG) where they can express stuff. Adrian; I’m concerned about vendors showing up and telling operators how to run their network. The are operators / vendors wanting these to be RFCs? Joel: Not sympathetic to people who have proprietary interfaces that need to be documented. Bert: Doesn’t need to be driven by operators, but need to be validated by operators. If a vendor has already implemented, but doesn’t need to be standards tracks. Joel: It would be dangerous for us to toss at IS. But they can make demands on how these need to be reviewed. Nevil: Aaaaarg! I’d need a review group, like CFRG. Linda: Why do these need to be RFCs? Benoit: Glad you said this, we’ve been discussing things like this. We’ve been doing something similar building the catalog in the Hackathon, which has metadata. Joe Clark: The open source model has generally worked — maybe not for models, but for other stuff. People who use it say what works - if we do this, those that use it will say if it works. If you build it they will come… ? This will also showcase what can be built Bert: For historic perspective — people yelled that MIBs were ugly. We now have a tool that is not ugly. We should not be surprised that people want to use it!