Routing Area WG (rtgwg) IETF 85 Agenda ===================================================== CHAIR(s): Alia Atlas Alvaro Retana o Administrivia Chairs 10 minutes - Note Well - Scribe - Blue Sheets - Document Status Alvaro Retana/Cisco: We have a full agenda. Alia is not here this week. Nothing to bash on agenda. Document status overview. FRR and oFIB, revised drafts needed. ===== o Composite Link Framework in Multi Protocol Label Switching (MPLS) http://tools.ietf.org/html/draft-ietf-rtgwg-cl-framework Curtis Villamizar 10 minutes [Presentation] – Curtis Villamizar. Document status overview. Almost no discussion on the mailing list, hope that it is not an indication of lack of interest. Some small technical changes to the framework. Issues identified were moved to the separate section. List of issues - details on each. [Discussion] Dave McDyson: Express path drafts have been renamed to metric extensions, therefore you see it as abandoned. Little discussion on rtgwg list - discussion is in fact needed. Alvaro: Who has read the draft? Alvaro: Would be good if someone could volunteer to look at those issues. ===== o Operational management of Loop Free Alternates http://tools.ietf.org/html/draft-litkowski-rtgwg-lfa-manageability Bruno Decraene 15 minutes [Presentation] – Bruno Decraene [Detailed slides, presentation not captured.] Soliciting more feedback from deployments. [Discussion] Hannes Gredler/Juniper: Why is it important to control the LFA computation? Bruno: Good question. Could you send to mailing list? Uma Chunduri/Ericsson: Do you see any implementation that does this approach? And LFA activation granularity - is that per prefix? Bruno: convergence time is measured by sending probes, it is not tied to a prefix as such. Stewart Bryant/Cisco: There was a question about SPF monitoring - this is not about running a set of SPFs that need to be run to calculate the backup paths. Hannes Gredler: A comment to Stewart’s statement – would that mean that your backup SPF calculation is significantly slower? Stewart Bryant: No, but there are multiple runs of backup SPFs. Hannes: But the overall number is insignificant? Alvaro: Out of time, please take the discussion to the list. ===== o Applicability of LDP Multi-Topology for Unicast Fast-reroute Using Maximally Redundant Trees http://tools.ietf.org/html/draft-li-rtgwg-ldp-mt-mrt-frr Zhenbin Li 10 minutes [Presentation] - Quintin Zhao [Details not captured, in slides.] [Discussion] Alvaro: Who has read this? [a few hands.] ===== o Routing Extension for Fast-Reroute Using Maximally Redundant Trees http://tools.ietf.org/html/draft-li-rtgwg-igp-ext-mrt-frr Zhenbin Li 10 minutes [Presentation] - Quintin Zhao [Discussion] Alvaro: Any questions, comments? ===== o Tree Notification to Improve Multicast Fast Reroute http://tools.ietf.org/html/draft-wijnands-rtgwg-mcast-frr-tn IJsbrand Wijnands 10 minutes [Presentation ] – Ice Wijnands [Presentation details in slides.] [Discussion] In order for this to be fast, TN packets need to be originated and consumend in data plane. May reuse BFD infrastructure. Difference - TN are not periodic, only during the failure. Stewart: Good job in using more reserved bits. :-) [Name not stated]: When a node detects a failure, and it is itself on a backup path, does it send a TN down? Ice: Not when on the backup path in case of its failure. You may get more duplicated packets during the convergence. Alvaro: Let’s take it offline. ===== o Available Routing Constructs http://tools.ietf.org/html/draft-thubert-rtgwg-arc Pascal Thubert 10 minutes [Presentation] – Pascal Thubert ARC is another way of thinking, instead of being greedy and following the arrows (shortest path), go instead over this collection of arcs. [Discussion] Alvaro: Any questions? Alvaro: Of the people who have read the draft, how many have understood it? :-) [Name not stated]: How do you drive the extent of the ARCs/loops, making them grow by choosing the two nodes? What is the strategy of making the arc bigger or smaller? [Detailed explanation referred to the document] ===== o Applying Available Routing Constructs to bicasting http://tools.ietf.org/html/draft-thubert-rtgwg-arc-bicast Pascal Thubert 10 minutes [Presentation] – Pascal Thubert Bicasting - using two copies over two noncongruent paths. [Discussion] Alvaro: any questions, comments, observations? ===== o Arc and Ears (Discussion about Similarities/Differences) No draft -- comparison based on list discussion TBD 10 minutes [Discussion] Alvaro: This discussion is about the differences and similarities between ARC, EAR, MRT, etc. Pascal: If there is no breakage in the network, it results in the same topology. Stewart: Does MRT always do node failure as an attempted repair, or does it do leaf failure too? Pascal: For two failures - each ARC is a separate failure domain, and repair is also in a separate domain. Stewart: If the failure was on a node, would it still recover? Pascal: yes, as long as there is a possible path. [?]: Can ARC guarantee SRLG handling? Pascal: Data plane version can fix one breakage per ARC. Control plane version can fix more than one. Stewart: Would it be a good idea for the group of contributors to clarify all of those details in the draft? That likely would be a useful contribution. ===== o BGP Prefix Independent Convergence http://tools.ietf.org/html/draft-rtgwg-bgp-pic Ahmed Bashandy 10 minutes [Presentation] – Ahmed Bashandy Context and terminology. [Details in slides] [Discussion] John Scudder/Juniper: Given that is all internal to the router, why would IETF care? Ahmed: This is for information only. [Name not stated]: Maybe all vendors should document what they have done :-) ===== o 2547 egress PE Fast Failure Protection http://tools.ietf.org/html/draft-minto-2547-egress-node-fast-protection Minto Jeganathan/Hannes Gredler 10 minutes [Presentation] - Hannes Gredler. [Details in slides] [Discussion] Lucy Young/Huawei: In EVPN case supporting active-active, would this solution also apply? Hannes: There may be some pathological topologies where something along the lines of remote LFA is required, local repair would not be enough. It is a local repair technology for transient failures. ===== Alvaro: Thank you for coming. Read the drafts, and comment on the list. [End of meeting]