updated routing-wg minutes for RIPE 47
Joao Damas Joao_Damas at isc.org
Wed Apr 28 16:08:47 CEST 2004
Thanks to Arife Vural of the RIPE NCC for updating the minutes with the comments on the list. Joao Damas Routing wg co-chair ISC -- Minutes of the Routing Working Group held at RIPE 47, Amsterdam, Tuesday, 27th January 2004, 16:00 - 17.30. Session 1: * Administrivia (Joao Damas) Chair: Joao Damas Scribe: Arife Vural, RIPE NCC Attendees: Minutes from R46: No comments -> approved! Actions from R46: None * Call for Participation: Inter-Domain Routing Workshop Inter-Domain Routing Workshop will be held on 1-2 May, 2004 in Amsterdam just before RIPE48. Number of attendees is limited. If you are planning to attend, it's better to sign up as soon as possible. Email list: idrwq at ripe.net * Notification on RIPE Routing Registry courses Upcoming RIPE Routing Registry courses can be seen from http://www.ripe.net/cgi-bin/courselist.pl.cgi * RISwhois - mapping IP to ASN using RIS collected BGP data (Rene Wilhelm) A new interface to data collected by RIS project. That allows IP address matching to AS number. Questions: "Are withdrawals and updates are counted?" RIB data is used. "All AS are listed?" Yes "How do you handle AS-set?" It is recorded and it's listed as different objects. * IGP Fast Convergence in a ISP Backbone (Clarence Filsfils) This case study is about the convergence of IS-IS protocol and puts forward some suggestions on how to get a better convergence time. Questions: "Was there propagation delay?" Comeback on few slides. In the study case, we do not have too many hops because of that you do not have much propagation delay. "Which one is prioritized?" /32 is first converged, and then the rest. * Operator Experience from IS-IS Convergence Tuning (Henrik Villfoer) This study is done on the TeliaSonera network. Motivation for such a study is to get a fast convergence time, around 1 sec, for time sensitive new IP services and find the limits of the protocol, IS-IS. Mostly Cisco products are used. Questions&Comments: "How many LSPs were measured after tunning?" More than 100 LSPs. Clarence Filsfils proposed to tune the process-max-time since it should give smaller maximum flooding delays (among other things). There was a comment from audience about the convergence measurement which was done by TeliaSonera: "It would be more interesting if there would be more info about flooding loops." Henrik Villfoer: We actually only measure the number of test packets lost. Systematically inspecting the TTL would tell us if we had transient loops. I could do naught but agree but complemented with the information that manual inspection of a number of cases has shown that loops do exist but are very short lived and uncommon. * IP Fast Convergence Project (Nicolas Dubois / Benoit Fondeviole) France Telecom R&D did this study to how to improve IP convergence in various protocols. Also did study how to solve the problems they had in L2TP tunnels, and MPLS/VPN. Questions: "What do you mean by L2TP is used for INTERNET aggregation?" "When you say you have convergence problems on MPLS/VPN to which kind of problems do you refer?" BGP next-hopt change and NTP. "How does one update make such a change on BGP convergence?" Update to change the FIB. * AOB There is no other business.