110 likes | 267 Views
RTGWG. IETF-62 Minneapolis, MN. Agenda. Agenda bashing, aministrivia (chairs) [5m] 00:05 Document status (chairs) [5m] 00:10 Base IP FRR spec update (Alia) [20m] 00:30 Micro-loop prevention DT update (Alex) [30m] 01:00
E N D
RTGWG IETF-62 Minneapolis, MN
Agenda • Agenda bashing, aministrivia (chairs) [5m] 00:05 • Document status (chairs) [5m] 00:10 • Base IP FRR spec update (Alia) [20m] 00:30 • Micro-loop prevention DT update (Alex) [30m] 01:00 • Advancements in Advanced FRR (Alia, Mike, et al) [45m] 01:45
Document status • GTSM rev: will LC • IPFRR framework: no change • Basic FRR spec: updated • draft-atlas-rtgwg-ipfrr-ip-mib-01 • Will move to WG status • Objections now or on the list
Micro-loop prevent DT report • Methods: • Incremental cost changes • Sync’ed FIB installation • Path locking via safe neighbors (PLSN) • Ordered FIB installation (OFIB) • Tunnel-based • Combination of PLSN with OFIB or tunnels
Micro-loop DT • Eliminated: • Incremental cost change • Multiple convergence cycles per topo event • Excessively long convergence: O(h) • Sync’ed FIB installation • Tight synchronization (within 10ms) at the LC level required • Puts strong constrains on implementations (clock resolution, timer accuracy) • Service dependency on NTP • Operational concerns (“is my network sync’ed?”, “what happened?”)
DT: PLSN • Doc: draft-zinin-microloop-analysis-00.txt • Pros: • Easy to understand (route metric-based) • Const convergence time • Works with any topology changes • Allows SRLGs • Eliminates ~90% of 2-hop loops • Coverage similar to FRR • Can be augmented with OFIB or tunnels to get 100% • Cons • Coverage less than 100% • Looped traffic may congest loop-free • Asymmetric link costs require stricter safety condition (downstream BF)
DT: Ordered FIB install • Pros: • 100% coverage, including asymmetric link costs • SRLGs supported, may require per-group timers • Completely control-plane based • Cons: • Worst case convergence time can be long • May become greater than current convergence times • Increases risk of second topo event • Can be improved with signaling
DT: Tunnels • Pros: • Covers 100% • Const convergence time • SRLGs? • Cons: • Requires “covert” announcements in RPs • With basic FRR: requires an additional data-plane mechanism • Tunnels impose different ops and sec considerations • Distributed version has relatively high complexity
DT: next • Next step is come up with recommendation for the basic FRR • PLSN looks most promising and allows further extensions if deemed necessary • DT should get back to the WG by Paris (optimistically within 1 month)