Download as PDF, PPTX
Uploaded byBangladesh Network Operators Group
Fast Reroute in SR-MPLS by Md Abdullah Al Naser
As networks increasingly demand faster convergence and enhanced resilience, Segment Routing over MPLS (SR-MPLS) has emerged as a robust framework to simplify traffic engineering and improve failure recovery. This technical session will delve into Fast Reroute (FRR) mechanisms within SR-MPLS, with a focus on Topology Independent Loop-Free Alternate (TI-LFA). I will explore how TI-LFA enables sub-50ms protection against link and node failures while ensuring optimal coverage across any topology. The talk will also address key challenges like microloop formation during convergence and discuss practical strategies for microloop prevention using SR policies and ordered FIB updates. Through real-world examples and lab-tested topologies, attendees will gain a deeper understanding of how to design and deploy scalable, fast-converging SR-MPLS networks with high availability and minimal service disruption.
More Related Content
Similar to Fast Reroute in SR-MPLS by Md Abdullah Al Naser
More from Bangladesh Network Operators Group
Fast Reroute in SR-MPLS by Md Abdullah Al Naser
- 1. Md Abdullah AlNaser [email protected] Fast Re-route in SR-MPLS
- 2. 2 FRR Intro Fast Re-routein SR-MPLS
- 3. 3 3 Fast Re-route (FRR) •Pre-computed backup path – Backup next-hop installed in forwarding table – Minimize reconvergence time – Can be used to protect tunnels (links/nodes) – Widely used in carrier backbone • Must have LFA – Loop Free Alternative • Doesn’t provide optimum path all the time • Very difficult to configure and manage
- 4. 4 4 Fast Re-route (FRR) R1R2 R3 R4 FRR with Loop Free Alternative (LFA) R5 1 1 2 n IGP COST Primary Next-hop Backup Next-hop 1 R6 1 CE1 CE2 1 Case-1
- 5. 5 5 Fast Re-route (FRR) R1R2 R3 R4 FRR with Loop Free Alternative (LFA) R5 1 1 1 1 Primary Next-hop 1 R6 2 Can’t be a Backup Next-hop R1 itself is R5’s next-hop n IGP COST Micro-loop happens !!! CE1 CE2 Case-2
- 6. 6 6 Fast Re-route (FRR) R1R2 R3 R4 FRR with Remote Loop Free Alternative (RLFA) R5 1 1 1 1 Primary Next-hop 1 R6 2 Can’t be a Backup Next-hop R1 itself is R5’s next-hop R6 can be R1’s backup next-hop Requires LDP tunnelling for rLFA n IGP COST CE1 CE2 Case-2
- 7. 7 7 Fast Re-route (FRR) R1R2 R3 R4 FRR with RLFA is also not always Loop Free R5 1 1 1 1 Primary Next-hop 1 R6 5 Can’t be a Backup Next-hop R1 itself is R5’s next-hop R6 can’t be R1’s backup next-hop Because R1 itself is in R5’s transit path n IGP COST We need next-gen TI-LFA with micro-loop avoidance capability and 100% coverage CE1 CE2 Case-3
- 8. 8 FRR with RSVP-TE FastRe-route in SR-MPLS
- 9. 9 9 FRR with RSVP-TE •Link Protection Primary Path R1 R2 R3 R4 CE1 CE2 R6 Backup Tunnel R5 PLR MP
- 10. 10 10 FRR with RSVP-TE •Node Protection Primary Path R1 R2 R4 CE1 CE2 R5 R6 Backup Tunnel R3 PLR MP
- 11. 11 11 FRR with RSVP-TE •SRLG Protection Primary Path R1 R2 R4 CE1 CE2 R5 R6 Backup Tunnel R3 PLR MP
- 12. 12 12 FRR with RSVP-TE TypicalFRR may provide sub-optimal path R1 R2 R3 CE1 CE3 R4 R5 R6 CE2 5 n IGP COST 1 1 5 5 5 1 ● Based on circuit concepts ● Sub-optimal Path ● Not ECMP friendly ● Causes link congestion Primary Path Expected Backup Path Actual Backup Path Link Protection Configured
- 13. 13 SR-MPLS Recap Fast Re-routein SR-MPLS
- 14. 14 14 SR-MPLS • LDP+RSVP controlplane is complex • SR control plane with IGP – SR is an extension of IGP – MPLS is still as the forwarding plane • No LDP, No RSVP-TE are required
- 15. 15 15 SR-MPLS R1 R2 R3R4 CE1 R5 R6 R9 R8 R7 CE2 IP VPN 16009 IP VPN 16009 IP IP VPN 16009 IP VPN 16009 IP VPN 16009 IP VPN 16009 IP VPN PHP IP VPN 16009 IP One of the major ideas behind SR is to keep “Transport Label” consistent “per segment” PUSH CONTINUE NEXT CE1 CE2
- 16. 16 16 SR-MPLS R1 R2 R3R4 CE1 R5 R6 R9 R8 R7 CE2 IP VPN 16009 IP PHP Traffic is forwarded “Segment by Segment” 16006 16004 IP VPN 16009 16006 16004 IP VPN 16009 16006 IP VPN 16009 16006 IP VPN 16009 PHP IP VPN 16009 IP VPN IP VPN 16009 PHP IP “Ordered list of segment” is encapsulated by source node PUSH CONTINUE NEXT
- 17.
- 18. 18 18 SR-MPLS Key Features • FRRwith TI-LFA – Ensure fast re-routing – Avoids micro-looping • Highly scalable TE – SR Policy; and no more tunnel • Can be a part of SDN (for inter-domain network) – PCEP and BGP-LS
- 19. 19 FRR with TI-LFA FastRe-route in SR-MPLS
- 20. 20 20 FRR with TI-LFA FastReroute with TI-LFA • Topology Independent LFA – Sub 50ms convergence time – Minimizes amount of dropped traffic during convergence • Protects links, nodes and SRLGs • Maintains better SLA
- 21. 21 21 R2 FRR with TI-LFA TI-LFALink Protection R1 R3 R7 CE1 R5 R4 R6 CE2 10 n IGP COST 1 5 1 10 1 Topology is fully functional R1’s Primary path is in forwarding table 1 IP VPN 16003 “Ordered list of segment” for primary path 10 1
- 22. 22 22 FRR with TI-LFA TI-LFALink Protection R1 is aware of the backup path in advance in case of link failure to R2 R2 R1 R3 R7 CE1 R5 R4 R6 CE2 10 n IGP COST 1 5 1 10 1 1 10 PLR Protected Link 1
- 23. 23 23 FRR with TI-LFA TI-LFALink Protection R1 prunes protected link before link failure occurs and find out post convergence path R2 R1 R3 R7 CE1 R5 R4 R6 CE2 10 n IGP COST 1 5 1 10 1 10 1 PLR Protected Link 1
- 24. 24 24 FRR with TI-LFA TI-LFALink Protection R1 defines the P Node R2 R1 R3 R7 CE1 R5 R4 R6 CE2 10 n IGP COST 1 5 1 10 1 P Node is such node which can reach source (R1) without using protected link 10 1 P PLR Protected Link 1
- 25. 25 25 FRR with TI-LFA TI-LFALink Protection R1 defines the Q Node R2 R1 R3 R7 CE1 R5 R4 R6 CE2 10 n IGP COST 1 5 1 10 1 Q Node is such node which can reach destination (R3) without using protected link 10 1 Q PLR Protected Link 1 P
- 26. 26 26 FRR with TI-LFA TI-LFALink Protection P Space and Q Space R2 R1 R3 R7 CE1 R5 R4 R6 CE2 10 n IGP COST 1 5 1 10 1 10 1 PLR Protected Link 1 Q P
- 27. 27 27 FRR with TI-LFA TI-LFALink Protection R1 builds SID-list based on each segment it found along the post convergence path R2 R1 R3 R7 CE1 R5 R4 R6 CE2 10 n IGP COST 1 5 1 10 1 10 1 PLR IP VPN 16003 24045 16004 Protected Link “Ordered list of segment” for post convergence path 1 Q P Q P
- 28. 28 28 FRR with TI-LFA TI-LFALink Protection R1 programs the precomputed SID-List it in forwarding table R2 R1 R3 R7 CE1 R5 R4 R6 CE2 10 n IGP COST 1 5 1 10 1 10 1 Protected Link PLR IP VPN 16003 24045 16004 IP VPN 16003 “Ordered list of segment” for primary path “Ordered list of segment” for post convergence path 1 Q P
- 29. 29 29 FRR with TI-LFA TI-LFALink Protection When protected link goes down R1 instantly use post convergence SID-list even before IGP converges new path R2 R1 R3 R7 CE1 R5 R4 R6 CE2 10 n IGP COST 1 5 1 10 1 10 1 Protected Link PLR Q P IP VPN 16003 24045 16004 “Ordered list of segment” for post convergence path 1
- 30. 30 30 R2 FRR with TI-LFA TI-LFANode Protection R1 R3 R7 CE1 R5 R4 R6 CE2 10 n IGP COST 1 5 1 10 1 Topology is fully functional R1’s Primary path is in forwarding table 1 IP VPN 16003 “Ordered list of segment” for primary path 10 1
- 31. 31 31 FRR with TI-LFA TI-LFANode Protection R1 is aware of the backup Node in advance in case of failure of R2 R2 R1 R3 R7 CE1 R5 R4 R6 CE2 10 n IGP COST 1 5 1 10 1 1 10 PLR Protected Node 1
- 32. 32 32 FRR with TI-LFA TI-LFANode Protection R1 prunes protected node before node failure occurs and find out post convergence path R2 R1 R3 R7 CE1 R5 R4 R6 CE2 10 n IGP COST 1 5 1 10 1 10 1 PLR 1 Protected Node
- 33. 33 33 FRR with TI-LFA TI-LFANode Protection R1 defines the P Node R2 R1 R3 R7 CE1 R5 R4 R6 CE2 10 n IGP COST 1 5 1 10 1 P Node is such node which can reach source (R1) without using protected node 10 1 P PLR 1 Protected Node
- 34. 34 34 FRR with TI-LFA TI-LFANode Protection R1 defines the Q Node R2 R1 R3 R7 CE1 R5 R4 R6 CE2 10 n IGP COST 1 5 1 10 1 Q Node is such node which can reach destination (R3) without using protected node 10 1 Q PLR 1 P Protected Node
- 35. 35 35 FRR with TI-LFA TI-LFANode Protection P Space and Q Space R2 R1 R3 R7 CE1 R5 R4 R6 CE2 10 n IGP COST 1 5 1 10 1 10 1 PLR 1 Q P Protected Node
- 36. 36 36 FRR with TI-LFA TI-LFANode Protection R1 builds SID-list based on each segment it found along the post convergence path R2 R1 R3 R7 CE1 R5 R4 R6 CE2 10 n IGP COST 1 5 1 10 1 10 1 PLR IP VPN 16003 24056 24045 “Ordered list of segment” for post convergence path 1 P Protected Node Q 16004
- 37. 37 37 FRR with TI-LFA TI-LFANode Protection R1 programs the precomputed SID-List it in forwarding table R2 R1 R3 R7 CE1 R5 R4 R6 CE2 10 n IGP COST 1 5 1 10 1 10 1 PLR IP VPN 16003 “Ordered list of segment” for primary path “Ordered list of segment” for post convergence path 1 Protected Node P Q IP VPN 16003 24056 24045 16004
- 38. 38 38 FRR with TI-LFA TI-LFANode Protection When protected node goes down R1 instantly use post convergence SID-list even before IGP converges new path R2 R1 R3 R7 CE1 R5 R4 R6 CE2 10 n IGP COST 1 5 1 10 1 10 1 PLR “Ordered list of segment” for post convergence path 1 Protected Node P Q IP VPN 16003 24056 24045 16004
- 39. 39 39 FRR with TI-LFA FastReroute with TI-LFA • SR must be implemented to have TI-LFA • 100% topology coverage • Avoids micro-loop • ECMP friendly • Protects IP and LDP traffic also
- 40. 40 Microloop Prevention Fast Re-routein SR-MPLS
- 41. 41 41 Microloop • Microloop: Transientpacket forwarding loop during path convergence • Reasons of Microloop: – Topological change propagation may need few times – Forwarding table may not be updated at the same time or in a proper order – Depends on CPU, memory and link utilization also – Intermittent link flapping R1 R2 R3 CE1 CE2 R4 Microloop 1 1 1 5
- 42. 42 42 • Without TI-LFA –Causes packet loss due to microloop Microloop R1 R2 R3 CE1 CE2 R4 Microloop 1 1 1 5 0 ms 200 ms 400 ms Link R1-R2 goes down R1 updates forwarding table R4 updates forwarding table Traffic dropped by R1 Microloop between R1 and R4 Path is converged IP VPN 16003
- 43. 43 43 • With TI-LFA;but no microloop avoidance enabled – Much faster convergence but still microloop is there Microloop R1 R2 R3 CE1 CE2 R4 Microloop 1 1 1 5 0 ms 40 ms 200 ms Link R1-R2 goes down R1 activates TI-LFA R1 updates forwarding table And removes TI-LFA path Traffic dropped by R1 R1 forwards traffic via repair path 400 ms R4 updates forwarding table Microloop between R1 and R4 Path is converged IP VPN 16003 24043 16004 Backup SID-List
- 44. 44 44 • With TI-LFA;and microloop avoidance enabled – Much faster convergence and no microloop is there Microloop Avoidance R1 R2 R3 CE1 CE2 R4 Microloop 1 1 1 5 0 ms 40 ms 400 ms Link R1-R2 goes down R1 activates TI-LFA R4 updates forwarding table Traffic dropped by R1 R1 forwards traffic via repair path 5000 ms R1 updates forwarding table And removes TI-LFA path Traffic forwarded via repair path (Continued) Path is converged and R1 deactivates TI-LFA Microloop avoidance feature enables R1 to update its forwarding table later; to let other routers to update themselves first
- 45. 45 45 • Microloop mayoccur incase of both local and remote link failure – Local link failure for R1 – Remote link failure for R5 Microloop Avoidance R1 R2 R3 CE1 CE2 R4 Possible Microloop 1 1 1 5 R5 CE3 1 IP VPN 16003 SID-List IP VPN 16003 24043 16004 Explicit SID-List • R1 floods topology change • R5 senses possibility of microloop • R5 uses algorithm to steers traffic via explicit SID-List (Stage 1) • R5 recomputes path • R5 discards explicit SID-List (Stage 2)
- 46. 46 Load Balancing inFRR Fast Re-route in SR-MPLS
- 47. 47 47 TI-LFA Load Balancing •Prefix-SID is ECMP aware • PLR can load balance over multiple PQ pairs – Selects one PQ pair per destination (based on hash function) R1 R2 R3 CE1 CE2 R7 1 1 1 1 R8 R9 R4 R5 R6 8 4 1 4 8 1 Q P PLR P Q IP VPN 16003 24045 16004 Backup SID-List 1 IP VPN 16003 24078 16007 Backup SID-List 2 Case 1:
- 48. 48 48 TI-LFA Load Balancing •Prefix-SID is ECMP aware • PLR can load balance over multiple PQ pairs – Selects one PQ pair per destination (based on hash function) R1 R2 R3 CE1 CE2 1 1 1 R8 R9 R4 R5 R6 8 4 1 4 8 1 Q P PLR Q IP VPN 16003 24045 16004 Backup SID-List 1 IP VPN 16003 24048 16004 Backup SID-List 2 Case 2:
- 49. 49 49 TI-LFA Load Balancing •Prefix-SID is ECMP aware • PLR can load balance over multiple PQ pairs – Selects one PQ pair per destination (based on hash function) R1 R2 R3 CE1 CE2 1 1 1 R4 R5 R6 8 4 8 1 Q P PLR IP VPN 16003 24045 16004 Backup SID-List 1 IP VPN 16003 24075 16007 Backup SID-List 2 Case 3: R7 P 1
- 50.