1

BGP Prefix Independent Convergence

draft-rtgwg-bgp-pic-00
Authors : Ahmed Bashandy, Cisco Systems Clarence Filsfils, Cisco Systems Pradosh Mohapatra, Cisco Systems

Presenter : Ahmed Bashandy

IETF85, Nov/2012 Atlanta, USA

BGP PIC

Ahmed Bashandy

IETF 85. Nov/2012, Atlanta, USA

Agenda

Problem Solution Recovery from various failure scenarios

BGP PIC

Ahmed Bashandy

IETF 85. Nov/2012, Atlanta, USA

Agenda

Problem Solution Recovery from various failure scenarios

BGP PIC

Ahmed Bashandy

IETF 85. Nov/2012, Atlanta, USA

What do we want to Achieve?

Serial nature of reachability propagation
BGP convergence is inherently slow

Can we adjust the data structure in time complexity that does not depend on BGP prefixes?
If a path is lost or gained, modifications should be independent of the number of prefixes If there is a recovery, recover as soon as possible independent of the number of prefixes

Objective
Make rere-convergence after topology change independent of the number of BGP prefixes
BGP PIC Ahmed Bashandy IETF 85. Nov/2012, Atlanta, USA

Terminology
Leaf: A prefix or local label container datastructure Path: a recursive or non-recursive path
May be primary or backup

Pathlist: an array of paths

Each path carries its own pathindex

OutLabel-Array: Array of outgoing labels and/or label actions*

A possibly different outlabel-array attached to each leaf Each entry represents an outgoing label and/or label action for a path in the pathlist**

Adjacency:
The L2 information to send a packet to a directly connected router

BGP PIC

Ahmed Bashandy

IETF 85. Nov/2012, Atlanta, USA

Agenda

Problem Solution Recovery from various failure scenarios

BGP PIC

Ahmed Bashandy

IETF 85. Nov/2012, Atlanta, USA

Basic Idea
BGP-PIC is really a FIB feature Hierarchical and shared forwarding chains On topology changes
FIB modifies pathlists immediately without modifying leaves Restore traffic very quickly BGP modifies leaves later at a slower pace

Behavior is internal to the router

Completely transparent to operator Incrementally deployable
BGP PIC Ahmed Bashandy IETF 85. Nov/2012, Atlanta, USA

Sample Topology: VPN in LDP Core

LDP Core
PE2 PE1 PE3 N/n

Two BGP paths* Two IGP paths

BGP PIC

Ahmed Bashandy

IETF 85. Nov/2012, Atlanta, USA

Forwarding Chain on Ingress PE

LDP-L11 VPN-L21 VPN-L31 VPN leaf 1 PE2 IGP leaf PE2 LDP-L21

PE2 PE3 Shared BGP Pathlist (even among VRFs) VPN leaf 2 VPN-L22 VPN-L32 BGP OutLabel Array
BGP PIC

NH1 NH2 Shared IGP Pathlist

Adj1

Adj2

PE3 PE3 IGP leaf

LDP-L12 LDP-L22 IGP OutLabel Array

Ahmed Bashandy IETF 85. Nov/2012, Atlanta, USA

10

Forwarding Chain on Egress PE2

BGP OutLabel Array Unlabaled Swap VPN-L(PE3) VPN leaf IP prefix Leaf for Connected route Towards CE

NH to CE PE3 address Shared BGP Pathlist VPN leaf Local Label

IGP Leaf through core Towards PE3

BGP PIC

Ahmed Bashandy

IETF 85. Nov/2012, Atlanta, USA

11

Agenda

Problem Solution Recovery from various failure scenarios

BGP PIC

Ahmed Bashandy

IETF 85. Nov/2012, Atlanta, USA

Core Link/Node Failure

A
PE1 PE3 PE2

12

N/n

Works whether BGP is running In the core or not

B
Failure
PE1

PE2

N/n
PE3

IGP/LDP Convergence

C
PE1

PE2

N/n
PE3

BGP PIC

Ahmed Bashandy

IETF 85. Nov/2012, Atlanta, USA

Core Link/Node Failure: FIB on PE1

BGP leaf

13

Core link Failure

BGP leaf

pathlist

IGP/LDP Convergence
BGP leaf
BGP pathlist NOT modified

PE2

PE2 IGP leaf

pathlist

IGP PE2 leaf NH1

pathlist

PE2
Only IGP pathlist is modified

IGP leaf

NH1

NH2
BGP PIC Ahmed Bashandy IETF 85. Nov/2012, Atlanta, USA

Unipath PE Node* Failure: Protection at Ingress PE

PE2 PE1 PE3 CE2 CE1

14

Works whether BGP is running in the core or not** N/n PE3 must be prepreprogrammed as a backup path

Core or Edge node Failure

PE2 PE1 PE3

CE1

N/n
CE2

FIB Correction

PE2 PE1 PE3

CE1

N/n
CE2

BGP PIC

Ahmed Bashandy

IETF 85. Nov/2012, Atlanta, USA

Unipath PE Node/PE-CE Link failure: FIB on PE1

PE3-LI PE2-LI

15

Failure
Backup PE

PE3 must be prepreprogrammed as a backup path

BGP Pathlist recalculation

PE3-LI PE2-LI

PE2

PE3

Outlabel-array does OutlabelNOT change

NH1

NH2

IGP leaf

PE3-LI

PE3

PE2-LI

PE3
Path keeps its original index

NH1 IGP withdraws The leaf for PE2

NH2

Use backup PE and backwalk to recalculate higher level LDIs

NH2
BGP PIC Ahmed Bashandy IETF 85. Nov/2012, Atlanta, USA

16

Unipath PE-CE Link Failure: Protection at Egress PE

PE2 PE1 PE3 CE2 CE1

More useful for BGPBGP -free core N/n

Locally detect failure 50ms recovery !!

PE2

CE1

Failure

PE1 PE3 CE2

N/n

PE2

CE1

FIB Correction

PE1 PE3 CE2

N/n

BGP PIC

Ahmed Bashandy

IETF 85. Nov/2012, Atlanta, USA

17

Unipath PE-CE Link failure: FIB on PE2

Swap unlabeled

Failure
Backup path Going to PE3

PE3 must be prepreprogrammed as a backup path

BGP Pathlist recalculation

Outlabel-array does OutlabelNOT change
unlabeled Swap

CE1

PE3

LDP LI To PE3

PE3 NH1 NH2

LDP Label To PE3

Path keeps its original index

unlabeled Swap

PE3 NH1 Locally detect failure 50ms recovery !! NH2

LDP LI To PE3

Swap incoming PE2 VPN label with PE3 VPN label then push transport label for PE3
IETF 85. Nov/2012, Atlanta, USA

NH2
BGP PIC Ahmed Bashandy

Conclusion

18

Single elegant design to handle many convergence/protection cases in a BGP prefix independent time

BGP PIC

Ahmed Bashandy

IETF 85. Nov/2012, Atlanta, USA