SSO
The owner of the control plane is the RP, the Route Processor. The Active RP sends and receives the hello packets.
The physical router-to-router connections terminate on the linecard. The linecard needs a FIB to pass traffic.
With multiple RPs, if one RP has a catastrophic failure, the other RP can take over (SSO + NSR) without dropping traffic.
To get zero packet loss during a RP failure, without notifying the peer or dropping any packets, these three technologies are required (SSO + NSR + NSF).
Terms
RIB — Routing Information Base.
This is where the RP stores its routes.
FIB — Forwarding Information Base.
This is the information necessary to program the linecard to pass traffic.
SSO — Stateful Switchover.
The RPs sync with each other and share state, (hopefully) enough state to prevent traffic disruption.
Checkpointing
All necessary information to perform the task is already on the standby RP.
Non-Stop Routing
The Control plane relationships and RIB are both checkpointed.
- AKA NSR
Non-Stop Forwarding
The FIB is checkpointed.
- AKA NSF
Graceful Restart
SSO/NSF/NSR are all vendor features that do no share state with the neighbor. GR is an IETF capability both devices must have turned on.
EoR — End-of-RIB.
This means the neighbor has shared the its entire routing table.
Graceful Restart
Restart Timer
If I drop the BGP session, Please wait this long before you stop forwarding me my traffic. (Default is 2 minutes)
Stale Timer
Once I send an open message, that means I’m working, so please give me this long before flushing my routes. (Default is 6 minutes)
Graceful Restart Mechanics
This is a BGP Example.
References
Cisco - Introduction to HA Technologies: SSO/NSF with GR and/or NSR