Nonstop routing (NSR) is beneficial for BGP peers because it reduces the likelihood of dropped packets during switchover from the active Route Processor (RP) to the standby RP. Switchover occurs when the active RP fails for some reason and the standby RP takes control of active RP operations. The BGP—IPv6 NSR feature extends BGP support for NSR to include the following IPv6-based address families:
IPv6 unicast + label
Figure 1. Basic 6VPE Network Configuration
The figure above depicts a basic deployment scenario. Provider edge (PE) router 1, P, and PE2 form a 6VPE cloud. The customer edge (CE) router 1 to PE1 connection is IPv6 (VRF). The PEs are HA/SSO and NSF capable. The P routers are capable of Multiprotocol Label Switching (MPLS) label preservation (NSF equivalent).
As the CE1 is customer equipment, the provider cannot determine that it must be upgraded to be NSF aware. If PE1 can perform NSR on its connection to CE1, then CE1 will not be aware or impacted when PE1 performs a switchover in SSO mode. For all other connections within the autonomous system, the operations may be NSF or graceful restart. This means the control plane will be reset, and all the immediate peers will be aware of it and will resend data to help re-establish the session, but forwarding will be uninterrupted.
Neighbors not operating under NSR are still expected to be NSF capable/aware. If the CE is already NSF aware (that is, it can handle a BGP graceful restart by its peers), then the PE-CE connection will not be NSR, and will instead follow the regular NSF processing model. This parallels NSR for VPNv4 and assists in conserving network resources.