Fast Egress Protection

Fast recovery from network failures has been always in the focus of network architects and designers. It is especially important, as many new application (voice/si …

Fast recovery from network failures has been always in the focus of network architects and designers. It is especially important, as many new application (voice/signaling in cellular networks, financial application, applications controlling industrial processes, just to name some examples) are very greedy for minimal traffic loss during network failures and subsequent traffic recovery. While many good techniques have been developed over years to handle failures of transit nodes or links in a good fashion (e.g. LFA, rLFA, TI-LFA, RSVP 1:1 protection, RSVP facility protection), handling failures of egress PE has been always a challenging task. One option to address this challenge is the BGP PIC Edge, where ingress PE pre-installs in its FIB next-hops to multiple egress PEs providing reachability for given prefix. While this technique provides failover times independent from the number of BGP prefixes, it’s robustness heavily depends, how quickly ingress PE can detect failure of egress PE. This can be relatively fast in small networks (IGP domain is small, so global IGP convergence is quick). However, as the networks grow (IGP domains become big), eventually up to the point, where network size mandates partitioning the network to multiple IGP domains or autonomous systems, fast discovery of egress PE failure becomes a challenge. This presentation provides a case study for an alternative approach to address this problem, where instead of BGP PIC Edge technique, architecture referenced as ‘egress PE protection’ (called sometimes ‘egress mirroring’, too) has been deployed in one of the networks of an Tier 1 provider.