BGP convergence
BGP convergence refers to the process by which the Border Gateway Protocol (BGP) reaches a stable state where all BGP routers have the same understanding of the best paths to various networks. This state is achieved after all changes in network topology and routing policies have been communicated and processed by all routers participating in BGP.
Convergence time is a critical performance metric for BGP because it determines how quickly the network can adapt to changes, such as link failures, and continue to route traffic efficiently. The speed of BGP convergence can be affected by various factors, including the size and complexity of the network, the number of routes, the configuration of BGP timers, and the implementation of specific BGP features designed to improve convergence time.
For instance, the BGP update-delay timer, which is triggered when BGP establishes its first peer, is by default set to 120 seconds. This timer delays the execution of the BGP best path algorithm until it expires, which can impact convergence time[1]. To achieve faster convergence, it’s important to have a well-tuned Interior Gateway Protocol (IGP) such as OSPF or ISIS, which can quickly react to changes in the network topology[1].
BGP convergence can also be influenced by the use of techniques such as BGP Prefix Independent Convergence (PIC) Core, which protects against core link or node failures by ensuring that the BGP next-hop remains the same even if the path changes, thus avoiding the need for BGP to reconverge[1].
Detecting link failures for BGP-only convergence typically relies on BGP KEEPALIVE timers, which have default values of 60 and 180 seconds. These timers can be tuned for faster detection of failures, but this requires careful consideration of the trade-offs between faster convergence and the potential for increased routing instability[2].
The BGP convergence process is also described in RFC 7747, which outlines a benchmarking methodology for measuring data-plane convergence using BGP. This document distinguishes between control-plane convergence, which involves the Routing Information Base (RIB), and Forwarding Information Base (FIB) convergence, which affects the actual data-plane traffic[3].
In summary, BGP convergence is the process by which BGP routers come to an agreement on the best paths to route traffic through the network. Achieving fast BGP convergence is essential for maintaining reliable and efficient network connectivity, especially in the event of topology changes. Various methods and features can be employed to optimize BGP convergence times, but these must be balanced against the potential for causing routing instability[1][2][3].
Citations:
[1] https://lostintransit.se/2016/02/25/ccde-bgp-convergence/
[2] https://ine.com/blog/2010-11-22-understanding-bgp-convergence
[3] https://datatracker.ietf.org/doc/html/rfc7747
[4] https://www.cs.bu.edu/faculty/matta/Teaching/cs655-papers/BGP-convergence-griffin.pdf
[5] https://en.wikipedia.org/wiki/Convergence_(routing)
[7] https://www.reddit.com/r/networking/comments/3n01bv/bgp_convergence_time_3_minutes/?rdt=57952
[8] https://blog.ipspace.net/kb/BGPHighAvailability/30-Controlling-BGP-Convergence.html