BGP Convergence and ASn allocation design in Large Scale Networks covered in this post and the video at the end of the post. This content is explained in great detail in my BGP Zero to Hero course as well as CCIE Enterprise Training. BGP is always known as slowly converged protocol. In fact this is wrong knowledge.
If you just mention about BGP Control plane convergence, can be true but we always ignore BGP Data Plane Convergence which is commonly known as BGP PIC (Prefix Independent Convergence) In this post, I will explain the BGP Path Hunting process which slows down the convergence process. Path Hunting is not only BGP but in general distance vector protocols convergence problem.
Effect of Path Hunting gets very problematic in densely meshed topologies such as CLOS or Fat Tree. Many Leaf and Spine switches might be in the network and when EBGP is used (As it is recommended in RFC 7938) Path Hunting should be avoided by allocation the Autonomous System number to the networking devices wisely. Otherwise, for the prefix which is not anymore advertised to network due to failure for example, BGP speaking routers try any path which was previously learned.
I thought the best way to understand the problem space and solutions, sharing video with the Lab configuration would be best, thus I recommend you to watch below video and if you find it useful, share your comments in the comment section below. In this video, I am explaining BGP Convergence in CLOS/Fat Tree topologies, demonstrating BGP Path Hunting problem, different BGP ASn (Autonomous System Number) allocation schemas for large scale networks with the configuration in a lab.
Also I am explaining the suggested Autonomous System allocation schema from RFC 7938 and share the different alternative ASn designs as well. If you are interested in BGP, Bellman-Ford Loop avoidance and convergence characteristics, or overall increasing routing knowledge, this video is for you!