limiting access at the edge or in your actual application. its a big difference

hey everyone, im thinking about rate limiting strategies for a new service. were expecting a lot of traffic and want to protect our backend from abuse or accidental overload. my initial thought is to implement it at the edge (like with a waf or cdn), but im wondering about the trade-offs. does anyone have experience with rate limiting at the application layer and how it compares? what are the pros and cons youve encountered in production