Designing a Rate Limiter: Centralized versus Distributed Approaches.

Hey everyone, I'm building a new service and gotta put in a rate limiter to stop abuse and make sure everyone gets fair access. I'm kinda stuck between two ideas: doing it all in one spot, maybe with Redis acting like a shared counter, or having each service instance handle its own rate limiting. The second way would be more spread out, with consistency happening over time. What's good and bad about both, thinking about speed, how consistent things are, and how hard it is to actually run them, especially when a lot of people are using it? Are there times when one way is way better than the other? And seriously, any stories from the trenches or just good advice would be awesome.