what u use for rate limiting in microservices tho

hey folks, i've been wrestling with rate limiting for a while now across different services. i've seen solutions using api gateways, sidecars, and even within the app code itself. each seems to have its trade-offs. what's your preferred method for implementing rate limiting, especially when dealing with a growing number of microservices and varying traffic patterns? any gotchas or performance considerations i should be aware of