gRPC vs REST for inter-service communication: our migration experience
Amara Santos
·289 views
our team recently completed a migration of about 15 inter-service APIs from REST over HTTP/1.1 to gRPC over HTTP/2. the primary driver was P99 latency, and we did see a significant improvement, averaging about a 40% reduction in P99s across the migrated services, which was great.
however, it wasn't a silver bullet, and we definitely hit some pain points. debugging with gRPC is significantly harder; no simple curl command, and network visibility tools aren't as mature as for plain http. managing protobuf definitions across multiple repositories became a versioning nightmare, and keeping everything in sync required a lot of tooling investment. also, the mental model shift for some developers took longer than expected.
i'm curious about others' experiences. has your team seen similar mixed results with gRPC? what strategies did you employ to mitigate the debugging and proto management overhead? for new services, are you still defaulting to gRPC, or are you picking based on specific use cases where its benefits outweigh the added operational complexity?
6 comments