Real World Migration from HTTP to gRPC in Ruby
summary
Company running ~100 microservices on Kubernetes migrates interservice calls from HTTP/1.1 to gRPC with Protocol Buffers, using a central apis central schema repo to generate Ruby/Go/Node clients distributed via Bundler/npm, plus a servicex utility lib (hot reload, server command) and custom gRPC interceptors for logging and metrics. Migration follows a POC then gradual traffic ramp on production. Result: average latency drops ~50% (e.g., 130ms to 65ms) and 90th-percentile latency falls to about a seventh.
AI-written summary · about
chapters (14)
- 0:00Intro
- 1:35Our products are built with 100+ microservices
- 2:46Latency caused by using HTTP/1.1
- 4:471. Preparation 2. Migration
- 5:26Without a monorepo, we have to share the same proto files across the multiple repositories
- 5:37How should we share proto files?
- 6:15apis: a central repository for schema management
- 8:04Application monitoring is required in production
- 10:22Need to create a productive development environment for gRPC
- 10:54Use a common utility library named "servicex"
- 12:39Pb: utility for using Protocol Buffers
- 13:59RAILS Use gRPC combined with parts of Ruby on Rails
- 17:35Average Latency decreased
- 19:21Summary 1. Why RPC
