Scaling robots and other stuff with Erlang
summary
Erlang/SD-Erlang is shown to handle robotics coordination and failures far better than a ROS-only setup: a prototype uses face tracking with ROS + Erlang where Erlang processes aggregate camera data and command motors, recovering quickly after brutal process kills while ROS often doesn’t notice. Scalable Distributed Erlang adds s_groups and scalable global namespace to avoid full-mesh replication and lets the system run across hundreds of nodes and handle >5 million registrations in benchmarks where classic distributed Erlang falls over.
AI-written summary · about
chapters (21)
- 0:00Intro
- 1:07OVERVIEW
- 1:41DISTRIBUTED SYSTEMS
- 2:04EVERYTHING IS A DISTRIBUTED SYSTEM
- 3:43NETWORK IS A
- 4:30WHY GO SOFTWARE DISTRIBUTED
- 5:29WHY GO HARDWARE DISTRIBUTED
- 6:07WHY GO DISTRIBUTED
- 12:17SCALABILITY & RELIABILITY IN ERLANG
- 14:49TRANSITIVITY & GLOBAL NAME REGISTRATION
- 15:50WHY TRANSITIVE CONNECTIONS
- 24:07FEASIBILITY STUDY COMPARING ROS AND ERLANG
- 24:34SOFTWARE ARCHITECTURES
- 25:40CONFIGURATION
- 25:59SCALABILITY: PROCESSES
- 26:32LATENCY ANALYSIS
- 29:07QUALITY VARIATIONS WITH WORKER FAILURES (50%)
- 36:18TRANSITIVITY AND NAMESPACE
- 38:13SHARED NAMESPACE DE BENCH
- 42:36SHARED NAMESPACE & TRANSITIVITY: ACO
- 46:54CONCLUSION
