Scalable Networking with Rust
summary
Claims a real 10X+ memory and CPU win vs Go by reimplementing a POP-style event router in Rust using MIO event loops per core and one thread per CPU instead of a single loop plus a generic threadpool. Production deployment handles petabytes of image traffic by passing TCP sockets between IO loops over channels, using a 4‑byte length header and minimal unsafe for byte-order tweaks. Works on Unix; socket-transfer approach is not Windows-friendly.
AI-written summary · about
chapters (4)
- 0:00Start
- 9:53using multiple cores as a multiple event loops
- 16:21pass tcp connection between multiple threads
- 17:42measuring the length of bytes
related talks
27mHow to switch from cozy NodeJS to scary Rust as a company.
47mA Tale of Two Asyncs: Open Source Language Design in Rust and Node.js
19mDeno: an experimental approach on V8 interoperability
31mPorting Turborepo From Go To Rust
22mMulti-tiered Node.js architectures
33mReaching Ludicrous Speed
28mPowering Up NodeJS Performance by Rust and WebAssembly
21mIntroducing gRPC-Rust