This crate implements two rate-limiting algorithms in Rust: * a leaky bucket and * a variation on the leaky bucket, the generic cell rate algorithm (GCRA) for rate-limiting and scheduling.
Add the crate ratelimit_meter to your Cargo.toml
file; the crates.io page
can give you the exact thing to paste.
Find them on docs.rs for the latest version!
Unlike some other token bucket algorithms, the GCRA one assumes that all units of work are of the same "weight", and so allows some optimizations which result in much more concise and fast code (it does not even use multiplication or division in the "hot" path for a single-cell decision).
All rate-limiting algorithm implementations in this crate are thread-safe and lock-free. Here are some benchmarks for repeated decisions (run on my macbook pro, this will differ on your hardware, etc etc):
``` $ cargo bench Compiling ratelimitmeter v0.4.1 (file:///Users/asf/Hacks/ratelimitmeter) Finished release [optimized] target(s) in 1.71 secs Running target/release/deps/ratelimit_meter-680be7c7547f40f9
running 0 tests
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out
Running target/release/deps/multi_threaded-b206ea78b9fc87cc
running 2 tests test benchgcra20threads ... bench: 185 ns/iter (+/- 71) test benchleakybucket_20threads ... bench: 667 ns/iter (+/- 16,193)
test result: ok. 0 passed; 0 failed; 0 ignored; 2 measured; 0 filtered out
Running target/release/deps/single_threaded-18617cd4f9e09b0d
running 8 tests test benchallower ... bench: 26 ns/iter (+/- 4) test benchgcra ... bench: 131 ns/iter (+/- 33) test benchgcrabulk ... bench: 143 ns/iter (+/- 24) test benchleakybucket ... bench: 156 ns/iter (+/- 27) test benchleakybucketbulk ... bench: 152 ns/iter (+/- 24) test benchthreadsafeallower ... bench: 50 ns/iter (+/- 8) test benchthreadsafegcra ... bench: 133 ns/iter (+/- 21) test benchthreadsafeleakybucket ... bench: 154 ns/iter (+/- 47)
test result: ok. 0 passed; 0 failed; 0 ignored; 8 measured; 0 filtered out ```
I am actively hoping that this project gives people joy in using rate-limiting techniques. You can use these techniques for so many things (from throttling API requests to ensuring you don't spam people with emails about the same thing)!
So if you have any thoughts about the API design, the internals, or you want to implement other rate-limiting algotrithms, I would be thrilled to have your input. See CONTRIBUTING.md for details!