Tyra (TYped Rust Actor)

Tyra is a production ready, cross-platform Typed Actor System Written in Rust.

Configuration

See default.toml for a list of all configuration parameters and their defaults.

Configuration can be adjusted by providing a ./config/tyra.toml.

Clustering

Through the current implementation of the SerializedMessage it's proven that this system can be clustered. Development will focus on implementing a proper clustering system once 1.0.0 has been released.

cargo run --example serialize to view/run the poc implementation

Benchmarks

There are a few different benchmarks within examples, prefixed with benchmark_

The most interesting things to look at are specifically (ordered from fastest to slowest):

• cargo run --release --example benchmark_single_actor_process_after_send_single_thread
• cargo run --release --example benchmark_single_actor_process_after_send
• cargo run --release --example benchmark_single_actor
  • probably the most common scenario
• cargo run --release --example benchmark_single_actor_single_thread
  • roughly as fast as the benchmark_single_actor but some executions are much slower

Don't just trust random benchmarks posted somewhere and do your own!

Nonetheless, it makes sense to post some generic data that gives at least some insight into the performance:

• Release: 0.8.0
• CPU: Ryzen 1700 with 8 cores and 16 threads running at 3GHz
• tyra config:
  • general.default_mailbox_size: 0 (setting a fixed mailbox_size would further increase performance)
  • general.default_message_throughput: 15 (obviously could also be optimized for single actor usage, but would probably not be a fair measurement)
  • thread_pool.config.default.threads_max: 10 (overwritten by _single_thread tests)
• running benchmark_single_actor
  • send+receive 10mil messages in at least 3.3 seconds
    • ~3mil messages per second
• running benchmark_single_actor_process_after_send_single_thread
  • send 10mil messages in at least 700ms
    • ~14mil messages per second
  • receive 10mil messages in at least 750ms
    • ~13mil messages per second
• running benchmark_single_actor_process_after_send
  • send 10mil messages in at least 700ms
    • ~14mil messages per second
  • receive 10mil messages in at least 1750ms
    • ~5.7mil messages per second

Documentation

docs.rs or generate your own with cargo doc

Quickstart

This code can be found in examples/quickstart.rs and can be executed with cargo run --example quickstart

use tyra::prelude::*;
use std::error::Error;
use std::time::Duration;

// define an `ActorMessage` that can be sent to `Actors` that implement the corresponding `Handler<T>`
struct TestMessage {}
impl TestMessage {
    pub fn new() -> Self {
        Self {}
    }
}
impl ActorMessage for TestMessage {}

// define the actual `Actor` that should process messages
struct TestActor {}
impl TestActor {
    pub fn new() -> Self {
        Self {}
    }
}
impl Actor for TestActor {}

// define a factory that creates the `Actor` for us
struct TestActorFactory {}
impl TestActorFactory {
    pub fn new() -> Self {
        Self {}
    }
}
impl ActorFactory<TestActor> for TestActorFactory {
    fn new_actor(&mut self, _context: ActorContext<TestActor>) -> Result<TestActor, Box<dyn Error>> {
        Ok(TestActor::new())
    }
}

// implement our message for the `Actor`
impl Handler<TestMessage> for TestActor {
    fn handle(&mut self, _msg: TestMessage, context: &ActorContext<Self>) -> Result<ActorResult, Box<dyn Error>> {
        println!("HELLO WORLD!");
        context.system.stop(Duration::from_millis(1000));
        Ok(ActorResult::Ok)
    }
}

fn main() {
    // generate config
    let actor_config = TyraConfig::new().unwrap();
    // start system with config
    let actor_system = ActorSystem::new(actor_config);
    // create actor on the system
    let actor = actor_system.builder().spawn("test", TestActorFactory::new()).unwrap();
    // send a message to the actor
    actor.send(TestMessage::new()).unwrap();
    // wait for the system to stop
    std::process::exit(actor_system.await_shutdown());
}

License

Licensed under either of

• Apache License, Version 2.0 (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
• MIT license (LICENSE-MIT or http://opensource.org/licenses/MIT)

at your option.

Contribution

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.