FlowPrioritize is a Rust library for priority-aware scheduling of mixed asynchronous workloads on top of the Tokio runtime.

It targets systems where low-latency tasks ("mouse flows") and heavy background tasks ("elephant flows") compete for the same executor resources. The scheduler uses real-time telemetry and queueing-inspired logic to reduce head-of-line blocking and protect tail latency.

• Explicit workload classification with FlowType::Mouse and FlowType::Elephant
• Dynamic scheduling based on real-time load (rho)
• EWMA telemetry for lambda and mu estimation
• Overload admission control (rho >= 1.0) for elephant flows
• Tokio-native API and lightweight integration

[dependencies]
flow-prioritize = { git = "https://github.com/LoganZ2/flow-prioritize" }
tokio = { version = "1.49", features = ["full"] }

use flow_prioritize::{FlowPrioritizeScheduler, FlowType};
use std::sync::Arc;
use std::time::Duration;

#[tokio::main]
async fn main() {
    let scheduler = Arc::new(FlowPrioritizeScheduler::new(
        4,
        0.4,
        Duration::from_millis(100),
    ));

    scheduler.submit(FlowType::Mouse, async {
        // latency-sensitive task
    }).unwrap();

    scheduler.submit(FlowType::Elephant, async {
        // heavy background task
    }).unwrap();

    let sched = Arc::try_unwrap(scheduler).unwrap();
    sched.shutdown().await;
}

See examples/simple_game_server.rs for a full runnable example.

cargo run --release --bin benchmark

This writes CSV outputs under experiment_data/.

python3 -m venv venv
source venv/bin/activate
pip install pandas matplotlib seaborn numpy
python scripts/plot_results.py

Figures are written to scripts/.

cargo build
cargo test

CI config is available at .github/workflows/ci.yml.

• src/: library implementation
• src/bin/benchmark.rs: benchmark runner used in the paper
• examples/: integration example(s)
• tests/: integration tests
• scripts/: plotting and analysis scripts

This project is licensed under the MIT License. See LICENCE.txt.

For questions about the software or reproducibility package, contact:

• [email protected]

• Open-source code on public GitHub: yes
• README with build/run/reproducibility instructions: yes
• License file in repository root (LICENCE.txt): yes