15 releases (8 breaking)

0.11.0 Nov 30, 2024
0.9.0 Nov 7, 2024
0.7.0-alpha.1 Jul 30, 2024
0.4.0 Dec 14, 2023
0.3.0 Nov 5, 2023

#497 in Game dev

Download history 91/week @ 2024-08-23 19/week @ 2024-08-30 4/week @ 2024-09-06 6/week @ 2024-09-13 25/week @ 2024-09-20 18/week @ 2024-09-27 4/week @ 2024-10-04 9/week @ 2024-10-11 5/week @ 2024-10-18 64/week @ 2024-11-01 78/week @ 2024-11-08 16/week @ 2024-11-15 282/week @ 2024-11-22 147/week @ 2024-11-29 50/week @ 2024-12-06

496 downloads per month
Used in 4 crates

MIT/Apache

205KB
3K SLoC

aeronet

crates.io docs.rs

A set of Bevy-native networking crates, focused on providing robust and rock-solid data transfer primitives.

Goals

  • Native to Bevy ECS
    • Network state is represented as entities, making them easily queryable
    • React to connections and disconnections via observers
    • Send and receive data by mutating components
  • Correct and non-panicking
    • Explicit error handling, where failure conditions are clear and documented
    • No unwraps - all panicking paths are a bug unless explicitly documented
  • Support for any network topology
    • Dedicated server/client, listen server, peer-to-peer
  • Swappable IO layer
    • Use whatever you like as the underlying byte transfer mechanism
    • You can use multiple IO layers at the same time, e.g. Steam + WebTransport

High-level networking features such as replication, rollback, and prediction are explicit non-goals for this crate. Instead, this crate aims to provide a solid foundation for implementing these features.

Crates

IO layer implementations

cargo run --example channel
cargo run --example websocket_echo_server -F server
cargo run --example websocket_client -F client

# WASM
cargo install wasm-server-runner
cargo run --example websocket_client -F client --target wasm32-unknown-unknown
cargo run --example webtransport_echo_server -F server
cargo run --example webtransport_client -F client,dangerous-configuration

# WASM
cargo install wasm-server-runner
cargo run --example webtransport_client -F client --target wasm32-unknown-unknown
  • aeronet_steam: over Steam's networking sockets
    • Native
    • 🛠️ WIP
cargo run --example steam_echo_server -F server
cargo run --example steam_client -F client

Integrations

cargo run --bin move_box_server
cargo run --bin move_box_client

Overview

Layers

aeronet is fundamentally split into multiple layers:

  • IO layer (abstraction) - aeronet_io
    • Defines what a Session is, and how it behaves
    • Handles core dis/connection logic, shared among all IO layer implementations
    • Performs setup for the layers above
  • IO layer (implementation) - aeronet_channel, aeronet_webtransport, etc.
    • Establishes and maintains a connection to a peer
    • Detects connection and disconnection, and reports it to the session layer
    • Allows sending and receiving packets unreliably
    • User-swappable - can have multiple in a single app
  • Transport layer - aeronet_transport
    • Handles fragmentation, reliability, and ordering of messages
    • Splits messages into packets, and reassembles packets into messages, which can be used by layers above
    • Allows receiving acknowledgement of sent message acknowledgements
    • Technically user-swappable, but most code above this layer relies on aeronet_transport specifically
  • Component replication, rollback, etc.
    • This is not provided as part of aeronet, but you can use a crate which integrates aeronet with one of these e.g. aeronet_replicon

Getting started

To learn about how to use this crate, it is recommended that you learn the architecture by skimming the examples and reading the documentation of important types such as Session. If you're not sure where to start, take a look at the echo_client and echo_server crates. The examples are designed to be self-contained and self-documenting, giving you an easy jumping-off point for learning.

Crates and items are thoroughly documented through rustdoc, and are the most likely to be up to date, so you should use that as the definitive reference for information on specific items.

Once you have a rough idea of the architecture, choose an IO layer implementation from the list at the top, add it and aeronet to your app, and start building!

Testing

For aeronet

aeronet and its subcrates use a combination of:

  • unit tests, using cargo, for individual, self-contained features
  • integration tests, using cargo, for testing code in the context of a full Bevy app
  • fuzz tests, using cargo-fuzz, for protocol-level features and parsing

Fuzz tests

To run the fuzz tests:

cd crates/aeronet_transport
cargo +nightly fuzz run <fuzz target>

For users

Visualizer

As a debug tool, you may want to see the state of your session over time while you are in the app. If using aeronet_transport, you can use the visualizer feature to enable an egui_plot visualizer which displays statistics on sessions. This includes data such as round-trip time and packet loss percentage.

Conditioning

Using a conditioner allows you to emulate poor network conditions locally, and see how your app copes with problems such as duplicate or lost packets, delay, and jitter.

Some example tools you may use are:

  • Linux
    • tc
      • sudo tc qdisc add dev lo root netem delay 250ms
      • sudo tc qdisc add dev lo root netem delay 200ms 50ms distribution normal
      • sudo tc qdisc add dev lo root netem loss 50%
      • sudo tc qdisc delete dev lo root
  • MacOS
  • Windows

aeronet does not provide support for conditioning within the networking crate itself, since conditioning testing is more effective (and representative of real-world results) when the conditioning is applied at the lowest level possible.

Versions

bevy aeronet
0.15 0.10.0
0.14 0.9.0

Dependencies

~12–52MB
~876K SLoC