aquatic/README.md
Joakim Frostegård 2e67f11caf
udp: add experimental io_uring implementation (#131)
* WIP: add udp uring support

* WIP: fix udp uring address parsing

* WIP: udp uring: resubmit recv when needed

* WIP: udp uring: add OutMessageStorage, send swarm responses

* WIP: udp uring: increase ring entries to 1024

* WIP: udp uring: add constants

* WIP: udp uring: use sqpoll, avoid kernel calls

* WIP: udp uring: disable sqpoll

* WIP: udp uring: use VecDeque for local responses

* udp uring: enable setup_coop_taskrun

* udp uring: add RecvMsgStorage

* udp: improve split of uring and mio implementations

* udp uring: clean up

* udp uring: initial ipv6 support

* udp uring: improve helper structs

* udp uring: clean up, use constants for important data

* udp: share create_socket fn between implementations

* udp uring: improve send buffer free index finding

* udp uring: work on SendBuffers.try_add

* udp uring: split into modules

* udp uring: Rename RecvMsgMultiHelper to RecvHelper

* udp uring: improve SendBuffers

* udp uring: fix copyright attribution in buf_ring module

* udp uring: stop always consuming 100% cpu

* udp uring: clean up

* udp uring: add handle_recv_cqe

* udp uring: move local_responses into SocketWorker

* udp uring: move timeout_timespec into SocketWorker

* Update TODO

* udp: make io-uring optional

* Update TODO

* udp uring: enqueue timeout before sends

* udp uring: move likely empty buffer tracking logic into SendBuffers

* udp uring: improve error handling and logging

* udp uring: keep one timeout submitted at a time

* udp uring: update pending_scrape_valid_until

* udp uring: add second timeout for cleaning

* Update TODO

* udp uring: store resubmittable squeue entries in a Vec

* udp uring: add comment, remove a log statement

* Update TODO

* Update TODO

* udp: io_uring: fall back to mio if io_uring support not recent enough

* udp: uring: add bytes_received statistics

* udp: uring: add bytes_sent statistics

* udp: uring: add more statistics

* Update TODO

* udp: uring: improve SendBuffers code

* udp: uring: remove unneeded squeue sync calls

* udp: uring: replace buf_ring impl with one from tokio-uring

* udp: uring: store ring in TLS so it can be used in Drop impls

* udp: uring: store BufRing in SocketWorker

* udp: uring: silence buf_ring dead code warnings, improve comment

* Update TODO

* udp: uring: improve CurrentRing docs, use anonymous struct field

* udp: uring: improve ring setup

* udp: uring: get ipv6 working

* udp: uring: make ring entry count configurable, use more send entries

* udp: uring: log number of pending responses (info level)

* udp: uring: improve comment on send_buffer_entries calculation

* udp: improve config comments

* udp: uring: add to responses stats when they are confirmed as sent

* Update TODO

* udp: uring: enable IoUring setup_submit_all

* Update README
2023-03-07 19:01:37 +01:00

9.4 KiB

aquatic: high-performance open BitTorrent tracker

CI

High-performance open BitTorrent tracker, consisting of sub-implementations for different protocols:

Name Protocol OS requirements
aquatic_udp BitTorrent over UDP Unix-like (using mio)
aquatic_http BitTorrent over HTTP over TLS (rustls) Linux 5.8+ (using glommio)
aquatic_ws WebTorrent, optionally over TLS (rustls) Linux 5.8+ (using glommio)

Features at a glance:

  • Multithreaded design for handling large amounts of traffic
  • All data is stored in-memory (no database needed)
  • IPv4 and IPv6 support
  • Supports forbidding/allowing info hashes
  • Built-in TLS support (no reverse proxy needed)
  • Automated CI testing of full file transfers

Known users:

Usage

Compiling

  • Install Rust with rustup (latest stable release is recommended)
  • Install cmake with your package manager (e.g., apt-get install cmake)
  • Clone this git repository and enter the directory
  • Build the implementations that you are interested in:
# Tell Rust to enable support for all SIMD extensions present on current CPU
# except for those relating to AVX-512. SIMD is required for aquatic_ws and
# recommended for the other implementations. If you run a processor that
# doesn't clock down when using AVX-512, you can enable those instructions
# too.
. ./scripts/env-native-cpu-without-avx-512

cargo build --release -p aquatic_udp
cargo build --release -p aquatic_http
cargo build --release -p aquatic_ws

Configuring

Generate configuration files. They come with comments and differ between protocols.

./target/release/aquatic_udp -p > "aquatic-udp-config.toml"
./target/release/aquatic_http -p > "aquatic-http-config.toml"
./target/release/aquatic_ws -p > "aquatic-ws-config.toml"

Make adjustments to the files. You will likely want to adjust address (listening address) under the network section.

Note that both aquatic_http and aquatic_ws require configuring certificate and private key files to run over TLS. aquatic_http only runs over TLS. More details are available in the respective configuration files.

Workers

To increase performance, number of worker threads can be increased. Recommended proportions based on number of physical CPU cores:

udp http ws
CPU cores (N) N N 1-7 >=8
Swarm workers 1 1 1 2
Socket workers N N N N-2

Access control

Access control by info hash is supported for all protocols. The relevant part of configuration is:

[access_list]
# Access list mode. Available modes are allow, deny and off.
mode = "off"
# Path to access list file consisting of newline-separated hex-encoded info hashes.
path = ""

The file is read on start and when the program receives SIGUSR1. If initial parsing fails, the program exits. Later failures result in in emitting of an error-level log message, while successful updates of the access list result in emitting of an info-level log message.

Prometheus

Exporting Prometheus metrics is supported. Activate the endpoint in the configuration file:

aquatic_udp
[statistics]
run_prometheus_endpoint = true
prometheus_endpoint_address = "0.0.0.0:9000"
aquatic_http / aquatic_ws
[metrics]
run_prometheus_endpoint = true
prometheus_endpoint_address = "0.0.0.0:9000"

Running

If you're running aquatic_http or aquatic_ws, please make sure locked memory limits are sufficient:

  • If you're using a systemd service file, add LimitMEMLOCK=65536000 to it
  • Otherwise, add the following lines to /etc/security/limits.conf, and then log out and back in:
*    hard    memlock    65536
*    soft    memlock    65536

Once done, start the application:

./target/release/aquatic_udp -c "aquatic-udp-config.toml"
./target/release/aquatic_http -c "aquatic-http-config.toml"
./target/release/aquatic_ws -c "aquatic-ws-config.toml"

If your server is pointed to by domain example.com and you configured the tracker to run on port 3000, people can now use it by adding its URL to their torrent files or magnet links:

Implementation Announce URL
aquatic_udp udp://example.com:3000
aquatic_http https://example.com:3000/announce
aquatic_ws wss://example.com:3000

Details on implementations

aquatic_udp: UDP BitTorrent tracker

Implements:

  • BEP 015: UDP BitTorrent tracker protocol (more details). Exceptions:
    • Doesn't care about IP addresses sent in announce requests. The packet source IP is always used.
    • Doesn't track the number of torrent downloads (0 is always sent).

This is the most mature of the implementations. I consider it ready for production use.

Performance

UDP BitTorrent tracker throughput comparison

More details are available here.

io_uring

An experimental io_uring backend can be compiled in by passing the io-uring feature. Currently, Linux 6.0 or later is required. The application will attempt to fall back to the mio backend if your kernel is not supported.

cargo build --release -p aquatic_udp --features "io-uring"

aquatic_http: HTTP BitTorrent tracker

Implements:

  • BEP 003: HTTP BitTorrent protocol (more details). Exceptions:
    • Only runs over TLS
    • Doesn't track the number of torrent downloads (0 is always sent)
    • Only compact responses are supported
  • BEP 023: Compact HTTP responses
  • BEP 007: IPv6 support
  • BEP 048: HTTP scrape support. Notes:
    • Doesn't allow full scrapes, i.e. of all registered info hashes

aquatic_http has not been tested as much as aquatic_udp but likely works fine in production.

Running behind a reverse proxy is currently not supported due to the difficulties of determining the originating IP address without knowing the exact setup.

Performance

HTTP BitTorrent tracker throughput comparison

More details are available here.

aquatic_ws: WebTorrent tracker

Aims for compatibility with WebTorrent clients. Notes:

  • Doesn't track the number of torrent downloads (0 is always sent).
  • Doesn't allow full scrapes, i.e. of all registered info hashes

aquatic_ws has not been tested as much as aquatic_udp but likely works fine in production.

Running behind a reverse proxy is supported, as long as IPv4 requests are proxied to IPv4 requests, and IPv6 requests to IPv6 requests.

Performance

WebTorrent tracker throughput comparison

More details are available here.

Load testing

There are load test binaries for all protocols. They use a CLI structure similar to the trackers and support generation and loading of configuration files.

To run, first start the tracker that you want to test. Then run the corresponding load test binary:

./scripts/run-load-test-udp.sh
./scripts/run-load-test-http.sh
./scripts/run-load-test-ws.sh

To fairly compare HTTP performance to opentracker, set keep_alive to false in aquatic_http settings.

Architectural overview

Architectural overview of aquatic

Copyright (c) 2020-2023 Joakim Frostegård

Distributed under Apache 2.0 license (details in LICENSE file.)

Trivia

The tracker is called aquatic because it thrives under a torrent of bits ;-)