aiofastnet 0.0.7

Fast transport/protocol networking for asyncio

Introduction

aiofastnet provides highly optimized asyncio loop.create_connection and loop.create_server implementations. It is a drop-in replacement for the standard asyncio functions for performance-sensitive networking code.

Internally, it reimplements parts of CPython’s transport/SSL stack with Cython and C to reduce overhead on hot I/O paths, especially for protocol libraries that spend significant CPU time in transport and SSL plumbing.

Installation

aiofastnet requires Python 3.9 or greater and is available on PyPI:

$ pip install aiofastnet

Basic Usage

Use it similarly to stdlib asyncio APIs by passing the running loop:

import asyncio
import aiofastnet

loop = asyncio.get_running_loop()
# Instead of
# transport, protocol = await loop.create_connection(...)
transport, protocol = await aiofastnet.create_connection(loop, ...)

# Instead of
# server = await loop.create_server(...)
server = await aiofastnet.create_server(loop, ...)

Benchmark

The benchmark below compares echo round-trips over loopback for TCP and SSL. The exact gains depend on workload, message sizes, and how much time your application spends in transport/SSL overhead. Benchmark source: examples/benchmark.py

Why to use aiofastnet?

• Lower transport and SSL overhead. aiofastnet reimplements the expensive parts of CPython’s transport and SSL stack in Cython/C, so more CPU time is spent in your protocol logic instead of framework plumbing. Python libraries often pay a large premium for SSL-heavy workloads; aiofastnet uses OpenSSL more directly and avoids several extra buffer copies present in standard asyncio and uvloop SSL paths to reduce that cost.

• Clearer and safer write() / writelines() behavior. aiofastnet tries to push data to the socket before returning. If the socket can not accept everything immediately, only bytes and memoryview objects backed by bytes are retained without copying; everything else, including bytearray and memoryview objects backed by other exporters, is copied before being queued. This makes it safe to reuse application write buffers after write() or writelines() returns.

• Better backpressure semantics for SSL transports. get_write_buffer_size() reports total queued output across the whole stack, and set_write_buffer_limits() applies to the whole stack as well, so flow control reflects what is actually buffered for transmission.

• Useful for library authors, not only applications. If you build WebSocket, HTTP, RPC, database, proxy, message-broker, or custom binary protocol libraries, your users can get better throughput and lower CPU cost without changing your public API.

• Drop-in replacement for loop.create_connection() and loop.create_server(). You keep the standard asyncio transport/protocol model and your existing integration points.

In short: if your library already fits asyncio’s transport/protocol model and performance matters, aiofastnet lets you keep the same architecture while replacing one of the most expensive layers underneath it.

When Not To Use aiofastnet

aiofastnet is not always the right choice:

• If your workload is dominated by very short-lived connections. aiofastnet focuses on optimizing the data path after a connection is established; it does not primarily optimize connection setup cost. In workloads where connections are created and torn down quickly, you should expect little or no performance gain compared to standard asyncio.

• If targeting Windows ProactorEventLoop is itself a requirement. On ProactorEventLoop, aiofastnet falls back to asyncio’s native connection and server implementation underneath, so you do not get aiofastnet’s custom transport implementation on that loop.

Platform Compatibility

aiofastnet is built and tested on Linux, macOS, and Windows.

On Windows it works with SelectorEventLoop and winloop. With ProactorEventLoop it falls back to asyncio’s native connection and server implementation underneath, because the proactor loop does not provide add_reader() / add_writer() hooks required by aiofastnet’s transport implementation. For transports created through aiofastnet, a compatibility wrapper preserves the documented write() / writelines() buffer-safety behavior.

Contributing / Building From Source

Contributions are welcome!

1. Fork and clone the repository:

   $ git clone [e-mail]:tarasko/aiofastnet.git
   $ cd aiofastnet

2. Create a virtual environment and activate it:

   $ python3 -m venv aiofn-dev
   $ source aiofn-dev/bin/activate

3. Install development dependencies:

   $ pip install -r requirements-test.txt

4. Build in place and run tests:

   $ python setup.py build_ext --inplace
   $ pytest -s -v
   
   # Run specific test with debug logs enabled
   $ pytest -s -v -k test_echo[asyncio-ssl-buffered-32-64] --log-cli-level DEBUG

5. Run the benchmark:

   $ python -m examples.benchmark