https://github.com/danijar/portal

Fast and reliable distributed systems in Python
https://github.com/danijar/portal

arrays client concurrency distributed numpy parallel parallelism performance python rpc server sockets throughput

Last synced: about 2 months ago
JSON representation

Fast and reliable distributed systems in Python

• Host: GitHub
• URL: https://github.com/danijar/portal
• Owner: danijar
• License: mit
• Created: 2023-08-27T20:01:09.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2024-09-21T19:48:09.000Z (3 months ago)
• Last Synced: 2024-10-09T13:13:56.475Z (3 months ago)
• Topics: arrays, client, concurrency, distributed, numpy, parallel, parallelism, performance, python, rpc, server, sockets, throughput
• Language: Python
• Homepage: https://pypi.org/project/portal
• Size: 222 KB
• Stars: 19
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
[![PyPI](https://img.shields.io/pypi/v/portal.svg)](https://pypi.python.org/pypi/portal/#history)

# 🌀 Portal

Fast and reliable distributed systems in Python.

## Features

- 📡 **Communication:** Portal lets you bind functions to a `Server` and call

  them from one or more `Client`s. Wait on results via `Future` objects.

  Clients can automatically restore broken connections.

- 🚀 **Performance:** Optimized for throughput and latency. Array data is

  zero-copy serialized and deserialized for throughput near the hardware limit.

- 🤸 **Flexibility:** Function inputs and outputs can be nested dicts and lists

  of numbers, strings, bytes, None values, and Numpy arrays. Bytes allow

  applications to chose their own serialization, such as `pickle`.

- 🚨 **Error handlings:** Provides `Process` and `Thread` objects that can

  reliably be killed by the parent. Unhandled exceptions in threads stop

  the program. Error files can be used to stop distributed systems.

- 📦 **Request batching:** Use `BatchServer` to collect multiple incoming

  requests and process them at once, for example for AI inference servers.

  Batching and dispatching happens in a separate process to free the GIL.

- ✅ **Correctness:** Covered by over 100 unit tests for common usage and edge

  cases and used for large scale distributed AI systems.

## Installation

```sh

pip install portal

```

## Example

This example runs the server and client in the same Python program using

subprocesses, but they could also be separate Python scripts running on

different machines.

```python

def server():

  import portal

  server = portal.Server(2222)

  server.bind('add', lambda x, y: x + y)

  server.bind('greet', lambda msg: print('Message from client:', msg))

  server.start()

def client():

  import portal

  client = portal.Client('localhost', 2222)

  future = client.add(12, 42)

  result = future.result()

  print(result)  # 54

  client.greet('Hello World')

if __name__ == '__main__':

  import portal

  server_proc = portal.Process(server, start=True)

  client_proc = portal.Process(client, start=True)

  client_proc.join()

  server_proc.kill()

  print('Done')

```

## Questions

Please open a separate [GitHub issue](https://github.com/danijar/portal/issues)

for each question.