Skip to main content

Scale Distribution Framework

Project description

Scaled

This project is aiming the target that provides simple and efficient and reliable way for distributing computing framework, centralized scheduler and stable protocol when client and worker talking to scheduler

Introduction

The goal for this project should be as simple as possible

  • It built on top of zmq
  • it has ready python version of Client, Scheduler, Worker
  • I will provide golang or Rust version of Scheduler, the goal for the Scheduler should be completely computer language agnostic, which means they follow the same protocol
  • Scheduler might support function based computing tree in the future

Installation

pip install scaled

if you want to use uvloop, please do: pip install uvloop, default we are using python builtin uvloop

How to use it

Start local scheduler and cluster at the same time in the code

import random

from scaled.client import Client
from scaled.cluster.combo import SchedulerClusterCombo


def calculate(sec: int):
    return sec * 1


def main():
    address = "tcp://127.0.0.1:2345"

    cluster = SchedulerClusterCombo(address=address, n_workers=10, event_loop="uvloop")
    client = Client(address=address)

    tasks = [random.randint(0, 100) for _ in range(100000)]
    futures = [client.submit(calculate, i) for i in tasks]

    results = [future.result() for future in futures]

    assert results == tasks

    client.disconnect()
    cluster.shutdown()


if __name__ == "__main__":
    main()

Start scheduler and cluster independently

use scaled_scheduler to start scheduler, for example:

scaled_scheduler tcp://0.0.0.0:8516

use scaled_cluster to start 10 workers:

scaled_worker -n 10 tcp://127.0.0.1:8516

for detail options of above 2 program, please use argument -h to check out all available options

Then you can write simply write client code as:

from scaled.client import Client


def foobar(foo: int):
    return foo


client = Client(address="tcp://127.0.0.1:2345")
future = client.submit(foobar, 1)

print(future.result())

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

scaled-0.13.tar.gz (23.7 kB view details)

Uploaded Source

Built Distribution

scaled-0.13-py3-none-any.whl (34.2 kB view details)

Uploaded Python 3

File details

Details for the file scaled-0.13.tar.gz.

File metadata

  • Download URL: scaled-0.13.tar.gz
  • Upload date:
  • Size: 23.7 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/4.0.2 CPython/3.11.1

File hashes

Hashes for scaled-0.13.tar.gz
Algorithm Hash digest
SHA256 79bc04cee505b588b5273e46a14480fab67bc788f9594342ac57da97104cec99
MD5 127774c3a62a4192d756363fa6797d4c
BLAKE2b-256 c3b585708343043e584eec647dd45a1eee794412279fd7088a97f90aba742822

See more details on using hashes here.

File details

Details for the file scaled-0.13-py3-none-any.whl.

File metadata

  • Download URL: scaled-0.13-py3-none-any.whl
  • Upload date:
  • Size: 34.2 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/4.0.2 CPython/3.11.1

File hashes

Hashes for scaled-0.13-py3-none-any.whl
Algorithm Hash digest
SHA256 eb6c40f1091e7715fff1000612d3799be76c6e395fab12b2fff8d4383469fd36
MD5 20726dda5c477a192cbb65b98b385288
BLAKE2b-256 0155c314894f12bc03ad0f7c96b77249aae87e75ce43a66d93fe8bcc68bcc8f7

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page