Skip to main content

Distributed computing for everyone in modern Python.

Project description

achilles

Extremely accessible distributed/parallel computing (BYOB - Build Your Own Beowulf) in modern Python based on:

  • Twisted;
  • cloudpickle;
  • multiprocessing; and,
  • lessons learned from studying Hadoop, Apache Spark, Apache Arrow, Parallel Python, Ray, Jug, Pathos, Disco, SCOOP, Cluster-Workers and the questions asked regarding these projects on StackOverflow.

The goal of the achilles framework is to drastically reduce the barriers to entry for developers to use all of the computational power available on their networks and provide engingeers with an accessible, highly flexible framework for building distributed/parallel computing into their applications applications.

As achilles continues to be actively developed, the project will continue to be designed in alignment with our four demands for the project:

  • achilles must be simple to use;
  • achilles must 'just work' at all scales, from running all of the components on an individual's laptop to running on powerful clusters in distributed systems;
  • achilles must be scalable and accommodate 'big data' workloads (i.e. files too large to open in memory on a single machine);
  • achilles should be fast as hell and regularly benchmarked against alternative solutions; and,
  • achilles must be SECURE. Currently, achilles is not suitable for deployment outside of localhost. We are currently working on an SSH implementation.

Consists of:

  • achilles_server.py
    • Runs a Twisted TCP server listening at the endpoint described in the .env configuration file. Establishes and maintains persistent connections with multiple achilles_nodes (potentially running across numerous machines), distributes computations among them at the instruction of the achilles_controller, and sends the result(s) to the achilles_controller in the response_mode specified in achilles_compute().
  • achilles_node.py
    • Connect to the achilles_server and receive ID assignment. Wait for a job to be started, receive a function over the wire, receive an argument over the wire and then keep feeding the server results/receiving new arguments until all arguments are exhausted.
  • achilles_controller.py
    • Connect to the achilles_server and instruct it which function to perform against which arguments with which response_mode. Verify the job and wait for the results to be returned - achilles_server will do the rest.
    • Available commands in achilles_controller command interface:
      • achilles_compute, cluster_status, kill_cluster, help

Configuration files:

  • achilles_config.yaml
    • Use pyYaml to load the configuration file for a job into achilles_controller.py. See below for additional information on how to structure this file.
  • achilles_function.py
    • achilles_function() is cloudpickled and distributed to all connected clients as a part of the self.startJob() handshake initiated after achilles_controller.py verifies the commencement of a job.
  • .env
    • Use python-dotenv to load the local file into the system's environmental variables for configuration and authentication.
    • Basic security precaution. To-do: explore encryption.

Installation

pip install achilles

Usage

Deploy achilles server:

  1. from achilles.lineReceiver.achilles_server import genConfig, runAchillesServer
  2. genConfig() -> enter preferred settings or use default
  3. runAchillesServer() -> run an achilles_server using the specified HOST and PORT in the generated .env file

achilles is in the early stages of active development and your suggestions/contributions are kindly welcomed. achilles is written and maintained by Alejandro Peña. Email me at adpena<3gmail.com.

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

achilles-0.0.167.tar.gz (16.0 kB view hashes)

Uploaded Source

Built Distribution

achilles-0.0.167-py3-none-any.whl (27.7 kB view hashes)

Uploaded Python 3

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