Permalink
Fetching contributors…
Cannot retrieve contributors at this time
84 lines (60 sloc) 3.12 KB

Overview

QMachine_ (QM) is a web service for distributed computing. Its design relaxes the usual requirements of a distributed computer so far that it can be powered completely by web browsers -- without installing anything. As a model for computation, QM has been detailed in a recent paper, QMachine: commodity supercomputing in web browsers. This manual details QM as a software platform, with particular focuses on how it works, how to use it, and how to contribute to the open-source project.

How it works

Coordination of a distributed computing effort requires a common data interchange format and a common communications protocol. As a web service, QM provides an Application Programming Interface (API) that uses JavaScript Object Notation (JSON) as its common data interchange format and Hypertext Transfer Protocol (HTTP) as its communications protocol. In short, QM's API allows JSON-encoded messages sent to URLs over HTTP to become a message-passing interface for distributed computing. The use of web-friendly technologies is deliberate -- QM is designed to be as universal as the World Wide Web itself.

Most QM-related software falls into two categories. API clients are programs that consume QM's API, and API servers are programs that provide the API. Web servers represent a third category, however, that is important only for hosting content such as the browser client library.

API clients

An API client is a program that consumes QM's API by sending JSON-encoded messages over HTTP to specific URLs, as defined by QM's :doc:`http-api`. The vast majority of programs will fall into this category, and most of these programs will use a client library for convenience. Currently, the only client library supported by the QM project is the :doc:`browser client <browser-client>`, which is written completely in JavaScript. An outdated Node.js_ client is now being resurrected, and a Ruby_ client is planned.

API servers

An API server, by way of contrast, is a program that provides QM's API by listening for and responding to specific HTTP requests, as defined by QM's :doc:`http-api`. There are two implementations to choose from: the :doc:`original reference version <nodejs>` written in Node.js and the :doc:`"teaching version" <ruby>` written in Ruby. The Node.js version is recommended for production.

Web servers

For convenience in hosting, basic web servers are packaged alongside the API servers. A web server listens for and responds to HTTP requests for files and other resources that are published online. A web server is not strictly necessary as part of QM, but it is useful for making the browser client available to web browsers. The web servers bundled with the API servers are present only for convenience, and it is trivial to use off-the-shelf web servers like Apache_ or Nginx_ instead. (In fact, QM's own homepage uses Nginx.)

How to use it

See :doc:`local-sandbox` and :doc:`paas-sandbox`.

How to contribute

See :doc:`source-code`.