https://github.com/basho/riak_ensemble

Multi-Paxos framework in Erlang
https://github.com/basho/riak_ensemble

Last synced: 11 days ago
JSON representation

Multi-Paxos framework in Erlang

• Host: GitHub
• URL: https://github.com/basho/riak_ensemble
• Owner: basho
• License: apache-2.0
• Created: 2013-10-21T09:49:02.000Z (about 11 years ago)
• Default Branch: develop
• Last Pushed: 2022-12-21T13:23:33.000Z (almost 2 years ago)
• Last Synced: 2024-04-14T04:03:24.757Z (7 months ago)
• Language: Erlang
• Homepage: http://www.youtube.com/watch?v=gXJxbhca5Xg
• Size: 3.47 MB
• Stars: 166
• Watchers: 94
• Forks: 74
• Open Issues: 27
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
[![Erlang CI Actions Status](https://github.com/basho/riak_ensemble/workflows/Erlang%20CI/badge.svg)](https://github.com/basho/riak_ensemble/actions)

(Note: Work-in-progress documentation [here](https://github.com/basho/riak_ensemble/blob/develop/doc/Readme.md))

`riak_ensemble` is a consensus library that supports creating multiple

consensus groups (ensembles). Each ensemble is a separate Multi-Paxos

instance with its own leader, set of members, and state.

Each ensemble also supports an extended API that provides consistent

key/value operations. Conceptually, this is identical to treating each

key as a separate Paxos entity. However, this isn't accomplished by

having each key maintain its own Paxos group. Instead, an ensemble

emulates per-key consensus through a combination of per-key and

per-ensemble state.

As mentioned, `riak_ensemble` supports multiple independent consensus

groups. Ensembles are created dynamically, allowing applications to

use `riak_ensemble` in whatever way best fits that

application. Each ensemble also supports dynamic ensemble membership,

using joint consensus to guarantee consistent membership transitions.

A given ensemble is configured to use a particular "ensemble backend".

An ensemble backend is an implemented behavior that defines how a given

peer actually stores/retrieves data. For example, `riak_ensemble`

ships with the `riak_ensemble_basic_backend` which stores data as an

`orddict` that is saved to disk using `term_to_binary`, whereas

`riak_kv` implements the `riak_kv_ensemble_backend` that is used to

interface with Riak's vnodes.

Better documentation is coming soon. For now, this talk from RICON

West 2013 discusses the design of `riak_ensemble` in the context of

adding consistency to Riak 2.0:

http://www.youtube.com/watch?v=gXJxbhca5Xg