StreamMachine is an open-source streaming audio server aimed at pushing innovation for radio stations that have spent too many years running old technology like Shoutcast and Icecast.
The project has two goals: emulating the traditional streaming experience and building support for new features that push the radio listening experience forward. Currently StreamMachine supports traditional Shoutcast-style streaming and HTTP Live Streaming.
StreamMachine is being developed by Eric Richardson ([email protected]) and Southern California Public Radio. SCPR has run StreamMachine in production since 2012.
StreamMachine's big idea is keeping a big buffer of audio data in memory, allowing clients to connect to "nearly-live" radio. Because audio data is relatively small, a station can easily keep hours of audio in memory. For instance, a 64k MP3 stream will store 8 hours of audio data in just a little over 200MB of memory.
Using the Rewind Buffer, players can add support for concepts like "Play the current program from its start", or "Play the 9am broadcast." Unlike podcasts, these functions are available immediately and keep the user connected to the station's live stream.
You can read more about the implementation of the RewindBuffer in the wiki.
StreamMachine includes support for writing analytics into Elasticsearch. Output includes listening events—chunks of listening that reflect a segment of duration delivered—and rolled-up sessions.
Unlike traditional streaming log outputs, StreamMachine's listening events allow deep realtime analysis of user beahvior, including playhead position and in-flight session durations.
With the new Source Mounts support in 0.7, StreamMachine now supports taking one input source and transcoding it into the variety of output formats you may want to offer your listeners.
StreamMachine is a Node.js application. It can run on one server, or in a master-slave configuration for load-balancing. StreamMachine is designed for Node 0.10 and higher.
Static configuration can be done via a JSON configuration file. Configuration changes will not be written back to the file.
For a dynamic configuration, StreamMachine can store its stream information on a Redis server.
Two modes are supported for incoming audio:
-
Icecast Source: The
SourceInmodule emulates the Icecast server's source handling, allowing source clients to connect to a mount point and provide a source password. -
Proxy: The
ProxyRoommodule can connect to an existing Icecast server and proxy its stream through the StreamMachine infrastructure.
Currently, MP3 and AAC streams are supported.
StreamMachine can operate as a single process or in a master-slave configuration.
-
Standalone: One process manages both incoming and outgoing audio. For development and small installations.
-
Master: Master handles source connections and configuration. Provides the admin UI and centralizes logging from slave processes. Handles no client traffic.
-
Slave: Slave handles all client requests. Stream audio is proxied from the master over a single socket connection.
For more, see the Mode documentation in the wiki.
-
Raw Audio: Realtime audio stream output appropriate for HTML5 audio tags and other web players
-
Shoutcast: Realtime audio stream that injects Shoutcast-style metadata at intervals
-
HTTP Live Streaming: Apple's HTTP Live Streaming protocol uses playlists and audio segments to replace long-lived connections. StreamMachine supports segment chunking and can produce HLS playlists for individual streams and adaptive stream groups.
-
Pumper: Allows chunks of audio to be downloaded as quickly as possible
Alerts are intended to provide a heads-up when something's going wrong in the streaming process. A similar alert will fire when the condition is ended. Where logging is intended to signal an event, alerts are about signalling that a condition exists.
Alerts can be sent via email or Pagerduty.
StreamMachine uses nconf to load configuration settings from a JSON file, from environment variables and from the command line.
For more, see the full documentation of configuration options.
Traditional streaming is done using long-lived connections that can make life painful when you need to deploy a new code release. StreamMachine is able to accomplish live restarts by doing a "handoff" where the old process passes all clients to the new process before shutting down.
To manage this process, StreamMachine ships with the "streammachine-runner"
command. The runner is a lightweight process that will manage the StreamMachine
process, initiating a handoff based on either a SIGHUP or a change to a
watched trigger file.
StreamMachine isn't the most user-friendly piece of software to install at the moment, but there are two options for quickly getting something running that you can play with:
To run StreamMachine locally with no service dependencies, try the included config/standalone.json. To do so:
- Install Node.js
- Download the StreamMachine repository
- Run
npm installto install the Node modules that we depend on - Run
./streammachine-cmd --config ./config/standalone.jsonto start the StreamMachine service
The included configuration has an MP3 stream configured at /test, with the source password "testing". The source input listener is on port 8002, and that is where the broadcast should be pointed. To connect and stream audio, you can use a variety of input stream providers:
./streammachine-util-cmd source --port 8002 --stream test \
--password testing ./test/files/mp3/mp3-44100-128-s.mp3
To stream via VLC or Liquidsoap for supporting playlists and multiple mp3 files, please refer to the documentation.
To listen, point an audio player to http://127.0.0.1:8001/test.
You can also use the Vagrantfile included in streammachine-cookbook to install a standalone configuration in a virtual machine.