VLMa/Documentation:Architecture

From VideoLAN Wiki
Revision as of 20:47, 14 February 2009 by Jpountz (talk | contribs)
Jump to navigation Jump to search

Architecture

The model

The 2 key objects of the model are the media and the programmation. Basically, a media stands for something that can be streamed whereas the programmation tells you how it should be streamed.

The threads

The daemon is made of two main threads:

  • a thread which monitors the state of the servers (up/down),
  • a thread which monitors the state or programmation (whether programs are really streamed or not) and asks for orders to be given again whether there is something wrong.

Most events are triggered by these two threads, the other events being triggered from user interaction.

Orders

When VLMa notices that some programs are not really streamed, orders are computed and then sent to the servers. Computing orders means:

  • identify available adapters,
  • identify medias that should be streamed,
  • put together medias into groups of medias that should be streamed by the same server (for example, because of their frequency in case of a DVB input),
  • map every media group to an available adapter,
  • send orders to the servers.

A programmation state

A media can have 3 states:

  • not programmed,
  • programmed but not streamed,
  • programmed and streamed.

To check whether programs are really streamed or not, there are stream watchers. Currently, there are two main implementations : the DirectMulticastStreamWatcher which joins the multicast group of the program (in case of an UDP multicast streaming strategy) and tries to receive data. If not enough data is received, the stream is known as not streamed. The other watcher, HttpStreamWatcher connects on the VLC http interface and verifies that no 404 error is encountered.

See also