The Physical Layout¶
The hardware we use is described in our general documentation.
The Ansible configuration sets up 9 groups of machines:
- DHCP and TFTP Server
- This is the gateway between the internet and the video team network. It enables PXE booting the other local machines and configuring them using Ansible automatically.
- Opsis
- These are the PCs that are connected to the Numato Opsis boards and capture the presenter’s laptop output for streaming to Voctomix.
- Voctomix
- These are the PCs that live-mix the video from the cameras and opsis capture for recording to disk and live streaming to the internet.
- NFS Server
- This provides an NFS share for SReview, as it requires a common file system between nodes.
- Grid Engine Master
- This is the master node controlling the grid engine that manages our encoding machines.
- Encoder
- These are the encoding nodes that are added to the grid engine and encode the recorded talks for review and upload.
- Streaming Back-end
- This receives the RTMP streams from the rooms, saves these to disk and presents this over HTTP using DASH in a variety of formats.
- Streaming Front-end
- These machines are caching proxies in front of the streaming back-end, geographically distributed around the world. Users probably connect to these, not the back-end, when they exist.
- Review
- The machines that will host and manage SReview, which is our review system for talks after they are recorded.
Using all 9 groups in a full-blown conference gets complicated very quickly. For this reason, we will go through two much simpler streaming and recording examples here:
- A single Voctomix machine that stands alone.
- Voctomix and Opsis machines with a gateway to the outside world.