multiple video sources, multiple instances... about rpos HOT 1 CLOSED

The correct ONVIF way is to have one ONVIF server (one instance running) and then list all the available video streams in the GetProfiles ONVIF message.

There is a 2 step process in ONVIF
Step 1: Call GetProfiles. This returns a long list of available video streams.
Each Profile has a Name, a Unique UUID Token and details about the image size and video codec.
This would list all your available streams. Your CCTV software should present you with the list.
Step 2: Call GetStreamURI with the unique Token from GetProfiles to get the RTSP address

ONVIF viewing software does this Bosch do this with their 4 input ONVIF IP Encoder (4 video sources, 1 IP address, 1 ONVIF server instance)

So for your proxy situation, you could modify RPOS to have a list of streams in GetProfiles (each with a unique token) and then to check the token in GetStreamURI to deliver a RTSP address for each stream

For Pi performance, if you could return the actual RTSP address of the camera then the Pi has very little to do other than some ONVIF messages to supply the RTSP address.

If the Pi was to be a transcoding RTSP proxy then you need to think about CPU power (and GPU offload) (transcoders need to decompress the H264 back to video frames, then rescale the image to a new size and then re-compress with a H264 encoder).

If the RTSP proxy just gets video from the camera and then re-delivers it to your ONVIF viewing application then you just need to think about the ethernet port on the Pi. It is 10/100 (chained to the USB2 bus).
You need to add up the bandwidth of the stream coming in and all the streams going out and make sure you are not near the bandwidth limit of the Pi (as you need to allow for large I-frame bursts)

So, RPOS can be made to do this.

Also note RPOS runs on any NodeJS platform. I run it on my Mac and Windows machines with alternative RTSP servers (ffserver).

from rpos.