I have created a small prototype where you can call a person if he/she is online on https://echo-land.herokuapp.com
It uses twilio's API to fetch STUN and TURN servers. Users create and accept offer and connect in a p2p fashion.
I have read many articles which mentions that this architecture(MESH) will fail when I create a room with multiple participant. Now I want to know that If i use Janus as a gateway where do I have to fit in to make the system functional?
In a mesh topology each user has to upload their video to each viewer, so if you had a 4 person conference call each user would need to upload video 3 times.
With Janus you upload your video once, and then Janus distributed to each viewer.
Janus has a videoroom construct that sounds like it could fit your design. You could have a user join a videoroom, and that would give you a concept of presence.