Improve scalability (M:N:K) - M apprtc edges, N signaling nodes, K sfu workers

[rainliu]

Current setup (1:1:K) - 1 apprtc + 1 signaling deployed at the same host (appr.tc pointing to 1 IP host), 2 sfu workers deployed at two different hosts

Ideal setup (M:N:K) - M apprtc edges sit at appr.tc (DNS load balanced to M IP hosts), N signaling nodes sit at xxx.xx (DNS load balanced to N IP hosts), K sfu workers sit at sfu.rs (DNS load balanced to K IP hosts)

Problem:

• how to make sure that different clients with the same room id accessing via https://appr.tc/v2/r/{room_id} reach the same signaling host? different clients, even with the same room id, may reach different apprtc hosts due to appr.tc DNS load balance.
• The issue only exists between AppRTC edges and Signaling nodes (m vs n mapping), not between SFU workers and Signaling nodes since each signaling node has in-memory room to sfu worker mapping list.
• can this issue be solved without external services?

[rainliu]

addressed in design doc https://github.com/webrtc-rs/apprtc/blob/master/docs/design/apprtc-signaling-architecture-design.md 's
section 9. Horizontal scale: M apprtc edges, N signaling nodes, K SFU workers