-
Architecture patterns for serving a building digital twin to many anonymous web
We are designing a building-management digital twin SaaS with these requirements:
- A large, unpredictable number of anonymous users view the 3D building model in a web browser, each with an independent camera
- Future phases add server-side simulation (HVAC/airflow CFD, people flow) and physical AI (robots using the twin as world knowledge)
Our current thinking: client-side rendering (web-native 3D + BIM-derived glTF/tiles from a CDN) for mass viewing, with heavy computation done server-side and only the results (fields, time series, object states) pushed to clients. GPU pixel-streaming would be reserved for a small number of premium seats, since per-session GPU cost scales linearly with concurrent users.
Questions for the community:
- Has anyone operated a building/infrastructure twin at scale with anonymous public access? Did you go client-side rendering, server-side streaming, or a hybrid — and would you make the same choice again?
- For those who run server-side simulation (CFD, crowd sim, energy models): how do you deliver results into a web viewer? Pre-baked meshes on a CDN, live data over WebSocket, or something else?
- Any lessons on keeping the “source of truth” (BIM master + semantic graph + live state) decoupled from the delivery layer, so the renderer stays replaceable?
Vendor-neutral experiences especially welcome — we’re trying to validate the architecture before committing.
-
0 Replies
Sorry, there were no replies found.
Log in to reply.