Live network simulation
Mesh routing vs centralised data centres
A request from a client hops through the nearest relay, then to the nearest available GPU provider. Compared with a typical centralised inference call out to a US-east data centre.
Requests routed
0
Mesh avg
0 ms
Centralised avg
0 ms
Speedup
—
Client Relay region GPU provider Centralised DC
Mesh route (current request)
Tallinn → relay Stockholm → gpu Frankfurt
27 ms
Centralised baseline
Tallinn → public internet → US East (Virginia)
110 ms