Round-Trip Time
also called RTT, Latency Floor
The time for a packet to travel to a destination and back — a physical floor that no application optimisation can reduce.
Light in fibre travels roughly 200 km per millisecond, and real paths are not straight. London to New York is around 70–80 ms round trip; London to Sydney around 250 ms. Those numbers are not negotiable.
The architectural consequence is that the only way to reduce latency below the RTT floor is to move closer — a CDN edge, a regional deployment, an edge function — or to remove round trips altogether.
Removing round trips is usually the larger win and the more available one: connection reuse (a handshake is one to three RTTs), HTTP/2 or HTTP/3 multiplexing, avoiding CNAME chains, batching requests, and eliminating sequential dependent calls in favour of parallel ones.
The diagnostic habit: before optimising server-side processing, work out how many round trips the interaction requires and multiply by RTT. Frequently that product, not the compute, is the latency.