Performance
What is RTT?
RTT is the time for a packet to travel to a destination and for the response to return, measured in milliseconds. The physical floor is about 1 ms per 100 km each way: light in fibre travels at roughly two-thirds of c and real routes are around 1.35 times the great-circle distance.
Why RTT matters
The physical floor is about 1 ms per 100 km each way: light in fibre travels at roughly two-thirds of c and real routes are around 1.35 times the great-circle distance. No provider beats that floor, which is why placement decides more about perceived speed than any hardware specification. What a good network does is get close to the floor — direct peering rather than three transit hops, and uncongested links that do not add queueing delay on top of distance. Note that ICMP ping and application-layer round trip are not the same measurement: routers routinely de-prioritise ICMP, so a ping can look worse than the TCP handshake your users actually experience. Every OnionVPS location page publishes an estimated RTT to every other region, computed from the formula above so the number is reproducible rather than promotional.
RTT in practice
When running a server, RTT is the ground truth for user-perceived latency. I watch it from a few vantage points, not just my own datacenter. If my app's RTT to a region is high, I first check whether the path is taking three transit hops or congested links; that is queueing delay I can fix. If it's distance, I move the instance. Getting it wrong means a slow-feeling service that no hardware spec fixes, because physical law sets the floor. The good news: RTT from an ICMP ping can mislead, since routers de-prioritise that traffic; measure with a real TCP handshake instead.
What people get wrong about RTT
The common error is treating ICMP ping as the definitive RTT number. Many assume a ping tells them what users experience, but routers de-prioritise ICMP, so the ping can look worse than the actual TCP handshake. The correction: use application-layer measurements for real user-facing latency.
RTT — common questions
How is RTT related to ping?
RTT and ping are often used interchangeably, but ping specifically refers to ICMP echo requests. RTT is the broader concept covering any round-trip, including TCP handshakes. Ping can be unreliable because routers de-prioritise ICMP, so it may overstate the RTT users actually experience.
What is the minimum possible RTT?
The physical floor is about 1 ms per 100 km each way, with real routes about 1.35 times the great-circle distance. No provider can beat this because light in fibre travels at two-thirds of c. Placement matters more than any hardware.
More from performance
- CPU steal time
- Steal time is the proportion of time a virtual CPU was ready to run but the hypervisor gave the physical core to another guest.
- Jitter
- Jitter is variation in packet delay over time, as distinct from the average delay itself.
- TTFB
- TTFB measures the time from a request being sent to the first byte of the response arriving.