Low RTT but sluggish remote mirroring? Understand WebRTC paths and latency

Learn why WebRTC may use a relay, how to identify the selected path, and why network RTT is not click-to-picture latency.

A remote phone can show “1 ms” latency and still feel sluggish. The number may measure only a network round trip, not the complete journey from your click to the updated picture on your computer. Before deciding whether latency is good, ask what was measured.

This guide draws on WebRTC and MDN documentation checked on September 13, 2026. It explains general concepts for WebRTC-based mirroring and control tools—not a particular product implementation or a device benchmark.

WebRTC and relay are not opposites

WebRTC can use a TURN server to relay traffic when a direct connection cannot be established. A WebRTC label therefore does not prove that traffic bypasses a relay. Using TURN does not, by itself, mean the session has stopped using WebRTC. WebRTC TURN guide

Signaling is another distinct part: it exchanges information needed to establish a connection. Reaching a website or seeing a peer online is not enough to establish that the picture path is ready. WebRTC connections and signaling

Separate the evidence into three layers:

Layer Question to ask Conclusion to avoid
Contact Has signaling been exchanged? The website loads, so video must work
Transport Which path is actually selected? A direct candidate exists, so it is in use
Operation Did the phone respond and the picture update? Low network RTT guarantees responsive control

This is a diagnostic sequence, not advice to disable a firewall or open every port.

Check the selected path, not any candidate

For tools exposing browser statistics, inspect the selected candidate pair. Developers can use the transport statistic selectedCandidatePairId to find the associated pair statistics; older browsers may not expose it. MDN field reference

Then inspect the local and remote candidate types for that pair. A relay type identifies a TURN relay candidate. Merely gathering one is not evidence that it is selected. MDN candidate types

That observation applies to the WebRTC connection inspected. Separate proxies, application-level forwarding and other connections require the application's own routing evidence.

What RTT leaves out

currentRoundTripTime is expressed in seconds and calculated from a recent relevant STUN request and response. An example value of 0.025 means 25 ms. It measures the pair's network round trip, not click-to-picture latency. MDN RTT definition

Consequently, that field alone cannot tell you whether queuing, device processing, capture, encoding, decoding or display is slow. For a constant 0 or 1 ms display, maintainers should check endpoints, units, rounding and whether missing data is being replaced with zero. These are investigation questions, not a diagnosis of a specific incident.

A useful support report

Start with a small, time-matched sample:

  • Software versions at both ends; local or remote use.
  • The reported mode and selected path—or “unknown.”
  • Whether the problem is slow startup, ongoing stutter or unresponsive control with a working picture.
  • The latency metric's name, unit and timestamp, with a normal and an abnormal sample.
  • Whether one harmless action took effect on the phone and when the computer showed it.

If both screens are nearby, one camera recording the input action, phone and computer can help show where the delay appears. Treat this as approximate evidence, not a precision benchmark. Do not subtract timestamps from separate devices unless their clocks are appropriately synchronized.

Remove IP addresses, session identifiers, tokens and private screen content before sharing a report. Low RTT is a clue, not proof that the whole mirroring pipeline is healthy.

参考来源