Network time analysis
Open a log or capture, or start from a demo. Everything runs on this machine.
Offset
Overlapping ADEV
Time error TE = local − reference; 0.1 Hz filter (ITU-T G.8260 definitions)
Offset reference − local · drag to zoom, double-click to reset
Stability
Dynamic stability (sliding window)
Power-law noise model S_y(f) = Σ h_α f^α, fitted to OADEV; dashed curves
Values, intervals and noise per τ
Offset distribution
With the selected detrending; the dashed curve is a Gaussian of equal mean and σ. Heavy tails mean queueing spikes or path changes.
Offset vs. queueing delay (wedge)
Round-trip delay
Floor packet percentage
One-way delays extracted per direction; floors assumed equal
Replay in the bench →Power spectral density
Predicted time error after loss of reference
Every loaded dataset; click a row to analyse it, tick to overlay it on the Analyze charts.
Every loaded dataset as one host, ranked by how close it is to the nearest limit. Load one log per host (or a Prometheus range query of the exporter with one series per instance); click a row to analyse it.
Difference over time
TDEV and MTIE of the difference
Individual stability of each source
TE and TEL
MTIE and TDEV of dTE_L
Error bound to UTC and measured offset
Rules applied
Worst samples
Windows
A free-running clock measured through a modelled network; every dataset carries the true offset, so filters can be scored.
Scenarios
Estimators
Errors are estimate − true offset after a 30-minute warm-up, mean over seeds; the bar is the RMS on a log scale, shortest is best.