What the experiments show
Missed real-time deadlines by scenario
Misses per 1,000 messages of the time-sensitive flow. A message misses if it arrives more than 150 ms after it was sent, or never arrives.
TCP recovery after total outages
Time from the moment both paths returned until the receiving application got data again. The outage itself (12 to 40 s) is not included.
What the protection costs
Two separate measurements per scenario (medians of runs). Extra bytes come from copies sent when a path loses packets, so they appear in the loss scenarios (marked ●).
Network bytes per application byte
Tunnel CPU per endpoint (% of one core)
Scope
Replay a recorded scenario
Two separately recorded runs of the same scenario, upstream mqvpn and LinkShift, on one time axis. Charts are drawn from the saved measurements.
Run an experiment now
Run one scenario on the emulated network (about one minute). Live runs are stored separately from the recorded evaluation.
Run
Previous live runs
The timing check shows whether the test machine's timers stayed within the project's limits during a run. It describes the test environment, not the network result.
Method and provenance
How the system and the experiments work, how each number is measured, and where it comes from.