№ 002 · Results

Same hardware, 1.34× the useful work

Dusk simulates 10,000 missions on processors that fail and are never repaired, with no ground support after launch. A shrinking quorum delivered a third more useful work than fixed triple redundancy.

The question

Fixed triple modular redundancy has a fixed threshold. Three processors vote. Two can still detect a disagreement. One cannot outvote anything, so the system stops.

On a mission with no repairs and no ground team in reach, that throws away the last processor's whole remaining life. The whitepaper proposes a shrinking quorum instead: vote while three are alive, compare while two are, self-check on one. Falling back from TMR to simplex is not new: the Space Shuttle's computers did it. The question is what it is worth when no ground team can command it, and how it compares with the autonomous designs spacecraft fly.

The model

Each mission is a day-by-day fault-injection run. Processors die on a Weibull lifetime. Deaths can be correlated, as a shared thermal or power fault would be. Upsets corrupt a day's result at the rate that gets past EDAC and scrubbing.

Every policy runs against the same fault history and the same upsets. Any difference between policies is the policy.

The result

Mean useful years per mission, over 10,000 missions:

Fixed TMR99.0
One processor, spares83.1
Mru shrinking quorum132.6

That is 1.34× the useful work of fixed TMR (95% CI 1.33× to 1.35×). It assumes no ground support after launch. With ground support that never ends, the gain falls to 1.00×: see when the ground can help. The quorum is never less on a mission by construction in this model, because the two policies are identical until TMR stops.

Useful output over the mission for five policies. The shrinking quorum stays above fixed TMR after year 50 and keeps working long after TMR has stopped. Two dashed baselines: TMR with an unchecked fallback, and cold standby.
Fig. 1 · Useful output over the mission, mean of 10,000 missions. The shaded area is total useful work. Drawn by dusk --bin charts; nothing by hand.

The cost

A lone self-checking processor misses a few of its upsets. Those become wrong results that nobody knew were bad. Fixed TMR never produces them, because it never lives long enough.

So the honest comparison is a price. Under the shrinking quorum, 20.1% of missions (95% CI 19.3% to 20.9%) deliver at least one wrong result. Under fixed TMR, 0.0% do. Fixed TMR comes out ahead only if one wrong result costs more than 136 years of useful work (95% CI 131 to 142). Against one processor with spares (standby simplex), the shrinking quorum makes 59% fewer wrong results.

Against autonomous baselines

Fixed TMR is one baseline, not what most spacecraft fly. Two more baselines also act with no ground team, on the same fault histories.

PolicyUseful yearsMissions with a wrong resultShrinking quorum's work against it
Fixed TMR99.00.0%1.34×
TMR, unchecked fallback166.287.2%80%
Cold standby, spares aging 0.1144.761.3%0.92×
Cold standby, spares aging 1.083.143.2%1.60×
Shrinking quorum132.620.1%·

TMR with an unchecked fallback steps down on its own but never checks its last processor. It does more useful work at every setting, with about 100 times the wrong results: the shrinking quorum makes 1.00% of them (95% CI 0.96% to 1.04%). That gap is what self-check buys.

Cold standby keeps one processor powered and switches on the next when it dies, the nearest model here of dual-string with autonomous switchover. An unpowered spare ages more slowly than a powered one, but nobody knows how much, so the rate is swept from 0.1 to 1.0. Below about 0.20, cold standby does more useful work than the shrinking quorum. Above it, the shrinking quorum does more. At every rate, cold standby has more missions with a wrong result, because it never votes.

Cold standby's useful work relative to fixed TMR, by spare aging rate, with the shrinking quorum and standby simplex as reference lines. Cold standby crosses the shrinking quorum near an aging rate of 0.2.
Fig. 2 · Cold standby by spare aging rate, three strings, 10,000 missions per point. Drawn by dusk --bin charts.

When the ground can help

Real spacecraft do not run TMR alone. When a string fails, a ground team commands a fallback by hand. Dusk adds that team as a baseline, for as long as support lasts.

Ground support endsGain over TMR
At launch1.34×
After 25 years1.31×
After 50 years1.25×
After 100 years1.11×
Never1.00×

Answer time barely matters. A fallback a day or six months late gives the same result to two decimals. What matters is whether anyone is still there.

How sure are we

The sweep varies every parameter the model cannot pin down. Across every row, the shrinking quorum returned 1.11× to 1.68× the useful work of fixed TMR. The gain is largest with early, random failures and smallest with sharp wear-out, where all three processors die close together.

Sensitivity of the gain over fixed TMR to each uncertain parameter. Every row stays above 1.1 times.
Fig. 3 · Sensitivity. 10,000 missions per row, seed 1.

What this does not claim

The parameters are plausible orders of magnitude, not values fitted to flight data. The claim is the shape of the trade, and the sweep shows where it holds. Power, thermal limits and duty cycling are not modeled. The ground team in the baseline is perfect, which flatters the ground. The aging rate of an unpowered spare is not measured. Correlated deaths in cold standby follow the powered fault history, which slightly under-counts them.

Cite this

@software{binns2026dusk,
  author = {Binns, Will},
  title  = {dusk: A Monte Carlo of Redundancy Policies
            on Hardware That Only Ever Decays},
  year   = {2026},
  url    = {https://github.com/mruspace/dusk}
}

Sources

  1. R. E. Lyons and W. Vanderkulk, “The Use of Triple-Modular Redundancy to Improve Computer Reliability,” IBM J. Res. Dev., 1962.
  2. W. Binns, “Mru: A Fault-Tolerant Operating System for Thousand-Year Autonomous Operation,” 2026. doi:10.5281/zenodo.23305299
  3. J. F. Meyer, “On Evaluating the Performability of Degradable Computing Systems,” IEEE Trans. Computers, 1980.
  4. B. Efron, “Bootstrap Methods: Another Look at the Jackknife,” Annals of Statistics, 1979.

→ Get in touch

Questions about these results?

Tell us about the hardware, where it is, and what you need. We will tell you plainly whether Mru fits, including when it does not.

  • We'll review your setup.
  • We'll respond in one business day.
  • No project yet? All questions welcome.

About: № 002 · Dusk results

Interested in

We use these details only to reply. No mailing list.