Skip to content

Experimental Evaluation

crushr architecture was selected through destructive testing.

Method

  • generate archive
  • apply deterministic corruption
  • measure recovery

Metrics

Metric Meaning
recovery_rate % of data recovered
name_retention % of filenames preserved
false_positive incorrect recovery

Result summary

  • distributed identity consistently outperformed central metadata
  • mirrored dictionaries retained names under partial corruption
  • FORMAT-15 failed due to leadership dependency

Conclusion

Architecture is evidence-driven, not theoretical.