Slide contents

1. 1. Modeling Reliability from a Dependency Graph

   Alexander Polomodov × Anatoly Krasnovsky

2. 2. The graph narrows the chaos search

3. 3. Source version changes the conclusion

4. 4. A hand-built model ages before the system

5. 5. Telemetry becomes an executable model

6. 6. The graph sees only the mandatory path

7. 7. Version 1 fixes the failure count

8. 8. Model and system measure one outcome

9. 9. DeathStarBench is complex but controllable

10. 10. At p=0.3 the model matched the system

11. 11. Five rates preserve the overall trend

12. 12. Bias changes across failure rates

13. 13. Replicas help inside visible structure

14. 14. Version 2 changes the failure model

15. 15. Shared fate lies outside the call graph

16. 16. Uncertainty forms the chaos backlog

17. 17. Research becomes a platform pipeline

18. 18. A pilot starts with one operation

19. 19. Build a model before chaos engineering

    Paper · ACM Digital Library

    Code and data · GitHub

    Author · Telegram

    The graph selects hypotheses. Chaos checks reality.