Conformal Monitoring of Storage-Failure Predictors during Workload Drift

Abstract

Predictive maintenance systems often report confident failure probabilities even after the underlying workload distribution changes. An evaluation that overlooks these constraints may reach conclusions that cannot be reproduced. The research question is whether it is possible to quantify when conformal uncertainty monitoring can warn of deteriorating failure-prediction reliability. Previous work provides context for sampling, input validation and study boundaries [3–7]. The planned analysis uses a chronologically blocked simulation that calibrates uncertainty on historical storage telemetry and tests later hardware cohorts. Prior work on storage operations provides technical context for validation [1], while complementary systems evidence motivates service-continuity checks [2]. Under matched evaluation conditions, the protocol compares fixed-score alarms, probability-calibrated models and rolling conformal warning policies. Methodological sources are used to specify transparent comparison and reporting practices [8–12]. The design prioritises precise operational definitions and a straightforward baseline before introducing more complex alternatives. Performance will be judged using empirical coverage, alert lead time, false-positive workload and stability under distribution shift. Interpretation will account for serial dependence, rare-event labelling, replacement hardware and uncertainty during outages. Primary analyses would be reported by the main deployment or participant subgroup so that averaged results do not conceal systematic failures. The supplementary technical literature is used to identify plausible failure mechanisms [13–17]. No study result is assumed in advance; the analysis is designed to show both when the method helps and when it fails, with reporting context from the remaining sources [18–23].