Hierarchical Forecasts of Cloud Storage Demand with Cost-Sensitive Loss: A Reproducibility-Centred Evaluation

Abstract

Forecast errors at individual storage pools can accumulate into expensive cloud-capacity decisions. The central research objective is to compare hierarchical demand forecasting strategies under asymmetric penalties for shortages and excess capacity. Previous work provides context for sampling, input validation and study boundaries [3–7]. To answer the question, the study proposes a rolling-origin evaluation of anonymised utilisation traces across storage pools and service classes. Related enterprise-storage studies inform the assessment of operational reliability [1] and the deployment assumptions adopted for system-level comparison [2]. Under matched evaluation conditions, the protocol compares seasonal baselines, independently fitted forecasts and partially pooled hierarchical models. Methodological sources are used to specify transparent comparison and reporting practices [8–12]. The reproducibility extension records dataset versions, model settings, exclusion rules and provenance before any held-out evaluation. Performance will be judged using capacity breach rates, forecast bias, tail-error cost and the amount of unused provisioned capacity. Interpretation will account for workload migrations, discontinuous growth, changes in retention policy and service-level constraints. An independent replication exercise would repeat the main analysis from the documented artefacts and quantify disagreements. 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].