Right-sizing Recommendations (RSR): Cloud Workload Conformal Prediction for Virtual Machines in Data Center Operations
06:00 · July 29, 2026 · arXiv cs.AI RSS
Managing cloud infrastructure efficiently, especially in environments of large cloud providers or hyperscalers, requires optimizing the use of physical resources to minimize costs and maximize performance. Selecting the right virtual machine (VM) sizes is crucial to achieving cost efficiency in these dynamic environments. However, traditional VM allocation and scheduling approaches often fail to account for the fluctuating and unpredictable nature of VM utilization, leading to inefficiencies such as over- or under-provisioning of resources. High-quality interval prediction helps accurately capture uncertainty in cloud resource demand and supports cloud operators in efficient instance provisioning. As an effective and reliable framework for constructing prediction intervals (PIs), conformal prediction (CP) is used for mid- and long-term forecasting tasks in cloud computing environments. This study proposes a new data-driven PI construction approach using bootstrapping conformal prediction for modern, dynamic, data-driven Right-sizing Recommendations (RSR) to enhance provisioning for diverse application workloads on hyperscalers. By learning workload utilization patterns, identifying correlations across multiple time series, and predicting medium- to long-term utilization trends, this research seeks to improve the efficiency of cloud and data center operations through an AI/ML-based provisioning pipeline. Our study demonstrates that AI-driven models, powered by machine learning regression techniques and evaluated using backtesting, achieve promising forecasting results for cloud resource utilization. Additionally, we rank the selected models to identify top-performing approaches for long-life VM candidates. The proposed framework enhances right-sizing recommendations and supports more cost-effective resource allocation in dynamic cloud environments.
Summary
Right-sizing virtual machines remains a core operational challenge for hyperscale cloud providers because static allocation rules cannot keep pace with fluctuating, tenant-specific CPU and memory demand. Over-provisioning wastes physical capacity and energy, while under-provisioning risks service-level violations. The paper therefore frames right-sizing as an uncertainty-aware forecasting task and proposes an end-to-end pipeline that combines machine-learning regression with conformal prediction to produce calibrated prediction intervals for mid- and long-term horizons.
Historical Azure VM telemetry is first treated as univariate time series. Several regressors, including gradient-boosted tree ensembles, are trained to forecast CPU utilization. Conformal prediction, augmented by bootstrapping, then wraps each point forecast with statistically valid intervals that quantify the range within which future utilization is expected to lie. These intervals feed a downstream provisioning stage that applies explicit constraints, such as keeping the 99th-percentile CPU load below 70 percent and peak memory below 60 percent of the target instance capacity, thereby translating forecast uncertainty directly into instance-size decisions.
Model selection relies on backtesting tailored to temporal data rather than conventional cross-validation. Performance is assessed both by point-forecast error metrics and by interval-quality measures that penalize overly wide or poorly calibrated bands. The resulting rankings highlight a small set of consistently strong regressors for long-lived VMs, offering practitioners concrete guidance on which algorithms merit deployment at scale. The authors also release the associated code and curated Azure-derived data sets to support reproducibility and further experimentation in production environments.
Why it matters
Provides actionable, uncertainty-quantified forecasting methods directly applicable to Dutch data-center operators and AI infrastructure teams; aligns with EU energy-efficiency and transparent-AI priorities; introduces novel CP application in cloud operations with strong technical depth and reproducibility.
