AIOps-based SLO-driven, cost-aware autoscaling framework for Kubernetes clusters

Vinoth Punniyamoorthy and co-authors (arXiv:2512.23415) present an AIOps-driven, SLO-first autoscaling framework for Kubernetes that integrates multi-signal control with lightweight demand forecasting to reduce SLO violations and infrastructure cost.

• Main announcement: The paper proposes a safe and explainable multi-signal autoscaling framework that combines SLO-aware and cost-conscious control with lightweight demand forecasting; experimental results report up to 31% reduction in SLO violation duration, 24% faster scaling response time, and 18% lower infrastructure cost versus default and tuned Kubernetes autoscaling baselines.
• Background and details: The work includes a gap-driven analysis of existing autoscaling approaches, evaluates using representative microservice and event-driven workloads, and emphasizes stable and auditable control behavior; the manuscript was submitted to arXiv (v1) on 29 Dec 2025, is available as PDF, HTML, and TeX source, and is licensed under CC BY 4.0.