Intelligent Cluster Optimizer blocks autoscaling when memory leaks detected

A team of researchers (Azra Karakaya, Erva Şengül, Ahmet Kaplan) published a paper introducing the Intelligent Cluster Optimizer, an open-source Kubernetes operator designed to solve resource over-provisioning with safety as a first-class concern. The core contribution is a five-layer safety pipeline that combines SLA monitoring, a circuit breaker, HPA/PDB conflict detection, a policy engine, and a memory-leak detector based on linear regression with R² scoring. Unlike typical predictive autoscalers that only alert on anomalies, this system actively blocks scaling recommendations when a memory leak is detected, preventing the operator from masking software defects by granting more resources to a leaking workload.

The system uses percentile analysis and Holt-Winters forecasting to generate right-sizing recommendations, balanced through multi-objective Pareto optimization at the per-container level. It also includes rollback and dry-run modes for human approval. The researchers validated the operator with 1,118 automated tests (80.3% code coverage) and a live deployment on Google Kubernetes Engine. Their what-if projections estimated cost savings of 20–40% from right-sizing, while the leak gate achieved 83% detection accuracy. This approach turns anomaly detection from a passive alerting tool into an active safety gate, a novel contribution for Kubernetes resource optimization.