Adaptive Query Optimization in Kubernetes Orchestrated Open Source Relational Databases Using Telemetry Driven Feedback Loops

Abstract

Relational query optimization has traditionally relied on static cost models built upon cardinality estimation and predefined access path selection strategies. While adaptive query processing techniques have improved robustness against data skew and runtime uncertainty, most open-source relational database systems remain largely unaware of infrastructure-level variability introduced by cloud-native orchestration platforms. In Kubernetes-managed environments, elastic scaling, pod migration, and multi-tenant resource contention frequently invalidate optimizer assumptions, resulting in suboptimal execution plans and performance instability.
This study proposes a telemetry-driven adaptive query optimization framework for Kubernetes-orchestrated open-source relational databases. The proposed architecture integrates fine-grained query execution metrics, cluster-level resource telemetry, and feedback control mechanisms to continuously refine cost models and selectively trigger re-optimization during runtime. By establishing a cross-layer communication channel between the database engine and the orchestration layer, the system dynamically responds to workload fluctuations and resource elasticity.
Experimental evaluation under variable workload intensities and scaling scenarios demonstrates significant reductions in query latency variance, improved cardinality estimation accuracy, and enhanced cluster resource utilization efficiency. The results indicate that telemetry-guided feedback loops provide a practical pathway toward autonomic, cloud-aware relational database systems. This work advances the integration of adaptive query processing with containerized infrastructure management and contributes toward the realization of fully self-optimizing, Kubernetes-native database architectures.