Distributed QoS anomaly detection with adaptive sampling: a middleware-integrated approach for cloud SLA compliance

Cloud computing necessitates robust quality of service (QoS) management to ensure adherence to service level agreements (SLAs), yet existing systems often lack efficient QoS violation detection mechanisms. To address this gap, this study proposes a novel QoS violation detection framework integrated into the CP-M&E middleware, focusing on scalability and communication efficiency in large-scale cloud environments. The framework introduces a distributed sliding-window algorithm to balance detection accuracy and overhead, coupled with a likelihood-based adaptive sampling technique that dynamically adjusts intervals to minimise monitoring costs. A quantitative communication-cost model further optimises parameter tuning for diverse operational scenarios. Experimental deployment in real-world cloud infrastructures demonstrated a 30%-40% reduction in communication overhead compared to traditional methods, alongside enhanced scalability for resource pools. Results validate the framework's superior performance in reducing false positives and maintaining SLA compliance, establishing its viability for modern cloud ecosystems.