A Statistical approach to Event Suppression in Cloud Environments
Mudit Verma, Harshit Kumar, Pratibha Moogi · 2025
There is an exponential growth in observability data—including metrics, logs, traces, and events in modern cloud-native systems. Among these, events represent a specific type of telemetry data that often requires immediate attention, as they signal discrete occurrences such as anomalies, threshold breaches, or critical system changes. While comprehensive observability is essential for effective monitoring and troubleshooting, the sheer volume of event traffic can overwhelm operators, drain resources, and drive up operational costs. This paper introduces a dynamic rate-limiting algorithm tailored to suppress duplicate events, addressing the dual challenge of maintaining comprehensive observability while reducing noise. By dynamically adjusting event forwarding rates based on frequency patterns, the algorithm ensures critical anomalies are represented while suppressing redundant event traffic. Leveraging statistical measures and adaptive reset mechanisms, this approach optimizes event pipelines for efficiency without compromising diagnostic utility. We also demonstrate the algorithm’s effectiveness, achieving an 8.6x reduction in forwarded events while retaining actionable insights, paving the way for scalable and cost-effective event management in cloud-native environments.