An Experimental Testbed and Infrastructure Monitoring Dataset Containing Anomalous Patterns

The dataset contains 18 workload profiles (15 profiles training and 3 for testing), gathered using containerized environment. The profiles contain introduced anomalies. The dataset can be used for model training for purposes of anomaly detection in microservice systems.

The Testbed provides means for generating synthetic time-series datasets for anomaly detection in microservice systems. Each dataset is produced by running controlled load tests (Locust) against a FastAPI + PostgreSQL stack, while the Orchestrator injects real anomalies (CPU spike, DB slowdown, IO stress, Network delay). The Collector records system metrics every second into CSV files.