Astrophysics > Instrumentation and Methods for Astrophysics
arXiv:2510.27146 (astro-ph)
Abstract:Objective: ServiMon is designed to offer a scalable and intelligent pipeline for data collection and auditing to monitor distributed astronomical systems such as the ASTRI Mini-Array. The system enhances quality control, predictive maintenance, and real-time anomaly detection for telescope operations. Methods: ServiMon integrates cloud-native technologies-including Prometheus, Grafana, Cassandra, Kafka, and InfluxDB-for telemetry collection and processing. It employs machine learning algorithms, notably Isolation Forest, to detect anomalies in Cassandra performance metric…
Astrophysics > Instrumentation and Methods for Astrophysics
arXiv:2510.27146 (astro-ph)
Abstract:Objective: ServiMon is designed to offer a scalable and intelligent pipeline for data collection and auditing to monitor distributed astronomical systems such as the ASTRI Mini-Array. The system enhances quality control, predictive maintenance, and real-time anomaly detection for telescope operations. Methods: ServiMon integrates cloud-native technologies-including Prometheus, Grafana, Cassandra, Kafka, and InfluxDB-for telemetry collection and processing. It employs machine learning algorithms, notably Isolation Forest, to detect anomalies in Cassandra performance metrics. Key indicators such as read/write latency, throughput, and memory usage are continuously monitored, stored as time-series data, and preprocessed for feature engineering. Anomalies detected by the model are logged in InfluxDB v2 and accessed via Flux for real-time monitoring and visualization. Results: AI-based anomaly detection increases system resilience by identifying performance degradation at an early stage, minimizing downtime, and optimizing telescope operations. Additionally, ServiMon supports astrostatistical analysis by correlating telemetry with observational data, thus enhancing scientific data quality. AI-generated alerts also improve real-time monitoring, enabling proactive system management. Conclusion: ServiMon’s scalable framework proves effective for predictive maintenance and real-time monitoring of astronomical infrastructures. By leveraging cloud and edge computing, it is adaptable to future large-scale experiments, optimizing both performance and cost. The combination of machine learning and big data analytics makes ServiMon a robust and flexible solution for modern and next-generation observational astronomy.
| Comments: | Accepted for publication in IAU Symposium 397: Exploring the Universe with Artificial Intelligence (UniversAI 2025), Cambridge University Press. Editors: C. Sterken, J. Hearnshaw & D. Valls-Gabaud |
| Subjects: | Instrumentation and Methods for Astrophysics (astro-ph.IM); Machine Learning (cs.LG) |
| Cite as: | arXiv:2510.27146 [astro-ph.IM] |
| (or arXiv:2510.27146v1 [astro-ph.IM] for this version) | |
| https://doi.org/10.48550/arXiv.2510.27146 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Emilio Mastriani Ph.D [view email] [v1] Fri, 31 Oct 2025 03:47:13 UTC (1,035 KB)