Paper 13101-77
Advanced log analysis for operations at Paranal Observatory
On demand | Presented live 16 June 2024
Abstract
The VLT at Paranal Observatory has been in operation for over two decades, and soon, the ELT will be managed by the same operational team. Maintaining operational efficiency and minimizing downtime with limited resources will be crucial. Previous research has shown that software logs effectively capture the telescopes' behavior, providing valuable operational insights. We've integrated various log analysis techniques from academic literature and industry best practices. These techniques allow engineers to monitor system health, analyze error sequences, detect anomalies, and reconstruct processes which improve maintenance and extract new insights. Additionally, we've utilized generative artificial intelligence and NLP transformer-based models, to infer observation behavior and predict execution failures. We have taken advantage of both the Paranal Datalab on-premises facility and Azure Cloud. In this work, we provide technical details and outline the key challenges and opportunities in adopting this technique within an astronomy facility.
Presenter
European Southern Observatory (Chile)
Mr. Juan P. Gil, with an M.Sc. degree in Mathematical Modeling, currently serves as the Deputy Manager of the Software Group at Paranal Observatory. In this capacity, he leads software practices, contributes to strategic planning, and serves as the technical authority for the VLTI operational software. With a career spanning since 2014, Mr. Gil has been a notable contributor to automated log analysis. Previously, he worked as a software engineer at the ALMA Observatory, where his involvement in log analysis commenced. His professional background also includes roles at CONICYT and UFRO in Chile.