Design and Implementation of the Fast Data Processing System for the Baikal-GVD Neutrino Telescope
V. A. Allakhverdyan, A. D. Avrorin, A. V. Avrorin, V. M. Aynutdinov, I. A. Belolaptikov, Z. Berusova, E. A. Bondarev, I. V. Borina, N. M. Budnev, V. A. Chadymov, A. S. Chepurnov, V. Y. Dik, A. N. Dmitriyeva, G. V. Domogatsky, A. A. Doroshenko, R. Dvornicky, A. N. Dyachok, Zh. -A. M. Dzhilkibaev, E. Eckerova, T. V. Elzhov, V. N. Fomin, A. R. Gafarov, K. V. Golubkov, T. I. Gress, K. G. Kebkal, V. K. Kebkal, I. Kharuk, S. S. Khokhlov, E. V. Khramov, M. M. Kolbin, S. O. Koligaev, K. V. Konischev, A. V. Korobchenko, A. P. Koshechkin, V. A. Kozhin, M. V. Kruglov, V. F. Kulepov, A. A. Kulikov, Y. E. Lemeshev, M. V. Lisitsin, S. V. Lovtsov, R. R. Mirgazov, D. V. Naumov, A. S. Nikolaev, I. A. Perevalova, A. A. Petrukhin, D. P. Petukhov, E. N. Pliskovsky, M. I. Rozanov, E. V. Ryabov, G. B. Safronov, B. A. Shaybonov, V. Y. Shishkin, E. V. Shirokov, F. Simkovic, A. E. Sirenko, A. V. Skurikhin, A. G. Solovjev, M. N. Sorokovikov, I. Stekl, A. P. Stromakov, O. V. Suvorova, V. A. Tabolenko, V. I. Tretyak, B. B. Uizutuev, Y. V. Yablokova, D. N. Zaborov, S. I. Zavyalov, D. Y. Zvezdov
TL;DR
This work presents a fast data processing system for the Baikal-GVD neutrino telescope designed to rapidly identify astrophysical neutrino events. The system leverages the detector's modular cluster architecture to run parallelized file processing on dedicated virtual machines, implementing a fast per-file pipeline and a full per-run pipeline that incorporates dynamic detector geometry and data quality monitoring, achieving end-to-end latencies around $1.5$–$18$ minutes for fast results and about $27$ hours for complete processing. It integrates the BARS framework for standard Baikal-GVD analyses, Luigi-based workflow orchestration, CephFS/EOS storage, and near-real-time alerting to enable timely multi-messenger follow-ups. While delivering substantial speedups over offline pipelines, the paper also notes current offline-only multi-cluster event building and delayed acoustic data integration as areas for future upgrades to enhance accuracy and timeliness in time-domain astronomy.
Abstract
We present a fast data processing system for the Baikal-GVD neutrino telescope, designed for rapid identification of astrophysical neutrino events. Leveraging Baikal-GVD's modular cluster architecture, the system implements parallelized file processing where raw data files undergo concurrent analysis across dedicated virtual machines. The system implements two pipelines: a fast per-file processing and a fully fledged (per-run) processing, which integrates dynamic detector geometry determined from acoustic and inertial positioning systems and data quality monitoring with a latency of about 27 hr. The fast processing pipeline delivers a total latency of about 1.5-18 minutes from event detection to reconstructed data availability, depending on water luminescence levels. This enables fast follow-up observations of transient astrophysical sources, fulfilling Baikal-GVD's role in multi-messenger networks. The article also highlights key features of the data acquisition system, the integration of advanced synchronization systems and a robust infrastructure for data handling and storage, ensuring efficient and reliable operation of the Baikal-GVD telescope.