MLIT expands its participation in development of JUNO experiment distributed computing infrastructure In January, the 25th meeting of the international collaboration of the JUNO (Jiangmen Underground Neutrino Observatory) experiment took place in Kaiping (Guangdong Province, China). JINR has been actively involved in this project since its foundation. Nikita Balashov, a first-category software engineer of the Meshcheryakov Laboratory of Information Technologies, participated in the meeting. The main goal of the JUNO experiment is to determine the neutrino mass hierarchy. Thanks to the huge volume of the scintillation detector and the high accuracy of energy measurements, the project opens up wide opportunities for fundamental research: from the precision determination of mixing parameters in the lepton sector of the Standard Model and the detection of geoneutrinos and solar neutrinos to the observation of neutrinos from supernovae and the search for new physics, including proton decay. JINR specialists are involved in various tasks in the JUNO project, from detector assembly to data analysis. The leading Laboratory for the work within the project is DLNP. Since 2017, MLIT has joined the project and has currently implemented the connection of the JINR computing cluster to the JUNO common distributed computing infrastructure on the basis of the Multifunctional Information and Computing Complex (MICC), as well as has organized data storage systems on disks and tapes. At the moment, the expansion of the Laboratory’s participation in the project is being discussed. At the January meeting in Kaiping, the key topic of discussion was the preparation for the experiment’s start, including a six-month procedure of filling the central detector with liquid scintillator, the commissioning of photomultipliers, the analysis of equipment response and the first data. Filling the central detector with liquid scintillator will begin in February 2025, and it is planned to finish by the next collaboration meeting in July. A preliminary plan for processing the data acquired during this period was presented as well. In December, there was made a decision to form a team at MLIT that would work on data storage systems for the neutrino experiments, the first of which would be the JUNO project. The new storage for JUNO will be implemented on top of the dCache system by MLIT specialists, namely, Roman Semenov, Alexander Baranov and Alexander Makhalkin. Nikita Balashov became the coordinator of this work. Valery Mitsyn and Vladimir Trofimov, who have impressive experience in operating various data storage systems, are acting as consultants. As Nikita Balashov explained, data acquisition is performed directly in the underground observatory, and then the data is transferred to the China Institute of High Energy Physics (IHEP) in Beijing. Long-term data storage is carried out with triple redundancy, namely, three full copies of the data are stored in IHEP, in the CNAF data center at the National Institute of Nuclear Physics (INFN, Italy) and at MLIT JINR on the MICC resources. Partial data storage is also provided by the National Institute of Nuclear and Particle Physics (IN2P3 or CNRS, France). Currently, fast data exchange with IHEP is organized via European networks, however, for efficient operation, it is required to organize a dedicated communication channel between IHEP and JINR. Despite the detector’s data collection rate of 40.7 GB/s, due to channel and storage limitations, the data is filtered and compressed for transfer at 60 MB/s, which is about 2 PB per year. “Since a fiber-optic channel to the border with China exists, then with a positive decision by the Chinese government, such a channel can be organized. And the collaboration is making efforts to assist in resolving this issue,” Nikita Balashov noted. To process data, a distributed computing system DCI (Data Center Infrastructure), which is essentially a grid infrastructure, was built within the JUNO project. “Event modeling has been underway on DCI for several years now, and event reconstruction will start soon,” Nikita Balashov added. Following the results of the JUNO collaboration meeting, the possibility of implementing two tasks by MLIT is being considered. The first of them is the transfer of the project’s data processing system based on DIRAC to an updated version of this distributed computing platform, DiracX. “The version of the DiracX platform was written anew, from scratch, and contains a new deployment model using advanced containerization technologies for the software environment,” Nikita Balashov clarified. MLIT specialists also plan to undertake the adaptation and development of the JSAB software product used by the JUNO experiment. In essence, JSAB is a user-friendly interface that simplifies physicists’ access to the distributed computing infrastructure for experimental data analysis. In the future, this experience can be applied to the computing needs of other JINR experiments, including at the NICA accelerator complex. In addition, since December last year, Nikita Balashov’s team has been deploying a new data storage for JUNO at the MICC. “We are currently deploying this storage in a configuration similar to Tier1 for the CMS experiment at the Large Hadron Collider at CERN. However, we plan to study other options to achieve architectural enhancements of the system,” he pointed out. Right now, JINR, represented by the Meshcheryakov Laboratory of Information Technologies, is one of the key suppliers of computing resources for the JUNO collaboration, along with IHEP and INFN-CNAF. For the needs of the experiment, the Laboratory continuously provides at least 2,000 computing cores, 1 petabyte for data storage on disks and 5 petabytes on tapes.
