Defense of PhD thesis on distributed computing took place at MLIT On 12 September, MLIT researcher Igor Pelevanyuk successfully defended his thesis for the degree of Candidate of Technical Sciences at the Meshcheryakov Laboratory of Information Technologies. The thesis focused on methods and software for creating distributed heterogeneous computing environments based on the DIRAC platform. The methods described in the thesis were elaborated, implemented, and successfully applied to processing data from the experiments at the NICA accelerator complex and several other experiments. MLIT Scientific Leader, Doctor of Technical Sciences Vladimir Korenkov was the scientific supervisor of the work. Head of the Department of Computing Technologies at the National Research University «MPEI», Doctor of Technical Sciences Victor Toporkov and Head of the Laboratory of Symbolic Calculations in High Energy Physics at SINP MSU, Candidate of Physical and Mathematical Sciences Alexander Kryukov were opponents. “Igor Pelevanyuk’s thesis is of great scientific, technological, and practical significance. He is actively involved in the development of the unique JINR Multifunctional Information and Computing Complex and has performed the integration of equipment of various architectures (grid, clouds, clusters, supercomputers, data storage systems), which is being effectively employed for a number of large-scale scientific projects,” the thesis’s supervisor Vladimir Korenkov pointed out. Today, the organization of geographically distributed heterogeneous computing environments is a necessary tool for solving the tasks of large-scale physics projects. With the continuous growth of volumes of data generated by modern scientific experiments and projects, their demands often exceed the capabilities of any available computing resources. The solution often lies in the shared use of multiple computing clusters, storage systems, supercomputers, and cloud infrastructures. This entails the organization of a unified computing environment capable of integrating available resources and using them for data storage and processing. Such resource pooling remains a challenge that requires a comprehensive approach. In his thesis, Igor Pelevanyuk set the objective of developing methods and software tools to organize distributed heterogeneous computing environments for data processing based on the DIRAC platform, which was originally created to solve tasks for the LHCb experiment at the Large Hadron Collider at СERN. Since 2017, Igor Pelevanyuk has been successfully developing such a DIRAC-based computing environment at JINR. It embraces the “Govorun” supercomputer, the Tier1 and Tier2 clusters, the NICA and BM@N DAQ Computing Center clusters, cloud infrastructures of JINR and the JINR Member States, the computing resources of several collaborations of the experiments at the NICA accelerator complex. EOS disk storage, Enstore tape storage, and CTA tape storage were integrated as storage systems. Initially, the DIRAC platform toolkit did not comprise task monitoring and data transfer tools, tools for analyzing the execution of large batches of tasks, and there was no possibility of the direct integration of cloud resources running on the OpenNebula platform. Igor Pelevanyuk developed new software tools for integrating cloud resources that support the OCCI protocol, as well as for directly integrating clouds using OpenNebula. Both modules were included in the DIRAC platform source code. He developed and implemented user task monitoring and data transfer systems that enable the analysis of the execution of individual user tasks. He proposed an original method for analyzing the execution of large batches of tasks in heterogeneous computing environments, which allows analyzing both task batches and the computing resources on which these tasks were executed. Tools for simplifying the mass launch of tasks and data transfer were also developed and implemented. All developed systems operate independently, allowing the proposed approaches to be applied within other computing infrastructures organized using the DIRAC platform. The developed method for analyzing the execution of large batches of tasks revealed a performance calculation error when working with AMD Epic processors, which was subsequently corrected by DIRAC developers. In addition to using the developed systems separately, their combined usage enables to detect anomalies during task execution, to analyze their causes, and predict the behavior of a distributed infrastructure when executing a large number of tasks. Some of the solutions developed by Igor Pelevanyuk were applied to the BES-III experiments and the international neutrino experiment JUNO at the Institute of High Energy Physics of the Chinese Academy of Sciences (Beijing). The DIRAC-based computing environment developed by Igor Pelevanyuk at JINR has found wide application in the flagship NICA project. Being responsible for distributed mass data processing in the ВМ@М experiment, in 2023, Igor Pelevanyuk developed a methodology for the mass processing of experimental data using available geographically distributed, heterogeneous resources and subsequently successfully applied it to process data from the 8th run of the BM@N experiment. Furthermore, he developed and implemented a method for data transfer between resources that support standard grid data transfer protocols and those that do not. Thanks to the developed and implemented software tools, a distributed, heterogeneous computing environment was organized, allowing the BM@N collaboration to process approximately 400 TB of experimental data from the 8th run in one week, effectively selecting computing resources for various types of processing tasks. Since 2019, the MPD experiment at the NICA collider has been actively employing the DIRAC-based distributed heterogeneous environment organized by Igor Pelevanyuk. Thanks to the software he developed, the processing of data required for MPD detector design, reconstruction algorithm testing, and physics analysis was significantly accelerated. The DIRAC platform at JINR was also successfully used in the initial stages of preparation for the SPD experiment at NICA. Applying methods developed by Igor Pelevanyuk, SPD collaboration members performed data modeling and analysis using the Tier1 and Tier2 clusters. In addition, the computing environment organized during the work of Igor Pelevanyuk was employed to solve data processing tasks within the Baikal-GVD experiment and the Folding@Home project. “He is always searching, always full of ideas, which he successfully implements. His distinctive feature is to see things through to tangible results,” Vladimir Korenkov commented on Igor Pelevanyuk’s work. Igor Pelevanyuk started his scientific career at MLIT JINR in 2013 after graduating with a master’s degree from Dubna University. He had already been actively involved with the Laboratory during his student years. In his early years at MLIT, Igor Pelevanyuk monitored grid services for the MLIT computing complex and the DIRAC system of the BES-III experiment in China. Four years later, work on organizing a computing environment based on the DIRAC platform at JINR began. At present, Igor Pelevanyuk, a researcher in the Distributed Systems Sector, continues to effectively develop the distributed environment at JINR and its Member States. Igor Pelevanyuk successfully combines his research activity with teaching at Dubna State University, where he lectures and conducts practical classes for students. He is also actively involved in JINR’s science popularization efforts.
