|
Abstract.
This article analyzes trends, challenges, and strategic directions in the development of high-performance computing infrastructure (HPCI) for scientific research, emphasizing the growing role of artificial intelligence as a driver of technological progress. It examines how AI and big data have accelerated the evolution of research HPCI, including applications in governmental, defense, and security domains. The objective is to identify global trends and strategic priorities in research HPCI and to assess Russia's approach to HPCI transformation. The study applies a comparative analysis of international programs in the United States, the European Union, and China; data from the TOP500 and Green500 rankings; and publications on AI-HPC convergence and energy efficiency. It analyzes quantitative indicators of performance, energy consumption, and architectural features of leading supercomputers, together with statistics on the Russian supercomputing fleet. The results show a shift toward hybrid CPU-accelerator architectures, large-scale GPU/TPU/NPU deployments, hierarchical memory, high-speed interconnects, containerization, and HPC-as-a-service models driven by AI and big data workloads. The United States prioritizes exaflop-class systems and scalable clusters for LLM, Europe emphasizes modular, energy-efficient systems and processor sovereignty, and China focuses on centralized programs, distributed data-center networks, and indigenous hardware. For Russia, the analysis highlights technological lag, dependence on foreign components, and the need to reorient infrastructure toward hybrid, AI-oriented architectures. Overall, integration of HPC, AI, and big data into a unified chain emerges as the key trend shaping national strategies. For Russia achieving targets in HPCI requires development of domestic processors, accelerators, software, and specialists training.
Keywords:
high-performance computing, distributed computing, artificial intelligence, computing infrastructure.
DOI 10.14357/20718632260201
EDN BZXPHZ
PP. 3-12.
References
1. Zatsarinnyi AA, Ivanov KV. Key problems of implementing artificial intelligence technologies in the interests of ensuring the military security of the state. Proceedings of the 10th International Interdepartmental Scientific and Practical Conference of Research Department No. 10 of the Russian Academy of Rocket and Artillery Sciences "80th Anniversary of the Great Victory: Historical experience and modern problems of Russia's military security". Moscow. 2025;43-52. (In Russ.). 2. Kasneci G, Gasser U, Hofmann TF, Kramer G, Muller G, Peus C, Schonenberger H, Kasneci E. Europe's AI Imperative - A Pragmatic Blueprint for Global Tech Leadership. arXiv [Preprint]. 2025. Available from: https://arxiv.org/abs/2502.08781 [Accessed 16 September 2025]. 3. Tuytlayeva EO, Odintsov IO, Moskovskiy AA, Marmuzov GV. Trends in the development of computing nodes of modern supercomputers. Bulletin of South Ural State University. Series: Computational Mathematics and Software Engineering. 2019;8(3):92-114. (In Russ.). doi:10.14529/cmse190305. 4. McKinsey & Company. AI power: expanding data center capacity to meet growing demand. 2024. Available from: https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/ai-power-expanding-data-center-capacity-to-meet-growing-demand [Accessed 23 July 2025]. 5. Huerta EA, Khan A, Davis E, et al. Convergence of artificial intelligence and high performance computing on NSF-supported cyberinfrastructure. Journal of Big Data. 2020;7:88. doi:10.1186/s40537-020-00361-2. 6. Micron. AI at the edge: future memory and storage in accelerating intelligence. Available from: https://my.micron.com/about/blog/storage/ai/ai-at-the-edge-future-memory-and-storage-in-accelerating-intelligence [Accessed 30 September 2025]. 7. NVIDIA. Efficient AI supercomputers with GPU accelerators. NVIDIA Blog. 2023. Available from: https://blogs.nvidia.com/blog/efficient-ai-supercomputers-sc23 [Accessed 02 October 2025]. 8. Rajbhandari S, Rasley J, Ruwase O, He Y. ZeRO: Memory optimizations toward training trillion parameter models. arXiv [Preprint]. 2020. Available from: https://arxiv.org/abs/1910.02054 [Accessed 30 September 2025]. 9. Pryakhina DI, Korenkov VV. Relevance of creating a digital twin for managing distributed centers for data collection, storage, and processing. Sovremennye informatsionnye tekhnologii i IT-obrazovanie. 2023;19(2):262-271. (In Russ.). doi:10.25559/SITITO.019.202302.262-271. 10. TOP500 supercomputer sites. 2025. Available from: https://www.top500.org/ [Accessed 01 October 2025]. 11. Green500 list. TOP500. 2025. Available from: https://www.top500.org/lists/green500/ [Accessed 01 October 2025]. 12. U.S. Department of Energy. Advanced scientific computing research budget request FY 2026. Available from: https://science.osti.gov/~/media/budget/pdf/sc-budget-request-to-congress/2026/FY-2026-Advanced-Scientific-Computing-Research-Budget-Request.pdf [Accessed 16 July 2025]. 13. Roaten M. Lab Powers Up to Plug In Next-Gen Supercomputers. National Defense. 2022 Jul 26. Available from: https://www.nationaldefensemagazine.org/articles/2022/7/26/lab-powers-up-to-plug-in-next-gen-supercomputers [Accessed 23 July 2025]. 14. Becciani U, Petta C. New frontiers in computing and data analysis - the European perspectives. Radiation Effects and Defects in Solids. 2019;174(11-12):1020-1030. doi:10.1080/10420150.2019.1683840. Available from: https://www.tandfonline.com/doi/full/10.1080/10420150.2019.1683840 [Accessed 31 July 2025]. 15. Banchelli F, Garcia-Gasulla M, Mantovani F, Vinyals J, Pocurull J, Vicente D, Eguzkitza B, Galeazzo FCC, Acosta MC, Girona S. Introducing MareNostrum5: A European pre-exascale energy-efficient system designed to serve a broad spectrum of scientific workloads. arXiv [Preprint]. 2025. Available from: https://arxiv.org/abs/2503.09917. [Accessed 06 August 2025]. 16. European Processor Initiative. Available from: https://www.european-processor-initiative.eu [Accessed 01 October 2025]. 17. Zhang N, Duan H, Guan Y, Mao R, Song G, Yang J, Shan Y. The "Eastern Data and Western Computing" initiative in China contributes to its net-zero target. Engineering. 2025;52:256-261. DOI: 10.1016/j.eng.2024.08.010. Available from: https://www.sciencedirect.com/science/article/pii/S2095809924005058 [Accessed 17 August 2025]. 18. Ezell S. A new frontier: sustaining U.S. high-performance computing leadership in an exascale era. Information Technology & Innovation Foundation. 2022 Sep 12;84 p. Available from: https://www2.itif.org/2022-hpc-leadership-exascale-era.pdf [Accessed 23 July 2025]. 19. Trader T. Three Chinese exascale systems detailed at SC21: two operational and one delayed. HPCwire. 2021. Available from: https://www.hpcwire.com/2021/11/24/three-chinese-exascale-systems-detailed-at-sc21-two-operational-and-one-delayed/ [Accessed 23 July 2025]. 20. President of the Russian Federation. Presidential address to the Federal Assembly, February 29, 2024. Kremlin. 2024. (In Russ.). Ava
|
