Distributed High Performance Computing Environment of Ural Branch of RAS M.L.Goldshtein, A.V.Sozykin, Institute of Mathematics and Mechanics UrB RAS, Yekaterinburg. - презентация

1 Distributed High Performance Computing Environment of Ural Branch of RAS M.L.Goldshtein, A.V.Sozykin, Institute of Mathematics and Mechanics UrB RAS, Yekaterinburg G.F.Masich, A.G.Masich Institute of Сontinuous Мedia Мechanics UrB RAS, Perm Distributed Computing and Grid-technologies in Science and Education, Dubna, July, 2012

2 2 Introduction Distributed high performance computing environment of UrB RAS is the infrastructure for eScience (cyberinfrastructure) Users: 40 Research Institutes of UrB RAS in 7 regions (Yekaterinburg, Arkhangelsk, Syktyvkar, Orenburg, Perm, Izhevsk, Chelyabinsk) Universities Industrial enterprises Project participations: Institute of Mathematics and Mechanics UrB RAS (IMM UrB RAS) – computational recourses and information systems Institute of Сontinuous Мedia Мechanics UrB RAS (ICMM UrB RAS) – backbone networks Supported by grants of UrB RAS No 12-P , RCP Urb RAS No RCP-12-I21

3 3 Overview High performance computing environment of UrB RAS consists of 4 parts: Computational resources Storage Networks Cloud platform

4 4 Computational resources Main computational resources are installed at the Supercomputer center of IMM UrB RAS: Supercomputer URAN, peak performance 160 TFlops, 5 th position in the TOP50 Computational resources in the regional scientific centers of UrB RAS Connection using Grid technologies (Globus) and SLURM

5 5 Example tasks Calculation of launching of a carrier rocket (Sojuz-2 and Rus- M) into orbit (Avtomatika, Yekaterinburg) Remote sensing image processing (Uralgeoinform) Aero-engine computational fluid dynamics (Aviodvigatel, Perm)

6 6 Example tasks Молекулярно-динамическое моделирование и восстановление межчастичных потенциалов UO2–PuO2 для прогнозирования поведения ядерного топлива в процессах изготовления, эксплуатации и переработки Моделирования внутренней динамики Земли и других планет с целью изучения истории и перспектив их развития Создание математической модели сердца, отличающейся от существующих применением уникальной модели сократительного процесса мышечных волокон. Ее использование в вычислительном эксперименте позволит решить целый ряд проблем актуальных для современной кардиологии Синтез диаграмм направленности антенных решеток с двойным фазовым управлением для задач навигации автономных летательных аппаратов Разработка in silico новых противоопухолевых, противовоспалительных и антиаритмических лекарственных средств

7 7 Grid projects Grid of Russian Federation by Ministry of Communication One of the resource centers Network 1Gb/sec, Grid gateway Setup and testing till the end of the 2012 Distributed information and computational environment of Ural Federal District Institute of Mathematic and Mechanics UrB RAS South Ural State University, Chelyabinsk All-Russian Scientific Research Center of Technical Physics Federal Nuclear Center, Snezhinsk Ugra Research Institute of Information Technology

8 8 Storage Storage resources are also installed at the Supercomputer center of IMM UrB RAS Main storage EMC Celerra NS- 480, 150TB SATA, NAS (NFS, CIFS) Storage is used by: Supercomputer URAN Cloud platform of UrB RAS Experimental facilities

9 9 Backbone networks Giga UrB RAS project: High-speed, fiber-optic network infrastructure (1-100 Gbit/sec) 5 regional scientific centers of UrB RAS (Yekaterinburg, Perm, Izhevsk, Syktyvkar and Arkhangelsk) Dark" optical fiber and the DWDM technology Current status: Communication channel Perm-Yekaterinburg 1 Gbit/sec 456 km Plans for 2012: Increasing the speed of channel Perm-Yekaterinburg to 2x10 Gbit/sec Building the communication channel Izhevsk-Perm

10 10 Cloud platform IaaS for computational tasks Technologies: Linux, oVirt, Apache DeltaCloud Servers from old computational clusters Usage: HPC application, integrated with clusters Distributed applications (natural language processing, grant UrB RAS RCP-12-P10) Education in the Supercomputing technologies

11 11 Matlab in the cloud Running task on the supercomputer: job = imm_sch_f(8,20,'my_function');

12 12 Conclusion High performance computing environment of UrB RAS Computational resources, storage, backbone network, and cloud platform Available for research organization of UrB RAS for free