Supercomputing Education

Moscow State University

1 Contents

Supercomputing Consortium of Russian Universities 3 “Supercomputing Education” Project (2010-2012) 4 Series of Books “Supercomputing Education” 7 Summer Supercomputing Academy 8 Department of Supercomputers and Quantum Informatics 9 Faculty of Computational Mathematics and Cybernetics: Supercomputing Educational Programmes 10 Trainings on Applications Development and Optimization 13 SIGMA: Collective Bank of Tests on Parallel Computing 14 Excursions to the Supercomputing Center 15

2 Supercomputing Consortium of Russian Universities hpc-russia.ru

Number of Consortium organizations in cities

> 5 organizations 2 - 5 organizations 1 organization

Arkhangelsk

Petrozavodsk St. Petersburg Kirov Kostroma Perm Tver Izhevsk Nizhny Novgorod Ekaterinburg Tomsk Moscow Vladimir Kazan Krasnoyarsk Kaliningrad Saransk Chelyabinsk Omsk Novosibirsk Minsk Penza Ufa Tambov Samara Irkutsk Belgorod Voronezh Saratov

Chisinau Rostov-on-Don

57 organizations Vladivostok

The agreement about the formation of the Technologies. On the Consortium’s initiative, Supercomputing Consortium of Russian Uni­ a number of scientific conferences, All­Russia versities was signed by Rectors of Lomonosov student schools, and workshops were held, Moscow State University, Lobachevsky Nizhny books published, and analytical reports on Su­ Novgorod State University, Tomsk State Uni­ percomputers issued. versity and South Ural State University at the The Supercomputing Consortium of Russian session of the Russian Union of Rectors on the Universities initiated and organized a success­ 23rd of December, 2008. The main objective of ful implementation of the Russian Federation the Consortium is to use the School of Higher Presidential Commission Project on the mod­ Education’s powerful potential to develop and ernization and technological development of implement Supercomputing Technologies in Russian economy “Supercomputing Education”. Russian education, science and industry. Currently, the Consortium comprises 57 full and associate members, including leading Rus­ sian universities, the institutes of the Russian Academy of Sciences and commercial compa­ nies, actively working in the Supercomputing

3 “Supercomputing Education” Project (2010-2012) hpc-education.ru

Project leader – Rector • Tomsk State University, of Lomonosov Moscow State • South Ural State University, University academician • St.Petersburg State University of IT, Me­ V. A. Sadovnichy chanics and Optics, • Southern Federal University, • Far Eastern Federal University, • Moscow Institute of Physics and Technology, • Members of the Supercomputing Consor­ tium of Russian Universities. The Russian Federation Presidential Com­ mission on the modernization and technologi­ System of Research and cal development of the Russian economy Project Education Centers for “Formation of Highly Qualified Supercomput­ Supercomputing Technologies ing Technology and Software Specialist Train­ ing System” is complete. (REC SCT)

The strategic project objective is the formation The system of REC SCT forms a structure in­ of a nationwide highly qualified Supercomput­ corporating a group of REC in different Russian ing Technology and Software specialist training Federal Districts. REC are formed on the basis system. of structural subdivisions of Schools of Higher The main executives of the project are: Education – the members of the Supercomput­ ing Consortium of Russian Universities, which • Lomonosov Moscow State University, possess considerable expertise in carrying out • Lobachevsky Nizhny Novgorod State Uni­ research work and conducting the SCT educa­ versity, tional activity.

REC SCT in Federal Districts of Russia

4 Body of Knowledge in Academic-methodological Associations of clas­ sical universities. Supercomputing Technologies

The development of the Body of Knowledge in Proposals on Expanding Federal Supercomputing Technologies occupies the cen­ State Educational Standards tral part in the educational and methodological maintenance for the SCT specialist training, Based on the developed Body of Knowledge in skill development system, and advanced retrain­ the field of supercomputing technologies pro­ ing. posals on expanding the Russian third genera­ These are the competencies every student tion Federal State Educational Standards have must have upon graduating from a correspond­ been formed on the following specialties for in- ing faculty or a course, completing a retraining depth study of supercomputing technologies: program, or training in special groups. It is the • “Fundamental Informatics and Information Body of Knowledge that shows how an educa­ Technologies”, tional process must be organized and the par­ ticular subjects upon which it must be based. • “Applied Mathematics and Informatics”, The main aspect of the Body of Knowledge is • “Mathematics”. the description of the subject area “Supercom­ puters and Parallel Computing” to clearly define Proposals were developed on forming the cur­ what to train and how to organize the process of ricula and skill development and advanced re­ training for each target-oriented group of train­ training programs and in the supercomputing ees. technologies for enlarged groups of specialties The structure of the created Body of Knowl­ and training directions: edge is consistent with the recommendations of • “Physics and Mathematics”, the international professional communities of • “Informatics and Computational Tech­ the ACM and IEEE Computer Society. niques”, Supercomputing Technology • “Information Security”. Training Courses

Extensive work was performed on the devel­ opment of the new courses and extensions to existing ones: the materials of 50 courses, cov­ ering all main parts of the Body of Knowl­edge in the area “Supercomputing Technologies” are publicly available on the project site. All train­ ing courses underwent the academic-meth­ odological commission expertise and received the approving resolution of the Academic-meth­ odological Council on Applied mathematics, Informatics and Information technologies of

5 Entry-Level Training of Supercomputing Technologies

Implementation of entry-level specialists training on supercomputing technologies: • 3269 people passed the training, • the training program consisted of not less than 72 study hours, • 60 universities from 54 cities of all Russian Federal Districts were involved.

Faculty Staff Retraining Programs

During the project fulfillment the programs of skill development and advanced retraining of Target-Oriented Intensive faculty staff have been successfully carried out. Training in Special Education • the training programs included 72 study Groups hours, A very complex part of the project was ful­ • 453 people from faculty staff passed the re­ filled: training, • the programs included 72 study hours, • the retraining consisted of 10 programs, • 40 special groups of trainees were formed, • it embraced 50 organizations from 29 Rus­ sian cities and towns, • 790 trainees successfully passed the ad­ vanced retraining, • all Russian Federal Districts were presented. • more than 15 training courses, • all Federal Districts of REC SCT system were presented.

Internet University on Supercomputing Technologies

Within the project “Supercomputing Educa­ tion”, in 2010-2012, 731 people from over 100 Russian cities completed the entry-level train­ ing program via distant online education. hpcu.ru

6 Series of Books “Supercomputing Education”

A complex program on the development and expert examination of educational and methodolog­ ical literature in Supercomputing Technologies for the Bachelor`s and Master`s Degrees Programs is successfully being implemented. The series “Supercomputing Education”, created under the project, includes monographs, textbooks and training manuals on the main sections of the Body of Knowl­ edge in Supercomputing Technologies written by the leading Russian and foreign specialists.

“Computational Mathematics “Practical Course on Parallel “Supercomputing Modeling in and Algorithm`s Structure” Computing Techniques” Climate System Physics” V.V. Voevodin A.V. Starchenko, E.A. Danilkin, V.N. Lykosov, A.V. Glazunov, ISBN 978-5-211-05933-7 V.I. Laeva, S.A. Prokhanov D.V. Kulyamin, E.V. Mortikov, ISBN 978-5-211-05976-4 V.M. Stepanenko ISBN 978-5-21106341-9

“High-Performance Computa- “Computationally Complex “MPI and OpenMP Parallel tions for Multiprocessor Tasks of Number Theory” Programming Technologies” Multi-Core Systems” E.A. Grechnikov, S.V. Michailov, A.S. Antonov V.P. Gergel Y.V. Nesterenko, I.A. Popovyan ISBN 978-5-211-06343-3 ISBN 978-5-211-05937-5 ISBN 978-5-211-06342-6

“Parallel Programming Tech- “New Computational Fluid “Principles of Parallel Pro- nologies for New Architecture Dynamics Algorithms for gramming” Processors” Multiprocessor Computers” C.Lin, L.Snyder A.V. Linev, D.K. Bogolepov, V.M. Goloviznin, M.A. Zaitsev, ISBN 978-5-211-06427-0 S.I. Bastrakov S.A.Karabasov, I.A. Korotkin ISBN 978-5-211-05962-7 ISBN 978-5-211-06344-0

And others...

“Parallel Programming Tools in “Parallel Computing on GPU: From 10 to 50 copies of the Shared-Memory Systems” Architecture and Program K.V. Kornyakov, V.D. Kustikova, Models” books in the series “Supercom­ I.B. Meyerov, ISBN 978-5-211 06340-2 A.A. Sydnev,A.V. Sysoev, puting Education” are delivered A.V. Shishkov ISBN 978-5-211-05931-3 to 43 Russian universities free of charge. In total, 31500 books of this series are conveyed to Russian universities during the project.

7 Summer Supercomputing Academy academy.hpc-russia.ru/en

Lomonosov Moscow State University, Rus­sian seminar “Extreme Scale Scientific Computing”, Supercomputing University Consortium, Com­ which will be addressed leaders of the interna­ putational Mathematics and Cybernetics faculty tional supercomputing community: J.Dongarra and Moscow State University Research Comput­ (USA), P.Backman (USA), T.Sterling (USA), B. ing Center carry out a special annual event – Gropp (USA), S.Matsuoka (Japan), M.Valero an International Summer Supercomputing (Spain), M.Parsons (UK), A.Bode (Germany), Academy. The Academy gives its participants B.Mohr (Germany) and others. an opportunity to master a wide range of spe­ During the Academy a great number of adja­ cializations in supercomputing technologies and cent educational arrangements are organized: parallel computing including training on Mos­ trainings and master-classes carried by Intel, cow State University supercomputers “Lomono­ Mellanox, NVIDIA, RogueWave companies, sov”, “Chebyshev” and IBM BlueGene/P. IBM Technologies Day, excursions to Lomono­ Academy curriculum consists of lectures on sov Moscow State University supercomputing topical issues of supercomputer technology and complex, a series of scientific-popular lectures high-performance computing, advanced train­ and many other events. ing courses (tracks), master classes and work­ The Academy teachers are leading foreign and shops on specific topics, use of supercomputer Russian scientists: T. Sterling (Indiana Un.), D. technologies and the development of parallel Kirk (NVIDIA), M. Kagan (Mellanox), academi­ algorithms and software for high-performance cians B.N. Chetverushkin, I.A. Sokolov, corre­ computing systems. The focus of the curricu­ spondent-members of Russian Academy of Sci­ lum of the Academy is given to problems of con­ ences V.V. Voevodin, L.N. Korolev, G.G. Ryabov, structing extreme scale applications for solving E.E. Tyrtyshnikov, professors V.A. Kryukov, urgent computationally complex tasks. A.N. Tomilin, and other well-known specialists In 2014, all students of the Academy will be in high performance computing and supercom­ participants of unique international scientific puting technologies area.

8 Department of Supercomputers and Quantum Informatics E-mail: [email protected] Hewlett-Packard, T-Platforms, TESIS and many Website: qi.cmc.msu.ru others. Phone: +7 (495) 939-34-04 Regular courses: Chair of the Department: • “Supercomputers and parallel data process­ Vladimir V. Voevodin, Corre­ ing” by Prof. Vl.Voevodin. sponding Member of Russian Academy of Sci­ • “Quantum computations” by Prof. Y.Ozhigov. ences, Professor, Dr.Sc. • “Physical basis of quantum informatics” by Staff members: Prof. S.Molotkov. • Popova N., Associate Professor, PhD, • “Parallel programming for high perfor­ Deputy Chair, mance computers” by Assoc. Prof. N.Popova • Ozhigov Y., Professor, Dr.Sc., Scientific and Dr. V.Bakhtin. Secretary of the Department, • “Neural networks” by Dr. Burjak. • Molotkov S., Corresponding Member of • “Parallel models of natural computing” by the Academy of Cryptography of Russia, Dr. Ershov and Assoc. Prof. N.Popova. Professor, Dr.Sc., Department offers two Master Programs: • Yakobovsky M., Professor, Dr.Sc., • Supercomputer systems and applications • Kronberg D. , Assistant Professor, PhD, • Quantum Informatics • Cherniavsky A., Assistant Professor, PhD, In summer, the department organizes two- • Nikolsky I., Assistant Professor, PhD. weeks Summer Supercomputing Academy which Area of scientific interests of the Department is the largest supercomputing educational event includes supercomputing technologies, theory in Russia. and practice of high performance computing, highly scalable computing technologies, exascale systems, software, tools and methods, quantum informatics, quantum cryptography. An important activity of the Department is focused on developing educational programs in the field of supercomputing and parallel com­ puting technologies. The Department’s team is actively involved at the national Supercomputing Educational Proj­ ect working closely with leading Russian uni­ versities, institutes of Russian Academy of Sci­ ences and IT-companies: Intel, IBM, NVIDIA,

9 Faculty of Computational Mathematics and Cybernetics: Supercomputing Educational Programmes The faculty of Computational Mathemat­ Vacancies: 10 - for 2014-15 regular academic year. ics and Cybernetics of the Lomonosov Moscow http://fao.cs.msu.su/content/departments State University (CMC MSU) will be responsible PhD internship (3-12 months) at for this area from Lomonosov Moscow State CMC MSU departments after prior ap­ University in the sphere of teaching Undergrad­ proval of subjects of scientific research. uate, Graduate, Master, and PhD students in the Vacancies: 10 - for 2014-15 regular academic year. framework of the following types of activity: http://fao.cs.msu.su/content/departments 1. Short-time Sandwich Programmes 4. Joint scientific research and giving visiting 1.1. High Performance Computer Sim­ lectures by CMC MSU Professors. ulation for Science and Engineering (in English, 170 hours for 1-2 months), Un­ CMC MSU fees: dergraduate or Graduate level (by choice); Tuition fee for 2013-14 regular academic year Vacancies: 15 - for fall semester 2014, 15 - for is 320 000 RUB for Bachelor and Masters’ stu­ spring semester 2015. dents and 261 600 RUB for PhD students annual 1.2. Nonlinear Optimization and Op­ (10 months); fees for 2014-15 regular academic erations Research in Economics (in Eng­ year will be announced by June, 2014. Tuition lish, 126 hours for 1-2 months), Under­ fee for short time sandwich programmes is es­ graduate or Graduate level (by choice); tablished separately for each case and specified Vacancies: 15 - for fall semester 2014, 15 - for in the programme description. Tuition fee for spring semester 2015. internships varies from 16 000 RUB to 24 000 RUB per month depending on hours taken. 1.3. Information Security Basics (in Russian, 200 hours for 2-3 months) Un­ Responsible person from CMC MSU: Alex­ dergraduate or Graduate level (by choice). ander Razgulin, Professor, Vice Dean of CMC Vacancies: 15 - for fall semester 2014, 15 - for MSU Tel / fax: +7 (495) 932-88-51 spring semester 2015. e-mail: [email protected] 2. Preparatory internship at MSU (10 http://www.cs.msu.ru/ (Russian) months) followed by full-time two-year Mas­ http://fao.cs.msu.su/ (English) ter’s Programmes in Russian or integrated six-year Master’s Programme in Russian. Vacancies: 20 - for 2014-15 regular academic year. http://master.cmc.msu.ru/?q=en/node/1596 3. Preparatory internship at MSU (10 months) followed by full-time PhD at CMC MSU departments after prior ap­ proval of subjects of scientific research.

10 Student Training Programme “High Performance Computer Simulation for Science and Engineering”

Director: Professor Vladimir Voevodin. Total Length: 170 hours. Period: 2-3 months starting from September 15, 2014. Language: English. Courses included: 1. Parallel Computing and Supercomputing Technologies (24 hours) 2. Applied Computing for Clusters and Su­ percomputers (48 hours) 3. New High Performance Scalable Algo­ rithms for Computational Fluid Dynamics (48 hours) 4. Geometric Modeling in Computer Aided Design (CAD) (50 hours) Advantages and specific features:

• Unique combination of novel high perfor- mance technologies and algorithms for en- gineering; • Lecturers are the founders of Supercomput- er Education in Russia.

Cost (in roubles)

Number of students 5 10 15

Fee for 1 student 353 000 196 000 144 000

CMC MSU authorized officer: Alexander Razgulin, Professor, Vice Dean of CMC MSU for international relations, e-mail: [email protected]

11 Title: Parallel Computing and Supercomputing Technologies Introduction: Computational technologies are everywhere: in science, in commerce, in industry, so to keep competitiveness means to be fluent in computational technologies. Even the politicians say: “To Out-Compete means Out-Compute”. This is not a fundamental science anymore, this is true for numerous areas of the real-world science, commerce, industry. But for the last decade computational technologies have transformed from sequential to essentially parallel ones and these changes will never go back. This is the second computer revolution. In the 50-s and 60-s we learnt programming technologies for serial computers but now we must learn parallel programming technologies, parallel algorithms, and parallel architectures. In the course we will discuss basic principles of parallel computing as well as deep relationships between algorithms, parallel programming technologies, and architectures. Level: Undergraduate or Graduate (by choice) Length: 24 lecture hours Period: 2 months starting from September 15, 2014 Language: English Prerequisites: Bachelor in IT or CS: knowledge of computer architectures, basic programming technologies, algorithms, graph theory, numerical methods. For Whom: The course will be interesting and useful for students who want to feel themselves comfortable and to be in high demand in the totally parallel computer world of the nearest future. Modules: The course will focus on the following components: • Introduction in parallel computing and supercomputing technologies; • Architectures of parallel computers; • Efficiency, performance, and benchmarking of parallel computers and supercomputers; • Parallel programming technologies; • Introduction in the theory of analysis of the informational structure of algorithms and programs.

Learning methods: Theory and practice, practice and experience. Combination of fundamental foundations of computer science with deep understanding of the latest real-world achievements in parallel computing and supercomputing technologies is the basic idea of the course. Well known companies like Cray, IBM, Hewlett-Packard, Intel, NVIDIA, AMD, and etc, why they are still in business and why they are in supercomputing? Well known companies like Boeing, Procter & Gamble, DreamWorks, Ford, BP, Shell, and etc, why they exploit supercomputing technologies for their businesses? The major principle of the course is to understand the basic idea of parallel computing and to use it efficiently in any other area later. Learning goals: After completing the course, student can demonstrate his/her ability to: Be fluent in the concept of parallel processing; Estimate efficiency of a particular parallel computing platform; Make a decision about a parallel computing platform which is necessary for solving a particular problem; Understand basic principles of modern parallel programming technologies; Be able to reveal and describe parallel structure of algorithms and programs. Assessment: Theory: individual assessment, written examination. Practice: implementation of basic numerical kernels on parallel computing platforms, description of informational structure and resource of parallelism of algorithms and programs. More information contact person: Name: Professor Vladimir Voevodin, Phone: +7(495)9395166, E-mail: [email protected] 12 Trainings on Applications Development and Optimization Using of domain-specific applications some­ ability). Accompanied with hand-on labs, train­ times requires a special knowledge or just a ings covered such toolkits as LWM, Score-P, simple advice how to start, which could be use­ Scalasca, Vampir, Extrae/Paraver, and Dimemas. ful even for professionals. Understanding of Also a special series of trainings by Rogue program runtime behavior – such as memory Wave specialists was arranged in MSU covered usage, logical structure, I/O influence etc. – is TotalView debugger and ThreadSpotter ana­ necessary for most optimization types and the lyzer. effective utilization of supercomputing resourc­ Advanced training for supercomputer system es. Keeping both cases in mind, MSU arranges administrators is held in MSU by popular de­ a number of trainings, intensive special courses mand of other Russian supercomputing centers. and hand-on labs for users of MSU Supercom­ Key topics of this training are supercomputers puting Center. hardware and software stack, users and access Specialists of Intel together with MSU-Intel control, batch systems, monitoring tools, com­ Center carried out a series of trainings devoted pilers, and parallel libraries. to efficient programming with Intel HPC soft­ As a CUDA Center of Excellence, MSU to­ ware tools. The key feature of the first event gether with NVIDIA offers a number of train­ was meeting with Intel specialists who helped ings and courses on GPU programming tech­ to understand and tune users own code during nologies. Topics covered are GPGPU evolution, the training. Target programming languages architecture of GPUs, CUDA, libraries (CUFFT, included C and Fortran with MPI, OpenMP, or Thrust etc.), multi-GPU technologies, profiling, combination of both. The latest event was con­ debugging and optimization, OpenACC. centrated on Intel Xeon Phi features as well as Intel Trace Collector and Analyzer practice. Accelrys Materials Studio is a popular model­ ing and simulation environment designed to al­ low researchers in materials science and chemis­ try to predict and understand the relationships of a materials atomic and molecular structure with its properties and behavior. After install­ ing the Materials Studio on “Lomonosov” super­ computer, specialists from Accelrys were invited to conduct the training devoted to this toolkit. More than 60 MSU Supercomputing Center us­ ers took part in this 4-days event. A number of trainings on performance analy­ sis were arranged in MSU in the frame of joint Russian-European projects HOPSA (HOlistic Performance System Analysis) and APOS (Ap­ plication Performance Optimization and Scal­

13 SIGMA: Collective Bank of Tests on Parallel Computing sigma.parallel.ru

SIGMA is an Internet service that helps to created, the only thing needed for students to test knowledge of students. The main feature of start the testing is Internet access. this system is the possibility to combine work of The main area of usage is parallel computing, specialists to create a distributed mode of high but SIGMA can bу easily adapted for using in quality set of tests, questions and exercises for other subject areas. testing students’ knowledge level. Every special­ ist can offer questions to be included into the Short overall statistics on using “SIGMA” can shared bank. Every teacher can make up his own be presented as follows: set of tests using the collective data bank or us­ • Universities participating: 20 ing his own questions. After the test has been • Teachers registered: 77 • Groups created: 178 • Students registered: 6521 • Tests created by teachers: 125 • Tests held: 360

14 Excursions to the Supercomputing Center

Research and Education Center “Supercom­ • Why is the Internet considered as the world puting Technologies Center“ organizes excur­ biggest computer? sions for schoolchildren to Lomonosov Moscow • How is it possible to make a supercomputer State University Supercomputing Center. system from school or home personal com­ During the excursions the schoolchildren get puters? information about different problems in super­ computing area: During these excursions a lot of other prob­ • Why has our world become computerized, lems are discussed and the real supercomputer and the computerized world – parallel? systems are shown. • Why is it not enough to use common com­ In 2013, more than 550 young people visited Supercomputing Center of MSU. puters to create a good car or even sport shoes and why a supercomputer is needed for these purposes? • Why can a modern computer occupy the whole hall, weigh 20 tons and why is it con­ sidered to be normal?

15 HPC.MSU.RU

MOSCOW STATE UNIVERSITY is one of the oldest and largest Russian educational and research centers. The history of high performance computing in MSU goes back to the mid fties of the 20th century when MSU Computing Center was founded in 1955 and equipped with up-todate computing hardware. Nowadays MSU is the leading Russian supercomputing center serving more than 600 research groups from MSU, other Russian universities and institutes of Russian Academy of Sciences.