N. I. Lobachevsky State University of Nizhni Novgorod


Print-Friendly Page

MODELS, METHODS, SOFTWARE AND PRACTICAL APPLICATIONS IN THE FIELD OF HIGH-PERFOMANCE COMPUTING:

 

- Optimization of realizations of system level packages for creating parallel programs;
 
- Development and realization of program systems for supporting high-performance computing on cluster systems;
 
- Development and realization of effective parallel versions of applied software packages;
 
- Development of effective parallel algorithms of global search;
 
- Realization of educational programs  for preparation of experts, professional skills development and retraining of personnel in the field of high-performance computing.

 

Main results:
Leading expert:
Main technological and research equipment:
Main partners and customers:
Key projects (sources of financing):
Main publications:
 

 

Main  results (comparable with the best world analogues):
- The algorithm for message broadcasting have been developed and implemented with the account of the specific features of the  architecture of clusters built on Power systems. The scheme for data transmission has been developed that enables effective realization of local message transmission in collective operations;
- Control system «Metacluster» has been developed and introduced as a control system for clusters at UNN. The main distinctive features of this system are the opportunity of exercising simultaneous control several computing clusters, and the opportunity of supporting various operational systems and integration with other control systems;
- Effective parallel algorithms for output and learning of probabilistic networks have been developed and implemented in PNL library;
- Analysis  and comparison have been performed as to the efficiency of implementation of Cluster OpenMP technology in Intel compilers  and the realization of MPI library; we have also offered and tried out (using PNL library as an example) a technique for increasing the efficiency of Cluster OpenMP applications;
- Adaptation of the control system of UNN’s high-performance cluster ("Metacluster" system) for connecting the cluster to the experimental grid segment of the "SKIF-GRID" program has been performed;
- The porting of the GROMACS package for solving molecular dynamics problems to suit Cell BE architecture has been performed, general methodology for transferring and optimization of existing programs for effective performance on the processors of the Cell family has been developed.

contents ↑

 

Leading expert:
- Victor Pavlovich Gergel, D.Sc. (Engineering), Professor.

contents ↑

 

Main technological and research equipment:
- High-performance computing cluster installed at the University of Nizhni Novgorod in the framework of the priority national project "Education" comprising 64 dual-processor servers (128 dual-core processors) Intel Xeon 5150 2,66 GHz, 4GB DDR2-667 FB-DIMM RAM, 80 GB HDD, 2 õ Gigabit Ethernet; total number of processors - 128, kernels - 256, peak performance – 2700 GFlops.

contents ↑

 

Main partners/customers:
- Moscow State University,
- Institute of Applied Mathematics of the Russian Academy of Sciences,
- Intel Corp.,
- Microsoft Corp.,
- T-Platforms Company.

contents ↑

 

Key projects (sources of financing):
- The project «Scalable parallel algorithms for output and learning of probabilistic networks within the framework of Probabilistic Network Library» was carried out with the support of Intel Corp. 2003-2005.
- The project «Optimization of open implementations of MPI standard for Linux on the clusters of Power architecture» was carried out with the support of the IBM grant (Faculty Awards Program), 2005.
- The project "Development of Metacluster system for controlling computing resources in multiprocessing computer systems  (clusters)”, 2006-2009.
- The project «Technology for development of  distributed parallel programs by means of Cluster OpenMP» was carried out with the support of  Intel Corp. in 2006-2007.
- The project «Establishing the Nizhni Novgorod node  of the pilot segment of the distributed grid system "SKIF" was carried out in the framework of project No. 2007-SG-04/5 of the scientific and technical program of the Union State “Development and use of hardware and software for  grid technologies of promising high-performance (supercomputer) computing systems of the SKIF family ("SKIF-GRID") in 2007-2008.
- The project «Adaptation of the GROMACS library for Cell BE architecture» was carried out  with the support of T-Platforms company in  2008.
- The project «Research of the problem of porting of  weather forecasting to the system of grid computing» has been performed with the support of  Intel Corp. in 2008-2009.
- The project «Development of a high-performance program system for quantum-mechanical calculations and modelling nano-sized atomic-molecular systems and complexes» is being carried out within the framework of the Federal Target Program «R&D in priority directions of development of the scientific and technological complex of Russia in 2007-2012», State Contract No. 02.524.11.4005, code: 2008-04-2.4-15-02-003, 2008-2009.

contents ↑

 

Main publications:
- Gergel V.P. The theory and practice of parallel computing. – Ì .: The Internet University of Information Technologies – INTUIT.RU, 2007. – 424 Pp.
- Gergel V.P., Strongin R.G. (2001). Fundamentals of parallel computing for multiprocessing computing systems. – N.Novgorod, NNSU (2nd edition, 2003).
- Strongin R.G., Sergeev Ya. D. (2000). Global optimization with non-convex constraints: Sequential and parallel algorithms. Kluwer Academic Publisher s , Dordrecht.

contents ↑



Our resources can help you to accelerate your business growth and profit!



Contacts