Summaries of the Issue
PROGRESS IN THE IMPLEMENTATION RESEARCH PROJECTS FTP "Research and development in priority areas of science and technology complex in 2007-2012"
HIGH-PERFORMANCE SOFTWARE FOR NANO-DIMENSIONAL ATOM-MOLECULE SYSTEMS MODELING Vladimir N. Vasilev, Boukhanovsky Alexander V, Kozlov Sergei A, Maslov Vladimir G, Rosanov Nikolay Nikolaevich
Approach for building of next-generation high-performance software aimed at nanostructure and nanocomplex modeling within the bounds of iPSE concept is proposed. Proof of high-level architecture of this software is given; structure and purposes of main problem-oriented services are defined; issues of human-computer interaction organization, data- and knowledge management, results visualization are considered.
CONCEPTION AND METHODOLOGICAL BASIS FOR INTELLIGENT GRID SYSTEMS BUILDING Yuri Nechaev, Boukhanovsky Alexander V, Vladimir N. Vasilev
The problem of intelligent Grid systems is discussed. The main attention is focused on formulating of conceptual basis and principals of data processing in complex dynamic environment. Formal description of logic space for knowledge model integration, including Neuro-Fuzzy systems and multi-agent technologies, is produced. Features of computing intelligence are considered, choice model, main definitions and axioms of intelligent Grid systems are formulated. Interpretation of Grid system as parallel computing architecture is given.
HIGH-PERFORMANCE GRID-APPLICATIONS DEVELOPMENT ENVIRONMENT. PART I: PRINCIPLES Alexey Larchenko, Anton Dunaev, Boukhanovsky Alexander V
Grid-service-based high-performance applications development principles are discussed. A new design approach based on a special type of workflow is proposed. The use of Grid-applications developer decision support is well founded. Basic principles of intelligent decision support system working are described.
HIGH-PERFORMANCE GRID-APPLICATIONS DEVELOPMENT ENVIRONMENT. PART II: ARCHITECTURE, IMPLEMENTATION AND APPLICATION Boukhanovsky Alexander V, Anton Dunaev, Alexey Larchenko
Issues to practical implementation of the decision support system for high-performance applications developer in PEG software family are discussed. Basic principles of the system working are described. Also assessments of composite applications effectiveness using task of climatic spectra approximation are presented.
HIGH-PERFORMANCE GRID-APPLICATIONS DEVELOPMENT ENVIRONMENT. PART III: KNOWLEDGE ACQUISITION AND FORMALIZATION Boukhanovsky Alexander V, Alexey Larchenko, Anton Dunaev
Acquisition, knowledge representation and formalization of Grid services performance is discussed in this paper. Hierarchy of analytical models as determinate function of random variable (parameters of Grid environment) that represents the application working time is considered. Indirect approach to knowledge acquisition about performance using simulation modeling is introduced. Identifying method of parallel speedup probabilistic capabilities is presented. It is based on the knowledge about application working time distribution.
HIGH-PERFORMANCE SOFTWARE FOR HYDROMETEOROLOGICAL EXTREME EVENTS SIMULATION. PART I: PROBLEM STATEMENT, MODELS, METHODS AND PARALLEL ALGORITHMS Boukhanovsky Alexander V, Ivanov Sergey V, Kovalchuk Sergey V., Oleg Zilberstein, Sergey Popov, Leonid Lopatoukhin, Michael Chumakov
Issues of new concept implementation based on computer simulation for new extreme hydrometeorological events characteristics producing are discussed. General stages of calculation process, main features of modern approaches, numerical methods and parallel algorithms for hydrodynamic and statistic simulation of wind, waves, currents and sea level fields are described. Main calculation operations performance estimation is considered, a proof for supercomputer architecture choice is given.
HIGH-PERFORMANCE SOFTWARE FOR HYDROMETEOROLOGICAL EXTREME EVENTS SIMULATION. PART II: SOFTWARE ARCHITECTURE DESIGN AND ASSESSMENT Boukhanovsky Alexander V, Ivanov Sergey V, Kovalchuk Sergey V.
High-performance software for hydrometeorological extreme events simulation design and development peculiarities are considered within the bounds of Service-Oriented Architecture paradigm (SOA). It is shown by investigation using prototype software and parametric performance model, that parallel decomposition type for each of calculation modules of developing software is defined by task features and parallel hardware and software architecture specific character. The approach for dynamic performance-efficient organization of parallel software architecture by using intelligent management service is proposed. Results of architecture assessment using computing experiments are presented.
HIGH-PERFORMANCE SOFTWARE FOR METOCEAN EXTREME EVENTS SIMULATION. PART III: CALCULATION RESULTS INTERPRETATION AND APPLICATION Ivanov Sergey V, Kovalchuk Sergey V., Leonid Lopatoukhin, Ekaterina Chernysheva, Boukhanovsky Alexander V
Forms and ways of extreme hydrometeorological characteristics representation based on computer simulation are discussed. General types of extreme are considered: in point, in field, by months and by directions, multidimensional characteristics. Extreme wave height and wind speed estimation for the Barents Sea and the Caspian Sea is described.
INSTRUMENTAL ENVIRONMENT FOR MASS MOBILE NEW-AGE ONLINE-SERVICES. PART 1. OPERATING PRINCIPLE, SOFTWARE ARCHITECTURE Oleg Zolotarev, Ekaterina Grinina, Boukhanovsky Alexander V
The issue on distributed system software designing and engineering process automation for mobile commerce is considered. It is shown that mobile service architecture and operating principle concept is defined basically by e-Payment regulation character. Instrumental environment supporting construction process for mobile service, suitable for penetration into Russian mobile market, is developed.
INSTRUMENTAL ENVIRONMENT FOR MASS MOBILE NEW-AGE ONLINE-SERVICES. PART 2. NFC TECHNOLOGY Ekaterina Grinina, Oleg Zolotarev, Ilya Pimenov, Nasonov Denis A, Boukhanovsky Alexander V
This paper presents comprehensive study on issues concerning mobile online service based on NFC local communication technology. The mobile ticketing Id is considered as an example.
SELECTED PUBLICATIONS SCIENTIFIC SCHOOL SESSION I "High performance computing technologies and systems"
INSTRUMENTAL ENVIRONMENT FOR MASS MOBILE NEW-AGE ONLINE-SERVICES. PART 2. NFC TECHNOLOGY Kovalchuk Sergey V., Sergey Vishnyakov
Issues on calculation algorithm mapping onto GPU-hardware parallel architecture are discussed in this work. Main features of considered hardware architecture and their impact on parallel algorithms performance are described. Performance estimation model for parallel algorithm using computational resources of graphics accelerator is proposed. Optimization of parametric function approximation using random search is taken as an example of calculation task.
IDENTIFICATION OF PARAMETRIC BOUND MODELS OF COMPLEX SYSTEMS Ivanov Sergey V
Approaches for complex systems models identification are considered in the paper. The distinctive features of models are parametric bound, multi-scalability and incompleteness of observations. Effectiveness of this approach is proved by several examples. Schemes of parameter identification for two models in system “ocean-atmosphere” and HIV population dynamic model are shown.
PARALLEL SIMULATION OF ASYNCHRONOUS CELLULAR AUTOMATA EVOLUTION Konstantin Kalgin
For simulating physical and chemical processes on molecular level, asynchronous cellular automata with probabilistic transition rules are widely used being sometimes referred to as Monte-Carlo methods. The simulation requires a huge cellular space and millions of iterative steps for obtaining the CA evolution representing a real scene of the process. This may be attained by allocating the CA evolution program onto a multiprocessor system. We propose a new parallelization method of asynchronous CA based on its stochastic properties. The experiment results are presented.
ARCHITECTURE OF SERVICE-ORIENTED DISTRIBUTED COMPUTED SYSTEM AND ITS REALIZATION ON THE BASE OF FRAMEWORK OSGI Andrey Gavrilov, Irina Dorovskih, Ilya Krasinsky
Development of distributed computing systems (DCS) is important for solving problems that require a great number of calculations. Analysis of existing widespread DCS has revealed several shortcomings. New DCS architecture is introduced that allows to avoid these shortcomings. The architecture is implemented in DCS based on OSGi framework. This system was used for solving particular problems and has shown its performance.
COMPARATIVE STUDY OF TABU-MACHINE AND HOPFIELD NEURAL NETWORKS APPLICATION TO SOLVE DISCRETE OPTIMIZATION PROBLEMS FROM DISTRIBUTED DATABASES DOMAIN Edward Babkin, Margaret Karpunina
We propose a new neural networks approach based on tabu-search to solve discrete optimization problems arising during the synthesis of optimal distributed database logical structures. The solutions with a quality on average greater by 3.8% than quality of the solutions, received with help of Hopfield networks, have been obtained. Thanks to optimization of Tabu-machine the operating time of data decomposition algorithm has been lowered in compare with an initial version of tabu-algorithm (on average more than twice) as well as with neural networks algorithm based on Hopfield networks (on average on 36%). In the course of the research the investigation of Tabu-machine parameters space has been carried out.
DEDICATED NUMERATOR FOR RADAR SIGNAL PROCESSING Anatoly I. Grushin, Maxim L. Remizov, Artem V. Rostovtsev, Nikolaev Dmitry D., Chin Kuang Kien
Real time complex matrix calculation is a typical radar signal processing problem. Implementation of recursive calculation of complex matrix 64x64 is considered in the paper.
TECHNOLOGY OF VIRTUAL TEST DESKS CONSTRUCTING IN DISTRIBUTED COMPUTING ENVIRONMENT Leonid Sokolinsky, Gleb Radchenko
We propose a technology of virtual test desks constructing in distributed computing environment based on a CAEBeans technology. We describe a structure organization of CAEbeans shells. We describe basic principles of CAEBeans system implementation.
PARALLEL SOFTWARE DEVELOPMENT IN PARJAVA Arutyun Avetisyan, Varvara Babkova, Michael Kalugin
The goal of the paper is development of a scalable parallel program with the help of ParJava environment. The problems concerning program optimization in order to increase its scalability are discussed, and ParJava tools are presented.
EFFECTIVE UTILIZATION OF COMPUTATIONAL RESOURCES FOR BERMUDA OPTIONS RATIONAL PRICES ESTIMATION Anna Gorbunova, Iosif Meerov, Andrey Nikonov, Andrey Rusakov, Alexander Shishkov
Our results in effective implementation of one algorithm of rational price determination for Bermuda options are reported in the paper. The approaches for optimization of calculations both in serial and parallel versions based on MPI and OpenMP are described. Computational results show linear speedup at clusters/SMP systems.
SIMULATION INVIRONMENT FOR PERCOLATING PROBLEMS Evdokiya N. Golovchenko, Dmitry Petrov
Task preparation and launching on modern multiuser systems have a number of peculiar properties. Execution of cluster experiments is connected with necessity of multiple authorization, state control of files distributed among network resources, synchronization of stages of the sequences of commands execution and other difficulties. A simulation environment was created for simplification of the process of interaction with cluster systems. Using this environment water flooding of oilfield was carried out. Oilfield was approximated with percolating meshes with 106 – 1010 vertices. Steady and nonsteady states of influence on oilfield were explored. Diagrams of oil distribution, water, pressure and vacancy over the field were obtained.
SOFT DEVELOPMENT OF COMPETITIVE MARKET MODELING FOR COMPUTATIONAL CLUSTERS Maxim Nesternko, Denis Leonov, Ekaterina Bolgova, Alexey Kirillov
The approach to computational difficult problem – the finding of optimal coalition and gain distribution in n-players cooperative games is considered. The parallel algorithm of cooperative games solving is proposed. The testing results of algorithm program realization are analyzed.
ADVANCED RESEARCH IN HIGH-PERFORMANCE COMPUTING AND COMPUTER MODELING. BRIEF SCIENTIFIC REPORTS
SIMULATION OF LARGE-SCALE VORTEX STRUCURES ON PARALLEL COMPUTERS Alexander Babakov, Pavel Novikov
The given work is devoted to studying of landing modules aerodynamics for supersonic and transonic modes of a running flow. The near wake structure, formation and interaction of large-scale vortical structures are investigated for hypersonic flows influence on aerodynamic characteristics of equilibrium and non-equilibrium chemical reactions.
LOCALLY-REFINED RECTANGULAR MESHES Anton Sukhinov
Two-dimensional hierarchical locally-refined rectangular meshes with dynamic adaptation to the solution are considered in this paper. The mesh structure, adaptation criterion and parallel algorithm are described.
INTERPRETATION AS RESEARCH METHOD FOR DYNAMICAL PROPERTIES OF PARALLEL PROGRAM ON INSTRUMENTAL COMPUTER Manuk Akopyan
Analysis and tuning of parallel program on computational platform is a resource-intensive process. Our approach consists of the model simulation of parallel program using instrumental computer to analyze its dynamic characteristics.
Method of solving the problem of numerical modeling of seismic waves propagation in the 3D models of inhomogeneous media is briefly presented in this paper. The method for constructing the 3D model of elastic media is described. The parallel implementation of the program for numerical modeling of elastic waves propagation is described. Some results of numerical modeling that can be used in interpreting geophysical field experiments with complicated media are presented.
PARALLEL COMPUTING WITH «CELL» MICROPROCESSORS Sergey Vishnyakov, Alexander Mordvintsev
Adaptation of various kinds of computational tasks to CBEA architecture is considered in this work. Ray tracing through 3D-scene that has voxel representation is taken as example – an application that allows user to interact with 3D-scene is developed.
STOCHASTIC SIMULATION OF COMPLEX NETWORKS Ilya Kolyhmatov
Research devoted to design and analysis of parallel algorithms for stochastic simulation of complex networks with respect to heterogeneity and nonstationarity is represented in the article. Proposed algorithms were applied to several problems in mathematical epidemiology.
METROLOGICAL SUPPLY FOR DISTRIBUTED COMPUTATION AND SYSTEMS Boukhanovsky Alexander V, Alexander Semyonov
Modern state of metrological supply for distributed computation and systems technologies problems is considered. Objects, methods and facilities used in the subject field are discussed from measurement unity supply point of view.
There is a significant grow in graphic co-processors utilization during last years because of new programming tools emerging. However it’s still a rare case when using it for parallel computing. The article discusses perspectives of this new research field on the example of MVS-EXPRESS supercomputer prototype built with graphic accelerators.
MAIN MEMORY SPEED ANALYSIS AND PC PERFORMANCE MODELING Konstantin Solnushkin
In this work main memory speed has been researched as a function with memory subsystem parameters: bus frequency and behavior mode of memory controller. Performance model has been introduced and approved by experiment. It determines linear dependence between PC performance and main memory speed.