|
|
|
Surname:
|
Vazhenin
|
|
First Name:
|
Alexander
|
|
Phone:
|
+81-0242-37-2717
|
|
Fax:
|
+81-0242-37-2733
|
|
E-Mail:
|
vazhenin@u-aizu.ac.jp
|
|
2. Education
Ph.D. (Candidate of Sciences) in Computer Science at the Institute of Informatics
Systems of the Siberian Division of the Russian Academy of Sciences with the subject: "Methods and Tools for
Parallel High-accuracy Arithmetic in Vertical Processing Systems" in 1993.
M.Sc. (Computer Engineering) at the Novosibirsk State Technical University in 1978.
3. Employment
1999-2000: senior
researcher at the Institute of Computational Mathematics and Mathematical Geophysics (former Computing Center)
of the Siberian Division of the Russian Academy of Sciences, Novosibirsk
Duties: research and development of multimedia tools for high-performance and high-accuracy computations,
architecture and software of massively parallel systems, parallel algorithms; supervision of BS, MS and Ph.D. theses.
1997-1999: visiting researcher
(associate professor) at the University of Aizu, Japan
Duties: participation in research and development of experimental multimedia system based on VIM-technology
1987-1997: researcher, senior researcher at the Computing Center of the Siberian Division of the Russian
Academy of Sciences, Novosibirsk
Duties: participation in research and development of experimental heterogeneous parallel computing
system "Siberia" oriented to high-performance processing of geophysical information; management of a
team of researchers and programmers in the field of parallel high-accuracy computations, architecture and software
of massively parallel systems, parallel algorithms; supervision of BS, MS and Ph.D. theses.
1978-1987: engineer, researcher, and senior researcher at the Novosibirsk State Technical University
Duties: participation in research, design and development of adaptive computing/operating systems
for an automatic capture and processing of experimental data; supervision of master theses.
4. Research Interests
Parallel Architectures
Filmification and Visualization of Methods and
Information Processes
Operating Systems
Heterogeneous Processing
Parallel Algorithms in Numerical Analysis
Multi-precision Parallel Arithmetic
Software Engineering in Computational Science Applications
5. Research and Development
VIM: MULTITIMEDIA PROGRAMMING
TECHNOLOGY
The Visualization of Methods (VIM) technology is to represent and specify
application methods using different multimedia elements. According to this technology, each method is considered
as a plan of actions, which should be implemented to solve a problem. To specify a method (a task to computer)
we use abstract films that are a series of stills showing some dynamics (transformations). Each still reflects
a computational step. This allows considering a film as a "dynamical shape" and "an idea" of
computation. The analysis and evaluation include a consideration of new multimedia signs for human-computer interface
based suitable combination of sounds, color, movement, and text for presentation large-scale algorithms. Developing
VIM-technology concerns in creating original multimedia signs to represent and implement sequential, parallel and
associative algorithms as well as approaches providing composition and superposition of methods. The great attention
is paid to the special visual language, which is oriented to input and verify arithmetic/logical expressions, as
well as special user-oriented interface using multimedia icons and high-level program keyboards. The automatic
generation of an executable code is based on the film-oriented compiling and computer-oriented coding.
An important issue of the visual language is the presence of icons to synthesize computational formulas. In this
case, the big attention should be paid to the investigation of representation model and retrieval of icon images
as well as creation of the new ones from a set of existing icons. Our representation model and similarity evaluation
also simply the retrieval progress of linear transformations, include rotation and reflection, of an image.
An important part of investigations is related to the matrix computations, which are widely used for solving many
scientific and practical problems. The linear algebra problems (searching for norm or condition number of a matrix,
partial or full eigenvalue problems, solving the system of linear algebraic equations (SLAE) or reversing of a
matrix, etc.) are appeared widespread either independently or as a part of a large project. There is produced a
big number of different methods that is why the problem of using and understanding them becomes very urge.
WWW-ORIENTED MULTIMEDIA TOOLS
FOR APPLIED PROGRAMMING
The current multimedia systems and Internet allow geographically distributed
users to send, receive and manipulate media-rich information. That is why we have now very interesting WWW-applications.
One of them can be a trend improving traditional programming techniques because of involving a lot non-professional
people in software engineering. The goal of this project is in research and developing of the WWW-oriented Workplace
of Applied Programmer (W4AP). From the user's viewpoint, this workplace represents an application oriented virtual
mashine being accessible via standard WEB-browsers. We use modern technologies like object-oriented programming,
clustering computational and multimedia resources, networking as well as parallel, distributed and client/server
platforms and Internet Computing. The developing multimedia software having the following important features:
- Convenient,
intuitively understandable, simple in learning WWW-interface based on standard browsers like Netscape, MS Explorer,
etc.
- Unified environment hiding from user unimportant
hardware and system software details.
- Effective client/server subsystem for storing,
manipulating and serving the user's projects.
- Accessibility, openness and multi-platform support
of both client and server parts of a WWW-server. This means the creation of the Computing Informative WWW-server
supporting different programming systems (GNU C/C++, MPICH, MPI, LAM, JAVA, etc.) as well as different operating
systems (MS Windows, Solaris, Linux, etc.).
- Safety of all system elements based on combination
of standard and special security tools as well as control of an applied program implementation.
- Visual programming VIM-technology based on manipulating
with multimedia objects called films, which can be considered as self-explained multimedia units using to represent
computations and data.
- The template manipulating subsystem providing automatic
or semi-automatic generation of the user's application programs.
The Java-based programming systems are used for the
server creation to support the software flexibility, the high-level of human-computer interaction and effective
network communications.
To support distance learning, lectures developing should be based on distributed architecture using client-server
approach. To use these courses, a student should download the programming kernel of course allowing synthesis,
visualization and mapping necessary films on client-machine. In this case, server functions are reduced to form
control sequences pointing what is necessary to implement and/or visualize. This approach can significantly decrease
communication loads because of reducing/excluding data transfer such as images, audio/video, huge information arrays
(matrices, numerical data, etc.). This improves efficiency of communications and makes lectures more information
intensive.
METHODS,
TOOLS AND ALGORITHMS FOR HIGH-ACCURACY COMPUTATIONS
The goal of this project is to investigate the propagation of rounding
errors in solving the problems of linear algebra, mathematical physics, finite element method as well as the methods
overcoming these errors. The parallel algorithms that are stable to the errors of this kind will be developed.
The guaranteed result accuracy for an acceptable time is obtained by using the multi-precision arithmetic, dynamically
changed length of operands, and parallel processing. The original approach named SPARTH (Super-precision Parallel
AriTHmetic) was proposed to create a parallel/distributed system consisting in computational nodes each of which
has a variable and programmable word length. This allows adapting conventional high-parallel algorithms to SPARTH-computations.
Software tools were designed for a massively parallel
SIMD-system ES-2720 with the STARAN-like architecture. From the user's viewpoint, this programming system can be
represented as a virtual vector processor with the programmable length of vector operands named SPARTH-processor.
The portable message passing SPARTH-system was developed and tested on the following systems: IBM PC, Sun, multitransputer
workstations, Power X'Plorer (Parsytec), MVS-100 (NII KVANT, Russia), iPSC/860 (Intel), Silicon Power Challenge,
etc. A set of new parallel/distributed algorithms using the dynamically changed length of operands was designed
for solving problems of linear algebra, polynomial evaluation, calculation of transcendent functions, etc. These
programs were tested on systems mentioned above.
The current direction of this research is to combine
SPARTH- and VIM-technologies. This leads to the creation of the visual SPARTH-oriented software.
CONCENTRATED
HETEROGENEOUS SUPERCOMPUTING
The co-called Combined Architecture is based on the cooperation of a powerful
basic computer of parallel architecture with a set of specialized coprocessors. This approach can be considered
as a concentrated heterogeneous processing in contrast to distributed heterogeneous systems based on commercially
available computers of diverse architectures connected by networks.
In a Combined Architecture (CA), solving of any problem is considered as interaction of several processes, so that
execution of each process is delegated to a specialized subsystem that is most efficient in implementation of this
process. The subsystems are controlled in such a way that their balanced operation might be ensured, and special
complementing features of subsystems might be best exploited. For each subsystem a structure is chosen which best
corresponds to the function it should perform. This approach can be used, for example, for designing the general-purpose
computers of advanced architectures combined massively parallel SIMD-systems and systolic array processors. The
programming system is intended for simulation, exploration of possibilities and evaluation of efficiency of such
Combined Architectures in solving various types of problems. A language VEPRAN was suggested for describing CA-algorithms,
and estimating quality of their software/hardware realizations.
COMPUTING
INFORMATION SYSTEMS FOR A CAPTURE AND PROCESSING OF EXPERIMENTAL DATA
A family of systems consisting
of the SM-3 and SM-4 host-computers and specialized coprocessors of different kind was designed. The SM-3/4 architectures
are similar to the PDP-11/05 and PDP-11/40 computers (DEC, USA) correspondingly.
The original hardware resources, applied program environment, and real time operating systems were developed.
The operating systems key-point is the presence of the complimentary kernel including standard and special software.
The last one represents a set of high-performance tools (drivers) for effective hardware control as well as for
high-speed capturing of experimental data. This operating system includes also the special editors for creating
the experiment program defining how to scan detectors, volume of capturing data, how to process experimental data,
information to identify experiments as well as software for statistical processing.
The high-accuracy and high-performance data processing is supported by special devices called Adaptive Normalizing
Processors (ANP), which implement the automatic matching of a dynamic range of values of input data (analog or
digital) with a given input range of processing devices. The algorithm implementing of such matching can be selected
according to the processing type (statistical or spectral analysis of random processes, digital signal processing,
etc.). The original ANP-architecture was patented.
The extension of the standard FORTRAN-DEC was proposed to provide the effective control of coprocessors by means
of high-level language. A new data type named "the event-variables" was included in the standard FORTRAN
language. These variables allow the user in getting get status information about the selected coprocessor and
organizing fast and accurate data exchange between coprocessors and the FORTRAN program in any required time. The
Applied Program Package for Spectrum Analysis called PACSAN was developed to provide processing of experimental
data. The automatic control system supporting dynamical tests of the gas-dynamic engines was build on the basis
of the systems described.
6. Teaching
Associate Professor
and Scientific Secretary, Chair of Parallel Computing Technologies,
State Technical University, Novosibirsk, 1996-1997.
Leading Lecturer of the following courses:
"Design and Analysis of Parallel Algorithms", "Architecture
of Computing and Information Systems", "Languages and Systems of Parallel Programming".
Associate Professor, Chair of Applied Mathematics, State Technical
University, Novosibirsk, 1996.
Leading Lecturer of the following courses:
"Architecture of Computing Systems"
Supervisor of the Student Research Projects
on "Numerical Methods", "Methods of Optimization"
Associate
Professor, Chair of Computing Processing of Economic Information, State
Technical University, Novosibirsk, 1994.
Lecturer of Course of "Principles of Computer Architecture"
Assistant
Professor, Chair of Computing
Systems, State University, Novosibirsk, 1993-1996.
Leading Lecturer of the following course:
"Languages and Systems of Parallel Programming"
Supervisor of over 40 graduation theses for BS and MS degree at State Technical University and State
University, Novosibirsk, 1980-1997.
Referee of many graduation thesis at State Technical University and State University, Novosibirsk,
1980-1997.
Member
of State Examining Boards for BS and MS Degree in Computing Science
and Applied Mathematics at the State Technical University (1986, 1987, 1992-1997), and State University (1994,
1995), Novosibirsk.
7. Academic Activities
Chairman
of the Program Committee of the
Second Aizu International Student Forum-Contest in Multimedia (Aizuwakamatsu, Japan, 1999).
Chairman of
the Program Committee of the Sixth
International Conference on Distributed Multimedia Systems (DMS'99) (Aizuwakamatsu, Japan, 1999).
Vice-chairman
of the Organizing Committee of the Aizu International Student Forum-Contest
in Multimedia (Aizuwakamatsu, Japan, 1998).
Reviewer of the IEEE Computer Society.
Participant
of International Conferences with accepted papers: CONPAR90/VAPP-IV,
(Zurich, Switzerland, 1990), PaCT-91 (Novosibirsk, USSR, 1991), PaCT-93 (Obninsk, Russia, 1993), ParCo93 (Grenoble,
France, 1993), ASE-93 (Bath, UK, 1993), PARA94-L (Lyngby, Denmark, 1994), pAs'95 (Aizu-Wakamatsu, Fukushima, Japan,
1995), PARA95-L (Lyngby, Denmark, 1994), PaCT-95 (St.-Petersburg, Russia, 1995), AMCA-95 (Novosibirsk, Russia,
1995), The Second Int. Conf. "Real Numbers and Computers" (Marseille, France, 1996), pAs'97 (Aizu-Wakamatsu,
Fukushima, Japan, 1997), PaCT-97 (Yaroslavl, Russia, 1997), The 3rd Int. Conf. "Real Numbers and Computers"
(Paris, France, 1998), ICSE’98 (Kyoto, Japan, 1998), PDPTA'98 (Las-Vegas, USA,1998).
Member
of Organizing Committee of the International Conferences "Parallel
Computing Technologies": PaCT-91 ( Novosibirsk, Russia, 1991), PaCT-93 (Obninsk, Russia, 1993), PaCT-95 (St.-Petersburg,
Russia, 1995), PaCT-97 (Yaroslavl, Russia, 1997).
Visiting
scientist in the Universites
D'Aix-Marseille I & IRPHE (Marseille, France, 1997), the Institute of Technical Cybernetics (Slovak Acad .Sci.,
Bratislava, Czechoslovakia, 1988).
Scientific
Secretary of Supercomputer Software Department, Computing Center of
Siberian Division of the Russian Academy of Sciences, Novosibirsk, 1990-1996.
Invited
Speaker and Participant of
many Russian Conferences, Seminars and Workshops (1980-1996) on Computing Systems for Scientific Research, on Parallel
Architecture, Algorithms, and Programming.