Student of DonNTU Usachov Alex

Usachov Alex
Faculty: Computer Science
Speciality: Computer Systems and Networks
Theme of student's work:

"Development and research of programmatic facilities for an imitation design in GRID environment"


Scientific adviser: Voronoi Sergey
RUS UKR ДонНТУ Student's portal of DonNTU 
  About author

Theme of student's work

Actuality of work

In spite of already existent facilities in Grid infrastructures, there is a row of important scientific tasks, without the decision of which the fully use of Grid technologies is impossible in science. One of such is an effective management resources in the up-diffused environment. Absence of minimum on expenses algorithms of design in time during the imitation of model considerably reduces efficiency use of Grid infrastructure. And, today an actual task is research of algorithms in this area.

A purpose

The purpose of work is creation of programmatic facilities of the up-diffused system imitation design in a grid network, realizing distributing of tasks between the clients of the design system, management, research and analysis of efficiency system work, a design on the basis of treatment events. Development of the up-diffused system imitation design, by which clients can apply a multiple design, where the same model uses different datains.

Scientific novelty

The new algorithms of the up-diffused design process control are offered in Grid to the environment, which abbreviate general time of imitation design.


Calculable Grid

Grid is an infrastructure, built on basis Internet, which provides scaleable, safe, fast-acting mechanisms for a discovery and access to the remote calculable and informative resources (possibly different architecture). Creation of ultraspeed computer networks which use the opened universal interfaces and protocols was preceded, from one side, appearance of Grid technologies, decision of tasks in computer safety, such as authentification, authorizing of resources and access control to the resources.

Presently two-tier organization of resources Grid develops actively, Grid built from being in the different points of network cluster in which incorporated plenty of calculable resources. This calculable resources are fully selected in Grid and can not be used. During such organization calculable resources unite in clusters and are under the management of local manager.

A basic task of Grid without depending on realization is coordinating of collections resources. The use of the agent systems for planning of tasks in Grid will enable to untie two basic problems - scaleableness and adaptiveness. Agents give resources and accomplish the search of certain resource. An agent is set on executive computers the proprietors of which want to give the resources in Grid [1].

Implementation of tasks in Grid.

The method of organization of grid software got general confession based on the opened architecture of services of OGSA, in accordance with grid characterized as a programmatic system, which consists of the up-diffused components - services which cooperate between itself by standard, opened and universal protocols and interfaces.

Meaning mainly grid of calculable type, it is possible to examine his functioning as process of maintenance of the standardized requests for implementation of calculations, designed as tasks for the general use operating systems, this implementation of these tasks are produced on resources which get out from a general pool. We will transfer the basic stages of treatment task (figure 1):

  • Planning of resources. Special component of software - a planner selects executive resources from a general pool - task will be executed.
  • Delivery of executable files and input files on executive resources. Job processing.
  • Executable.
  • Upon termination of task delivery resulting files is on the servers storage (in particular, into working place of user).

Figure 1 Chart of the inquiry and virtualization of resources processing

Presentation of time at imitation design.

The basic task of imitation design is a correct reflection of order and temporal relations between changes in the designed system on the order of implementation events in a model. There are three concepts of time in imitation design: physical, model and processor.[4] Physical time behaves to the designed system. Model time is reproducing of physical time in a model. Under processor time imply time of implementation simulation model on a computer.

The design of the difficult systems can demand the considerable expenses of processor time. Therefore other task of imitation design is diminishing of processor time. It can be attained by the use of the multiprocessor or up-diffused systems.[8] An imitation design can be divided into successive and parallel implementation of treatment events.

Time management at the up-diffused imitation design.

Three sources of the development have the up-diffused imitation design: design, requiring for the implementation of plenty calculable resources, soldiery applications and computer games with the use of Internet. Under up-diffused imitation a design the up-diffused implementation of the single program simulation model is understood on the multiprocessor or multicomputer system.

At the up-diffused design unlike a successive design primary unit there is a not object, but called logical process. A logical process is successive podmodel.[6] Every logical process has an own set of objects and own control program. The own local list of events and own clock of local design of time has a logical process. Logical processes cooperate exceptionally by passing of messages.

Passing of messages can be carried out logical processes directly by facilities of the operating system. But a chart, resulted on a fig. 2., is more perspective


Figure 2 Chart of implementation in up-diffused model

Description of algorithms.

For now large count of conservative algorithms and their modifications, in use at the design certain classes of the real systems is known. The criterion of the use of one or another conservative algorithm for the design of the concrete systems is attitude of overhead costs of this algorithm toward winning, to got at parallel execution[12].

If conservative algorithms eliminate potential possibility of origin paradox of time even, optimistic algorithms "hope" that at parallel execution of logical processes potential possibility of origin paradox time will not become reality. In the case of origin of paradox of time optimistic algorithms will realize the "recoil" of logical process to the value of model time, in which a report, causing the paradox of time[12], was sent him. A recoil plugs in itself liquidation of consequences improper execution of logical process taking into account a report, causing the paradox of time. This mechanism got the name "Deformations of time".

For the estimation of efficiency application of the different going near a time management the model of the up-diffused system design is developed with the library of realizing conservative and optimistic algorithms. On the basis of experimental estimations by this model algorithms will be chosen subject realization in the system of design GRID.

A mechanism of work in my system.

Development of my work a chart is underlaid made from two clients and server which manages all. General chart of the system is presented. The system is managed by a server, all operations of the system pass through him. A client has two shells - www-shell and shell of windows, developed on Visual C++ MFC.



Application of Grid allows considerably to promote speed and quality of calculations, especially in the case of drawing on loosely-coupled heterogeneous calculable complexes. The Grid system is described in this work. A superficial picture is considered of algorithms at an imitation design.


At a writing of this author's abstract final job is not finished yet. Definitive end: December, 2009.


Список литературы:

  1. Березовский П.С., Коваленко В.Н. Состав и функции системы диспетчеризации заданий в гриде с некластеризованными ресурсами, ИПМ им. М.В.Келдыша РАН, Москва, 2007.
    http://www.keldysh.ru/papers/2007/prep67/prep2007_67.html
  2. Шелестов Ю.А., Скакун С.В., Куссуль О.М. Агентный подход к реализации модели поведения пользователей GRID-систем, Институт космических исследований НАНУ-НКАУ, Москва, 2004.
    http://nbuv.gov.ua/portal/natural/Npdntu/2008/ikot/08saypgs.pdf
  3. Гороховський С.С., Римарчук В.К. Технологiя сiтки (GRID) i використання агентних платформ для задач планування // Наукові записки НаУКМА. Комп'ютерні науки.- 2005.
    http://www.library.ukma.kiev.ua/e-lib/NZ/NZV36_2005/04_gorokhovsky_ss.pdf
  4. Фостер Я., Кессельман К., Тьюке С. Анатомия Грид. Создание Масштабируемых Виртуальных Организаций, перевод с англ.: Карягин Д.А., ИПМ им. М.В.Келдыша РАН, Москва 2004.
    http://gridclub.ru/library/publication.2004-11-29.7104738919/publ_file/
  5. Коваленко В.Н., Корягин Д.А. Организация ресурсов грид, ИПМ им. М.В.Келдыша РАН, Москва 2004.
    http://www.keldysh.ru/papers/2004/prep63/prep2004_63.html
  6. Коваленко В.Н., Коваленко Е.И., Шорин О.Н. Разработка диспетчера заданий грид, основанного на опережающем планировании, ИПМ им. М.В.Келдыша РАН, Москва 2005.
    http://www.keldysh.ru/papers/2005/prep133/prep2005_133.html
  7. А.Е. Дорошенко, А.П. Розенблат, К.А. Рухлис, Ю.М. Тырчак МОДЕЛИ И СРЕДСТВА ПРОГРАММИРОВАНИЯ Грид-СИСТЕМ, Институт программных систем НАН Украины, Киев, 2005.
    http://usachov1987.narod.ru/Doroshenko_1.doc
  8. Аветисян А.И., Гайсарян С.С., Грушин Д.А., Кузюрин Н.Н., Шокуров А.В. Эвристики распределения задач для брокера ресурсов Grid, Институт системного программирования РАН, Москва, 2004.
    http://www.citforum.ru/nets/digest/grid/index.shtml
  9. Фостер Я. Разные оттенки ГРИД. Еженедельник "Computerworld", #13, 2005 год // Издательство "Открытые системы".
    http://www.osp.ru/cw/2005/13/038_1.htm
  10. David W. Emerging Distributed Computing Technologies, Department of Computer Science, Cardiff University, Cardiff, 2004.
    http://www.cs.cf.ac.uk/User/David.W.Walker/IGDS/GridCourse.doc
  11. Дал О.И. Нигард К. Симула - язык для программирования и описания систем с дискретными событиями // Алгоритмы и алгоритмические языки. Вып. 2. М.: ВЦ АН СССР, 1967
    www.gpss.ru/paper/okoln_2.pdf
  12. Шеннон Р. Имитационное моделирование систем - искусство и наука. М.: Мир 1978
    www.twirpx.com/file/21030/

Alex Usachov