Contemporary Problems of Social Work


INFORMATION AND TELECOMMUNICATION TECHNOLOGIES..

Optimization of the computational process the fitness function of genetic algorithm in distributed systems processing data

Автор/Author: Sizov V.A., Uralskiy N.B.

Аннотация/Annotation:
Скачать/Download | Посмотреть онлайн/View online

Список литературы/References:

1. Averchenkov C. I., Kazakov, P. C. Evolutionary modeling and its application / 2 ed., - M. : Flint,

2011

2. Babin, D. C., Crow, S. M., “Genetic algorithm for constructing schedules for multiprocessor

computing systems”,Donetsk state Institute of artificial intelligence, Ukraine “Artificial

intelligence”,2005

3. Bukatov A. A., C. N. Datsyuk, A. I., Gigolo Programming of multiprocessor computing systems/

Rostov-on-don: LLC “CWR”, 2003

4. Volkov D. Phenomenon of big ideas/Open systems №. 8, 2013 www.osmag.ru

120 121

CONTEMPORARY PROBLEMS OF SOCIAL WORK VOLUME 1, No. 2, 2015

5. Gorbunov Century, Elizarov,, Eisymont L. Ectatomminae supercomputers achievements and

prospects/Open system No. 7, 2013, www.osmag.ru

6. Efimov, A. C., C. N. Samoilenko, E. N. Paraskova Organization functioning of distributed

computing systems when processing sets scalable tasks/Vestnik of Tomsk state University,

Management, computing and information science № 2(15), 2011

7. Jaspal Singh, Harsharanpal Singh, Efficient Tasks scheduling for heterogeneous multiprocessor

using Genetic algorithm with Node duplication”, Indian Journal of Computer Science and

Engineering Vol. 2 No. 3 Jun-Jul 2011.

8. Kalyaev A. I. About one method of multi-agent organization’s cloud-based network of

computers private users - Taganrog: Institute of Multiprocessor Computing Systems, Southern

Federal University, 2012

9. Kazunori Kojima, Masaaki Ishigame, Goutam Chakraborty, Hiroshi Hatsuo, and Shozo Makino

“Asynchronous Parallel Distributed Genetic Algorithm with Elite Migration” International

Journal of Information and Mathematical Sciences 4:2 2008.

10. Medvedev V.I. Features of object-oriented programming in C++/CLI, C# and Java/ 2 ed., -

Kazan: RIC “School”, 2010

11. Portal artificial intelligence, Methods of selection in genetic algorithms /http://www.

aiportal.ru/articles/genetic-algorithms/methodsselection.html

12. Radchenko, I. Distributed computing systems, Chelyabinsk: Photographer, 2012

13. Rudenko Y. M. Distribution of software modules on the nodes of the computer network with a

common field memory for graph-schemes of parallel algorithms/Science and education № 10,

2011

14. Salnikov A. N. System design and support the execution of parallel programs/Dissertation on

competition of a scientific degree of candidate of physical-mathematical Sciences, -M, 2006

15. Sizov V.A. The design of software and informationware of the complex of related tasks in

computer network/Automated control systems, 1995

16. Uralskiy N. B. Sizov V.A. The development of optimization models and algorithms schedules

the execution of software modules in the computer network based on an evolutionary

approach/ Materials of III international scientific-practical conference “Problems of stability

and security systems in the sphere of housing and communal services”, 2012

17. Hohlova, M. N. The theory of evolutionary modeling/”Cepheus”, 2013

18. Coi Y. R., Spitsyn Century, Evolutionary approach to setting up and training of artificial neural

networks/Neuroinformatics, vol. 1, №. 1, 2006

19. Chipiga A. F., Voronkin R. A. the Problem of convergence of the majority of the genetic

algorithm/Artificial intelligence, 2005

20. Shapovalov T.S., The scheduling of tasks in distributed computing systems using genetic

algorithms/Russian Academy of Sciences Fareast the computing center, -Habarovsk, 2010

21. Shapovalov T.S., Peresvetov V.V. Genetic algorithm scheduling for distributed heterogeneous

computing systems/computing methods and programming, -Taganrog, 2009

References in Roman script:

1. Averchenkov v. I., Kazakov P. S. evolyutsionnoe modelirovanie i ego primenenie / 2-e izd., -

M. : Flinta, 2011

2. Babin, D. S., voronoy S. M., “geneticheskiy algoritm postroeniya raspisaniy dlya

mnogoprotsessornyh vychislitel’nyh sistem”,Donetskiy gosudarstvennyy institut

iskusstvennogo intellekta, Ukraina “iskusstvennogo intellekta”,2005

3. Bukatov A. A., S. N. Datsyuk, A. I., Al’fons Programmirovanie mnogoprotsessornyh

vychislitel’nyh sistem/Rostov-na-Donu: OOO “KVR”, 2003

4. Volkov D. yavlenie bol’shih idey/otkrytye sistemy №. 8, 2013 www.osmag.ru

5. Gorbunov V., Elizarov,, Eysymont E. I. L. Ectatomminae superkomp’yuterov dostizheniya i

perspektivy/otkrytye sistemy № 7, v 2013, www.osmag.ru

6. Efimov, A. S., S. N. Samoylenko, e. N. Paraskova Organizatsiya funktsionirovaniya

raspredelennyh vychislitel’nyh sistem pri obrabotke naborov masshtabiruemyh zadach/

Vestnik Tomskogo gosudarstvennogo universiteta Upravlenie, vychislitel’naya tehnika i

informatika № 2(15), 2011

7. Jaspal Singh, Singh Harsharanpal, effektivnogo planirovaniya zadach dlya geterogennyh

mnogoprotsessornyh ispol’zovaniem geneticheskogo algoritma s uzla dublirovaniya”,

Indiyskiy zhurnal Informatika i vychislitel’naya tehnika Izd. 2 No 3 Iyun-Iyul 2011.

8. Kalyaev A. I. Ob odnom metode mul’tiagentnoy organizatsii oblachnyh seti komp’yuterov

chastnyh pol’zovateley - Taganrog: NII mnogoprotsessornyh vychislitel’nyh sistem Yuzhnogo

federal’nogo universiteta, 2012

9. Kazunori Kodzima, Masaaki Ishigame, Goutam Chakraborti, Hiroshi Hatsuo, i Shozo Makino

“asinhronnyh parallel’nyh raspredelennyh geneticheskiy algoritm s elitnym migratsii”

Mezhdunarodnyy zhurnal informatsii i matematicheskih nauk 4:2 2008.

10. Medvedev V. I. osobennosti ob#ektno-orientirovannoe programmirovanie v s++/CLI, C# i

Java/ 2-e izd., - Kazan’: RITs “shkola”, 2010

11. Portal iskusstvennogo intellekta, metody selektsii v geneticheskih algoritmah /http://www.

aiportal.ru/articles/genetic-algorithms/methodsselection.html

12. Radchenko, I. raspredelennyh vychislitel’nyh sistem, Chelyabinsk: fotohudozhnik, 2012

13. Rudenko Yu. M. raspredelenie programmnyh moduley po uzlam vychislitel’noy seti s obshim

polem pamyati dlya graf-shem parallel’nyh algoritmov/Nauka i obrazovanie № 10, 2011

14. Sal’nikov A. N. Sistema proektirovaniya i podderzhki ispolneniya parallel’nyh programm/

dissertatsiya na soiskanie uchenoy stepeni kandidata fiziko-matematicheskih nauk, -M, 2006

15. Sizov V. A. proektirovanie programmnyh i informationware kompleksa svyazannyh zadach v

seti EVM/Avtomatizirovannye sistemy upravleniya, 1995

16. Ural’skiy B. N. Sizov V. A. Razrabotka modeley i algoritmov optimizatsii raspisaniy vypolneniya

programmnyh moduley v komp’yuternoy seti na osnove evolyutsionnogo podhoda/ materialy

III mezhdunarodnoy nauchno-prakticheskoy konferentsii “Problemy ustoychivosti i

bezopasnosti sistem v sfere zhilishno-kommunal’nogo hozyaystva”, 2012

17. Hohlova, M. N. teoriya evolyutsionnogo modelirovaniya/”Tsefey”, 2013

18. Pou Yu. R., Spitsyn V., evolyutsionnyy podhod k nastroyke i obucheniyu iskusstvennyh

neyronnyh setey/Neyroinformatika, tom. 1, №. 1, 2006

19. Chipiga A. F., Voronkin R. A. problema shodimosti mazhoritarnogo geneticheskogo algoritma/

iskusstvennyy intellekt, 2005

20. Shapovalov T. S., planirovaniya zadaniy v raspredelennyh vychislitel’nyh sistemah s

ispol’zovaniem geneticheskih algoritmov/Rossiyskoy Akademii nauk Dal’nevostochnyy VTs-

Habarovsk, 2010

21. Shapovalov T. S., Peresvetov v. v. geneticheskiy algoritm planirovaniya dlya raspredelennyh

geterogennyh vychislitel’nyh sistem/vychislitel’nye metody i programmirovanie, -Taganrog,

2009

Содержание статьи/Article:

The need for solving large-scale problems, the development of cloud technologies and

working with big data lead to the rapid development of distributed models provide computing

resources [12, 6, 8]. At present, the list of “Top 500” includes computer systems of various

configurations, but the experience of the last ten years has shown that for effective solution

of practical problems of superscalar cores type, light or heavy, is clearly not enough - need

different kernel architecture and functionality of the so-called hybridity. Property of hybridity

means heterogeneity at the level of processor cores. At a higher level there is another type of

heterogeneity - heterogeneity, when the composition of the supercomputer are whole segments

of specialized processors and networks that are optimized for any task or class of tasks [5].

Under forecasts occurrence in the near future postmurovsky element base and new

principles of construction of supercomputers will lead to creation of specialized heterogeneous

supercomputers zettaflops level in 2020, and jotaflops - after 2024.

The added complexity of computing systems will continue, it will manifest itself in a

complex hierarchy of communications and memory strengthening of hybridity/heterogeneity

108 109

CONTEMPORARY PROBLEMS OF SOCIAL WORK VOLUME 1, No. 2, 2015

and dramatically increasing management complexity. Therefore, extremely important to

search for new methods, algorithms and ways of managing computational resources [5]. To

date, developers have the task not so much build computer configuration, when run on which

performance specially created test programs would exceed the threshold of one of exaflops,

how many maximum effective execution mode real-world applications that involve all allocated

resources of a supercomputer [4].

The desire to achieve greater productivity and efficiency, without changing the existing

programming model [4, 3], require the use of effective systems of parallelization. In this

regard, the urgent task today is the synthesis of the optimal logical structure of the complex of

information-dependent tasks (IDT) for distributed systems [15, 13].

Optimization model the logical structure of the complex of information-related tasks in the

computer network are developed based on the measure of the feasibility of parallelizing IDT in

the computer network (sun) - γ. When γ > 1 the main task is to synthesize the optimal logical

structure of the complex IDT on the criterion of the minimum solution time (F1 ).

When γ ≤ 1 from the viewpoint of reducing the solution time of complexants the effect of

parallelization in computer networks (CN) is negative. Therefore, as a criterion, it is advisable

to use the minimum total in-process interface (F2).

Thus, the overall pattern of the optimal design of the logical structure of the complex IDT

can be represented in the form of the algorithm, depicted in figure 1. As the initial information

used characteristics of the sun, and the canonical structure of the complex IDT in multilevel

parallel form. Task F1* mainly differs from the tasks F1 that takes into account the time spent

on the protocols. Problem solving techniques are discussed in the next Chapter. Meaningful

formulation F1(F1*) is formulated as follows. On the known characteristics of nodes VS, data

channels, topology, sun, canonical structure of a complex IDT, you must define the logical

structure of complex IDT in the form of sets of operating modules (OM) and relations between

them, as well as placement of DG in the armed forces, which would ensure the minimum time

solution IDT and implementation of network, system and structural constraints.

Task F1 is the following.

Find:

{ } { } ( ) ( )

1 1 1

min max nm

r

N R L

nm r c nm с l з nm з l o

m r m rm l m l m пр n r l

t t t

E E t = = =

⎡ ⎤

⎢ + + φ ⋅ + φ ⋅ + ⎥ +

⎣ ⎦

Σ Σ τ Σ

1

' '

' '

1 1 ' 1 '1 1

N M N M L

B nmBX nm nm c

l l l n m пр

n m n n m l

t t



= = =+ = =

⎡ ⎤

+ ⎢ Ψ Ψ + φ ⎥

⎣ ⎦

Σ Σ Σ Σ Σ (1)

at limitations

– unambiguous allocation procedures on modules

1 1

1, , 1, ,

N M

nm

r

n m

E rr R

= =

Σ Σ = ∀ =

1

'

1 1 1

0, , 1, ,

L M M

nm nm

l l

l m mm

n n N



= = = +

Σ Σ Σ Ψ Ψ = ∀ = (2)

'

1

0, , ', , 1, , ' ;

L

BX nm B n m

l l

l

n n m m M n n

=

Σ Ψ Ψ = ∀ = >

– on uneven loading of the nodes of the CN

( ) ( )

1 1 1 1

1/ 1/

M N M N

nm nm nm

m n m n

M T T M T

= = = =

Σ Σ − σ ≤ ≤ Σ Σ + σ , (3)

where σ - is the permissible deviation from the average runtime of DG in the CN;

End

Ключевые слова/Tags1: Modular distributed data processing system. Genetic algorithms. Fitness function. Optimization of the computational process. The soft ware structure.