Article 4218

Title of the article



Tokarev Andrey Nikolaevich, Master's degree student, Penza State University (40 Krasnaya street, Penza, Russia),

Index UDK





Background. The research deals with the study of distributed computing systems of cluster and cluster-metacomputer type without a constant link between nodes.
The subject of research is the computational process in distributed systems with unreliable nodes. The purpose of the study is to reduce the impact of administrative failures on the progress of the computation process in a distributed system, and to reduce the load on the network and on the compute nodes and, thus, increase productivity.
Materials and methods. Research and development of a strategy for increasing performance in distributed systems with unreliable nodes is performed using the theory of reliability, the theory of Markov random processes.
Results. The analysis of reliability factors affecting the performance of distributed computing systems is made. The classification of failures is proposed depending on hardware or administrative reasons. Existing systems that prevent errors in calculations due to failures of various kinds are considered. A strategy for improving performance in distributed computing systems with unreliable nodes is proposed, which makes it possible to reduce the effect of administrative failures on the process of computing in a distributed system, and reduce the load on the network and increase the overall system performance.
Conclusions. Comparison of descriptive capabilities of the two computer models of the collision process of the dummy with the windshield of the car proved that the second computer model satisfies the requirements of the established quality representations
of the process of deformation and fracture of laminated safety glass in modern passenger cars. Analysis of existing systems preventing computational errors and maintaining the correctness of the computational process in distributed computing systems led to the conclusion that it is possible to provide sufficient performance and fault tolerance of a system with unreliable nodes without the use of special hardware. For this, an appropriate strategy has been proposed.

Key words

distributed system, calculations, cluster, subtask, reliability, fault tolerance

 Download PDF

1. Programmnoe obespechenie s otkrytym iskhodnym kodom dlya organizatsii dobrovol'nykh raspredelennykh vychisleniy i raspredelennykh vychisleniy v seti [Open source software for organizing voluntary computing and distributed computing in a network]. Available at:
2. Tokarev A. N. Aktual'nye problemy sovremennoy nauki: tr. 1-go Mezhdunar. Foruma [Actual problems of modern science: proceedings of the 1st International Forum]. Samara, 2005, pp. 113–116.
3. Goryainov V. T., Zhuravlev A. G., Tikhonov V. I. Statisticheskaya radiotekhnika: primery i zadachi [Statistical radio engineering: examples and tasks]. Moscow: Sov. radio, 1980, 544 p.


Дата создания: 11.12.2018 13:12
Дата обновления: 17.12.2018 08:45