350 rub
Journal Information-measuring and Control Systems №2 for 2011 г.
Article in number:
Comparative analysis of structure of the failure-safe duplicated computer complex
Authors:
I. Yu. Golubev, V. A. Bogatyrеv, V. F. Bezzubov
Abstract:
The purpose of this research is to compare the variants of construction of two-machine computer complexes (CC). It is spent at various approaches to an intermachine exchange aimed at ensuring the recovery computational process (RCP) after the occurrence of failures.
Two-machine CC are used as basic computational units in the highly distributed control systems. After accumulation of failures leading to loss of function of one of the computers, the whole stream of requests is distributed on a computer that has kept working. It is assumed that the query time in the system is critical, and in case of failure of one of the computers, the query can not be renewed without risk for the management process. Interrupted processing must be restored. The intercontrol of the computers and the mechanism of control points are to realize.
The structures of two-machine CC are considered, with the organization of interprocessor exchange via shared memory, without the realization of direct memory access (DMA) mode and with DMA, as well as an adapter connection between computers, using methods of the interface window and the so-called double DMA.
In the course of research for each of the structures are made reliability models. They are used for calculations and construction of curves of following dependencies:
- uptime of CC from the time of their operation;
- loss of productivity of CC, due to time consuming to ensure the RCP after the occurrence of failures, from the permissible time for RCP;
- performance of CC at the time of processing requests from the intensity of the receipt of requests.
These calculations show the advantage of organizing the relationship of computers of two-machine CC via adapters connection between computers. Decisions based on the use of shared memory, characterized by a low efficiency of query processing time and greater loss of productivity while reducing the allowable time of the RCP.
The construction of two-machine CC based on the adapters that implement the double DMA is recommended to ensure the lowest time of the RCP among the considered structures.
Pages: 8-12
References
- Половко А. М. Основы теории надежности. СПб.: БХВ-Петербург. 2006.
- Ларионов А. М., Майоров С. А., Новиков Г. И.Вычислительные комплекса системы и сети. Л.: Энергоатомиздат. 1987.
- Беззубов В. Ф. Сравнительный анализ методов обмена в многопроцессорных системах // Вестник компьютерных и информационных технологий. 2006. № 4. С. 51-56.
- Богатырев В. А. Комбинаторно-вероятностная оценка надежности и отказоустойчивости кластерных систем // Приборы и системы. Управление, контроль, диагностика. 2006. № 6. С. 21-24.
- Богатырев В. А. Оценка надежности и оптимальное резервирование кластерных компьютерных систем // Приборы и системы. Управление, контроль, диагностика. 2006. № 10. С. 18-21.
- Богатырев В. А. Оптимальное резервирование системы разнородных серверов // Приборы и системы. Управление, контроль, диагностика. 2007. № 12. С. 30-36.
- Богатырев В. А. Метод оценки надежности вычислительных систем при функциональной неоднородности компьютерных узлов // Приборы и системы. Управление, контроль, диагностика. 2006. № 12. С. 20-22.
- Патент СССР. № 1679493 G 06 F 13/00. Устройство для сопряжения ведущей и ведомой ЭВМ / В. Ф. Беззубови др.
- А. с. СССР. № 1462341 G 06 F 15/16. Устройство для сопряжения ЭВМ. / В. Ф. Беззубов.
- А. с. СССР. № 1798946 Н 05 К 10/00, G 06 F11/20. Резервированная вычислительная система / В. Ф. Беззубови др.