Song Kwang Suk, Yeo Hwan Geun, Han Chang Ho, Moon Tae Soo, You Choong Yeol, Lee Kwang Bae, Kim Hyen Ug, Yoon Chung Hwa

The Transactions of the Korea Information Processing Society (1994 ~ 2000), Vol. 2, No. 6, pp. 985-994, Nov. 1995
10.3745/KIPSTE.1995.2.6.985,   PDF Download:

Abstract

In this paper, we propose two hot standby architectures which not only provide high system availability but also lose little data on fault occurrence. In order to evaluate the performance of the proposed hot standby architectures, the warm standby architecture which is made from the hot standby architecture by eliminating its synchronization unit is considered. After system unavailability for each architecture is computed by using the corresponding Markov state diagram, the results are compared and evaluated. As the results, in most cases, hot standby architectures have higher availability than warm standby architecture. Also, hot standby architecture with external synchronization unit always maintains a little higher availability than hot standby architecture with internal synchronization unit. Active set time and personnel recovery rate for each architecture have little effect on system availability. However, in the case that data recovery time is too long, system availabilities of hot standby architectures and warm standby architecture degrade rapidly. In this case, the performance degradation of hot standby architectures is severe, and system availabilities of hot standby architectures eventually become lower than system availability of warm standby architecture.

Statistics
Cite this article