Sample sizes for system availability

Abstract

This paper presents a methodology for determining sample sizes for system availability under the assumption that the time between failures and the times to repair are independently and exponentially distributed. For specified "x-nines" availability and confidence levels, a table for sample sizes is generated for different failed and repair events. In order to demonstrate five nines or higher availability at a given confidence, the sensitivity in sample sizes in the tails of the distribution is explored. The proposed method is also utilized to compute the confidence interval estimates for higher nines of availability. For easy use, spread sheets are created in EXCEL and Staroffice to determine the sample size for any combination of higher availability and confidence level.