Workshop On Dependability Benchmarking 2002

Assessing the quality of service of a computer system is a difficult task. A lot of work has been conducted on evaluating quality of service attributes like performance, robustness, and dependability. Two approaches used for evaluating performance and robustness are modeling and benchmarking. For evaluating dependability, modeling can be used either alone or combined with fault injection [Sie92, Kan91]. However, less work has been conducted on building dependability benchmarks. A dependability benchmark can be defined as “a way to evaluate the behavior of components and computer systems in the presence of upsets, allowing the quantification of dependability attributes or the characterization of the systems into well defined dependability classes” [Mad01]. This paper focuses on the quantification part of the definition. The goal of this paper is to propose the use of Bayesian estimation methods for quantifying dependability attributes. We first will give a brief overview of two estimation Schools in Section 2. We will then illustrate our proposal by focusing on a key parameter for fault-tolerant systems, the coverage factor. We will introduce the coverage factor in Section 3 and present coverage factor estimations in Section 4.