Software complexity factor in software reliability assessment

Abstract

A straightforward early-stage software reliability assessment method was proposed and applied to large-scale, software-intensive programs (M.L. Yin et al., 1998). In this method, software size (in KSLOC) was used as the primary complexity factor, due to the limitation of detailed information available during the early stage of a program. Now that failure data are available from those programs, our next step is to evaluate and refine the method so that better assessment can be provided for future programs. In this paper, the performance of the early-stage method is addressed by comparing with other failure-data-based models. Since our major concern in the original method is the use of software size as the only complexity factor, the complexity issue is probed. The conclusions of this study are two folded. (1) The performance of the early-stage method is compatible with that of other failure-data based models, and (2) The early-stage method can be improved by adding the consideration of a functional complexity indicator n, which is derived from the McCabe's cyclomatic complexity (Kan, SH, 2003, T.J. McCabe, 1976).