Comparisons of Inferences on Software Reliability by Aggregated and Separated Recording of Fault Counts

• Activity Number: 497
• Type: Topic Contributed
• Date/Time: Wednesday, August 3, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Section on Physical and Engineering Sciences
• Author(s): Shuen-Lin Jeng*+
• Companies: National Cheng Kung University
• Keywords: Aggregated Recording ; Critical Fault-Detecting Time ; Nonhomogeneous Discrete-Compound Poisson ; Separated Recording ; Software Reliability

Abstract

In this study, we show that the separated recording of fault counts of each system in a software fault-failure processes (SFFP) is better than the aggregated recording of fault counts of all systems (used by Jeske et al., 2005). The aggregated recording approach may lose the information of variability from system to system and is difficult to identify different failure modes of SFFPs. Including the typical reliability metrics such as the average number of failures in a time interval, this study also proposes an alternative reliability index, critical fault-detecting time (CFDT) in order to provide more information for managers on judgments for software releasing times. Through simulations, our study suggests that the separated recording approach is more appropriate since it can provide more accurate estimates of CFDTs. The conventional NHPP, however, may not be suitable for some realistic practices when each failure event is caused by two or more faults. For this issue, our work considers a more general and feasible model, i.e., nonhomogeneous discrete-compound Poisson (NHDCP) model, for the software fault-failure processes.