A multistage methodology catering to different failure modes for estimating the failure intensity of software: an empirical study

Managing and measuring software reliability is one of the important problems in the software industry. Software reliability measurement issue becomes complex where the software experiences multiple failure modes. In reality, the complexity involved in reliability computation is dependent upon several factors, among them the most relevant ones are related to the failure process and how the failure is corrected. A study conducted in a software company revealed that, while testing the software product, certain types of failures are corrected while the test program is in progress and the others are corrected only at the end of the test program. The company wanted to estimate the failure intensity and the overall mean time between failures, taking into account the realities of the business practice in vogue. A methodology was proposed to meet this requirement and was tested using the data provided by the company. The details of the method in terms of the steps involved and how the method was used are presented in this paper.