Estimating Software Reliability Without Test Hours

Software reliability growth models [1] are used during testing to forecast reliability measures such as remaining defects, discovery rate, etc. The models require the user to collect the number of test hours since software failures are a function of usage hours, not calendar hours. The problem is that collecting the test hours between failures or failures over some number of test hours sounds easier than it is. This paper aims to present an alternative method that does not require the collection of usage time data. During software testing, the discovered defects will have varying levels of detectability, ranging from obvious to obscure. By tracking the detectability of the software defects, one can estimate the percentage of total defects found so far compared to the theoretical estimate of all inherent defects. The percentage of defects discovered (also known as Defect Removal Efficiency or DRE) is a measure of the overall stability of the software, which can be used to make a release decision and estimate other reliability figures of merit.