Reliability growth of fielded software

A reliability growth model is presented that translates quality measures to time-dependent MTBFs (mean times between failures). Software reliability growth stems from the detection and removal of faults that are latent in the code after delivery. The fielded reliability performance is measured over time; the `noise´ is smoothed; and a general form for the reliability growth profile is generated. The basic reliability growth profile is enhanced to include the impact of code updates, fault recurrence, and usage. It is concluded that the combination of theoretical analysis and actual field data presented should provide key guidance to make operational reliability decisions and determine the support analysis needs of fielded software