An analysis of system level software availability during test

Examines some of the issues associated with measuring software availability during a system-level test. Based on a client's request, a project was asked to estimate the stability of a communications system under a heavy load. The software reliability data that was gathered measured the performance of three key communication types (call types) during the test. This paper analyzes the software failure data for these three types of calls and estimates the software failure rate and associated availability for each call type and for the system as a whole. The fact that the original product requirements did not address the issues of software reliability and software reliability growth at the outset of the project proved to have significant consequences. This paper also discusses some of the issues and lessons learned resulting from this decision, in particular the difficulties in data collection and analysis.