This is an experience report on the application of software reliability engineering (SRE) in developing the Data Quality System (DQS). DQS is a knowledge-based software system developed by Lucent Technologies. It was developed to synchronize data spread among several disparate databases. The first application of DQS was to telecommunications network databases. Because the development of DQS was a large custom software development project, SRE was employed to assist in measuring product quality and readiness for production use. In this report, we first describe what we wanted to accomplish in applying SRE methods in the DQS project. The AT&T best current practice outlined the prerequisites for applying SRE during testing. We describe how we addressed these prerequisites in the DQS project. In particular, we discuss specifics on: what assumptions were made regarding an operational profile; how test cases were selected to conform to operational profile usage; and how test log forms were designed and used to collect failure data and to manage the test effort itself. Next, we share the results of reliability growth modeling during the system testing and field trials. Finally, we highlight what we learned in our initial implementation of SRE and what plans were recommended for subsequent implementations.

This panel session will discuss how Systems Engineers and Software Architects can benefit porn using SRE. Members of the panel will discuss how their respective organizations use (or anticipate using) SRE earlier in the software development life-cycle. The intent is to stimulate and encourage those involved in Systems Engineering and Software Architecture to adopt and use SRE in their work.

The use of software reliability growth models has become popular. However, the use of most such models is delayed until late in the software product life-cycle (i.e. during testing) as they require failure time data to determine the model parameters. Currently, there is only one model, the basic execution time (BET) model that can be applied earlier in the product life-cycle as its parameters can be related directly to software characteristics. The derivation of the BET model implies uniform execution of program instructions. However, branches and loops in software programs results in non-uniform execution of program instructions. This paper introduces an extension of the BET model, the extended execution time (EET) model that takes into account non-uniform execution of instructions. The non-uniform execution is accounted for by a parameter in the EET model that can be related to measurable characteristics of the software product. This paper further shows the EET model reduces to the BET when instruction execution is uniform and has behavior similar to the Musa-Okumoto logarithmic model in the limit as instruction execution becomes highly non-uniform.>

The measurement and analysis aspects of software reliability are described with the aim of providing software engineers and managers a sense of where and how software reliability measurements can be applied to their projects. Some background for understanding software reliability measurement is provided, and activities associated with measuring and analyzing software reliability are discussed in the context of the software product life cycle. The focus is on failures and the rate at which they occur. The thrust of reliability measurement and analysis is to specify what the customer needs in terms of reliability before a software product is built, to validate that these needs are met before delivery of the product to the customer, and to make sure that the customer's needs continue to be met after delivery.>