CASRE (Computer Aided Software Reliability Estimation) was developed as a software reliability measurement tool that is easier for nonspecialists in software reliability engineering to use than many other currently-available tools. CASRE incorporates the mathematical modeling capabilities of the public domain tool SMERFS (Statistical Modeling and Estimation of Reliability Functions for Software), and runs in a Microsoft Windows environment.
The command interface is menu driven; enabling and disabling of menu options guides users through the selection of a set of failure data, execution of a model, and analysis of model results. Input to the models is simultaneously displayed as text and as a high-resolution display that can be controlled to let users view the data in several different ways (e.g., time between successive failures, cumulative number of failures). Model predictions and statistical evaluations of a model's applicability (e.g., prequential likelihood ratio, model bias, bias trend) may be superimposed on the plot of the data used as input to the model. CASRE also incorporates earlier findings - that prediction accuracy may be increased by combining the results of several models in a linear fashion. Users can define their own model combinations, store them as part of the tool's configuration, and execute them in the same way as any other model.
What's new in Version 3.0?
Version 3.0 includes two trend tests that may be used to determine whether it is even appropriate to apply a software reliability model to a set of failure data. These two tests, the running arithmetic average and the Laplace test, allow users to determine whether a set of failure data indicates that the system's reliability is increasing during test, whether it is decreasing, or whether there is no discernable trend. If the failure data shows decreasing reliability or no discernable trend, applying reliability models to the data is not indicated. Version 3.0 also has a simpler input format than previous versions - for input data entered as the number of failures in a test interval of a specified length, eliminating a rarely-used model allowed the removal of three fields specific to that model. However, version 3.0 will still read input data formatted for previous versions. Finally, version 3.0 simplifies model selection to a certain extent. Several models have different variants - for instance, the Schneidewind model has three variants which treat the failure data in different ways. Previous versions of CASRE would let users select the model, and then force them to select one of the variants - for example, if a user wanted to run the Schneidewind model, they would first select that model, and then select one of the three variants. This made the selection mechanism more complicated, and allowed users to run only one model variant at a time. Version 3.0 treats each model variant as a separate model - users now simply select all of a model's variants they want to run. This makes the selection mechanism somewhat simpler and more consistent, and users are now able to simultaneously run all variants of a model.
This tool should be useful to software development organizations searching for ways to more effectively manage their development resources. Since CASRE has been designed with the non-specialist in mind, it may be easier for managers and developers to use than those tools requiring detailed knowledge of the models.