Research Results | Toward More Realistic Software Reliability Predictions


	Search:	Advanced Search

Software Architecture Assessment < Verification & Validation < Results Home

Click here to complete a short survey. The results of this survey will be used to help us improve the research program and this website.

Click here to view research projects that had new research results added in the last 90 days.

Toward More Realistic Software Reliability Predictions
Point of Contact	Katerina Goseva-Popstajanova Katerina.Goseva@mail.wvu.edu
Dates	October 2003 - June 2007
Problem	This project is focused on development of more realistic and accurate estimation and prediction of software reliability based on empirical studies. We address two important phenomena: uncertainty in software reliability due to errors in the operational profile and the effect of failure clustering on software reliability predictions. The proposed work on uncertainty in software reliability builds on our recent contributions on methods for uncertainty analysis developed in FY02 and FY03. We propose to run empirical case studies aimed at verification and validation of our methodology for uncertainty analysis and comparison of different methods based on real data. The same empirical studies will be used to build more realistic software reliability models that consider the correlation of successive software executions and analyze the effects of failure clustering on software reliability predictions.
Objective	The aim of this proposal is to develop more realistic and accurate predictions of software reliability basedon theoretical research and empirical case studies. In FY04 we will design and run a family of empirical case studies that will provide relevant data for studying two important problems in software reliability: uncertainty analysis and failure clustering. Our next objective is to continue our previous work on uncertainty analysis in software reliability due to errors in the estimation of the operational profile and component reliabilities. In this project we will apply and validate different methods for uncertainty analysis that we have developed during the FY02 and FY03 on empirical case studies and compare them using real data. In FY05 we will build architecture-based software reliability models based on empirical data obtained in FY04, apply different methods for uncertainty analysis and compare them with respect to several criteria. Our last objective is to focus on one of the weakest points in software reliability modeling, the assumption of independence among successive software failures which can be easily violated in practice. In FY05 we will develop methods for identifying the presence of failure clusters from the empirical data obtained from the experiments conducted in FY04 and develop improved software reliability models that consider failure clustering. In FY06 we will develop statistical inference procedures for modeling parameters, validate the new models on the empirical data, and study the effect of failure clustering on software validation and verification and software reliability.
Results	End-of-year briefing, NASA OSMA SAS 2004.ppt SAS 05 Executive Presentation.ppt Report on empirical cased studies.pdf SAS 06 Executive Presentation.ppt SAS 06 Technical Presentation.ppt SAS 05 Technical Presentation.ppt Report on Effect Failure Clustering.pdf
Keywords	software reliability, operational profile, uncertainty analysis, failure clustering, stochastic models, empirical case studies
Categories	Quality Control Software Reliability Software Architecture Assessment

Curator: Josh Stonestreet
NASA Official: Lisa Montgomery

NASA Privacy, Security, Notices