1 / 9

Toward more realistic software reliability predictions: A family of empirical studies

Toward more realistic software reliability predictions: A family of empirical studies. PI: Katerina Goseva – Popstojanova Students: Margaret Hamill & Ranganath Perugupalli Lane Dept. Computer Science and Electrical Engineering West Virginia University, Morgantown, WV

jills
Download Presentation

Toward more realistic software reliability predictions: A family of empirical studies

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Toward more realistic software reliability predictions: A family of empirical studies PI: Katerina Goseva – Popstojanova Students: Margaret Hamill & Ranganath Perugupalli Lane Dept. Computer Science and Electrical Engineering West Virginia University, Morgantown, WV katerina@csee.wvu.edu

  2. Problem • Problem: There are many open questions with respect to the realism of the underlying assumptions, accuracy, and applicability of software reliability models • Our goal: Contribute towards more realistic assessment and prediction of software reliability based on theoretical and empirical studies • Two important phenomena will be addressed • Uncertainty in software reliability due to errors in the operational profile • Effect of failure clustering on software reliability predictions

  3. Relevance to NASA

  4. Importance & Benefits • Develop new more accurate software reliability theory • Apply and validate theoretical results on large real-life empirical studies • 300 source files • 800,000lines of C code What are the parts of the system that need more extensive V&V? Identification of critical components -Frequently used components - Components that heavily affect the application - Components with low reliabilities

  5. Intended Approach historical data, UML Informed Approach component traces 1 p12 2 p23 p1E 3 p2E Build software architecture & Estimate operational profile E Conduct uncertainty analysis 1-R1 1 p12R1 1-R2 F 2 Entropy Perturbation analysis Method of moments Monte Carlo simulation p23R2 p1ER1 3 Growth models p2ER2 E Estimate components reliabilities & Build software reliability model Failed & non-failed executions 1-R3 Fault injection Approach

  6. Approach Instrument the software with a profiler Get test cases that failed Execution of test cases Version 3.2 test logs & change logs Bug tracking database Bugzilla Execute tests on newer versions & check logs Profiles for each test case Relational database of profiles for each test case Identify defect location for each test case that failed Mapping C functions to files; Mapping files to components Estimate components reliability Estimate transition probabilities & build operational profile Software reliability model Uncertainty analysis

  7. Accomplishments • Currently we are experimenting with theC properpart of GCC version 3.2.3 • 300 source files • 800,000 lines of C code • We are using 2426 test cases from the test suite of GCC version 3.3.3 • This is the largest application used for empirical software reliability estimation ever

  8. 1 2 • Analyzed the uncertainty based on entropy • Identified most frequently used components and components that affect large parts of the application START 3 15 4 14 17 5 13 6 12 END 7 Accomplishments • Identified location of the defects for the failed test cases using three different methods (two automatic and one manual) • Built the operational profile on component level based on execution profiles which contain over 2,000,000 C functions

  9. Next steps • Complete the experiments with open source software GCC • Apply and validate the methods for uncertainty analysis • Lessons learned • Conduct experiments with NASA software • Address the failure clustering phenomenon – theoretically and empirically

More Related