Project Title:

Quantifying Confidence in the Correctness of Parallel or Distributed Software

92-1-06.03-1219                                                   NAS01-19896
Quantifying Confidence in the Correctness of
     Parallel or Distributed Software
Reliable Software Technologies Corporation
1001 North Highland Street, Penthouse Suite
Arlington, VA  22201
Jeffrey M. Voas                                      (703-276-1219)

Abstract:

     This project will develop a new technique for assessing the
testability of parallel/distributed software systems. Since many
critical software systems are being considered for deployment in
parallel or distributed environments, and since the need for
extremely high levels of reliability exists, there must be a
feasible means of assessing reliability. Non-exhaustive, random
black-box software testing is frequently used for demonstrating
system reliability but is generally infeasible for high levels of
reliability. This project addresses whether particular software
systems can be tested to high levels of reliability by predicting
whether the testing has missed hidden faults. This is an improved
technique for assessing how confident a developer is in a
software's reliability. The project's objective is to augment a
software testability model created for sequential software systems
and apply it to parallel or distributed systems. Anticipated
results are a complete model for assessing the testability of
parallel or distributed systems.

Potential Commercial Application:

Potential Commercial Applications:  The benefit of testability
analysis, both for sequential and parallel or distributed systems,
is its power to reveal whether even thoroughly tested software is
still hiding faults although testing has suggested the contrary.
This analysis will benefit the commercial development of software
by reducing the likelihood of undetected faults.

                                     ***