Business Solutions

• IT Solutions
• IT Consulting
• IT Security
• Technical Resourcing
• Engineering Consulting
• Research and Development

Development of Methods for Minimal Test Pattern Generation (MTPG)

The dominance of software technology has dramatically increased in almost all aspects of our lives, and the trend is accelerating. Software systems are often very large, complex and powerful and their failure can have diverse effects ranging from personal inconvenience to catastrophic disaster. Yet, it is a widely known fact that many faults escape the testing phase and are discovered in the field, often causing customer dissatisfaction, high field maintenance cost, and perception of poor product quality.

Software testing is expensive, tedious and time consuming. Several recent studies indicated that testing consumes about 30 to 70 percent of an organization's software development resources. At the same time, software testing is critical for weeding out potential faults and for insuring that the system works as intended. To improve customer satisfaction and reduce development costs, it is important that software teams reduce testing costs without loosing its effectiveness, and send fewer faults to the field.

How can we reduce the testing cost and still find all the faults before the software is released? One way of improving testing effectiveness is by automating the testing activity so that, one software program tests another program and collects the results. Another way is to wisely determine which tests should be run.

This research project is focusing on developing methods and techniques to support the generation of minimal number of test patterns. These methods take a black-box approach to testing i.e. the information they need is independent of the implementation. The approach we have proposed use ideas from combinatorial design theory to minimize the number of tests needed for a specific level of test coverage of the input test space. The actual construction is based on test generation algorithms. In exhaustive testing the number of test cases needed is extremely large- almost always too large to be practical. Our method tries to find that for n-input parameters around n-tests would suffice. In some cases (especially with larger set of inputs to be tested) the minimum number of test cases required may be slightly more than test-inputs but certainly be much less than exhaustive testing. Thus savings due to proposed MTPG method is substantial when compared to the exhaustive testing.

Email to someone who can help.