Testing component -based distributed applications

Sudipto Ghosh, Purdue University

Abstract

Applications that utilize the broker-based architecture are often composed of several components that need to be tested both separately and together. An important activity during testing is the assessment of the adequacy of test sets. Testers use one or more adequacy criteria for this activity. Traditional test adequacy criteria have several limitations for commercial use. Therefore, a set of scalable interface-based test adequacy criteria have been identified and a testing method that uses these criteria proposed. This method incorporates elements from a component's interface description for computing coverage using multiple test adequacy criteria and performing interface mutation and fault-injection testing. A prototype testing and monitoring tool that implements the proposed method is described. The proposed interface-based test adequacy criteria were evaluated empirically and compared with control flow-based coverage criteria for their relative effectiveness an revealing errors and the cost incurred in developing the test sets. A formal analysis of the fault-detection ability of the testing methods was also carried out. In addition to the assessment of test adequacy, testers often need to assess the tolerance of an application to failures in its components or in the environment. The interface-based testing method has been extended to allow for fault-injection testing at the component interfaces. Experimental results are reported for fault-injection testing on a client-server system. A generic set of faults is identified for use while testing broker-based systems.

Degree

Ph.D.

Advisors

Mathur, Purdue University.

Subject Area

Computer science

COinS