S. Mankefors-Christiernin and A. Boklund (Sweden)
Random testing, parallelization, MPI, efficiency increase, scalability
Software testing is a tedious and time consuming but ut terly important task. There are several approaches to cre ating fault free software, in this paper we focus on auto matization and effectivisation of random testing tech niques. The execution of traditional random tests is time consuming and when applying complex operational pro files the time increases further. Using the integral formu lation of random testing it on the other hand becomes pos sible to create a kind of generic operational profiles. Hence the tests only have to be run one time independent of the profile applied. In this paper we present the integral formulation and a theoretical/ mathematic background on how it applies to a parallel computing environment. Fur thermore, we present an experimental evaluation of a par allel random testing implementation. For test suites con taining operational profiles the test time decreases to 9% when running over 16 worker nodes, making random test ing into a practically viable testing technique.
Important Links:
Go Back
