Software dependability evaluation based on Markov usage models |
| |
Authors: | Walter [Reference to J. Gutjahr] |
| |
Affiliation: | Department of Statistics, Operations Research and Computer Science, University of Vienna, 1010 Vienna, Austria |
| |
Abstract: | A general technique for computing optimal state transition probabilities for software tests, based on a Markov usage model, is developed. The optimization criterion is maximum precision of unbiased dependability estimates derived from the test results. Three different dependability measures are considered: (i) risk, (ii) safety, and (iii) reliability. As input, pre-information on failure probabilities and losses in case of failure related with single operations is used. The optimization itself is done by means of a numerical procedure which is fast because of the convexity of the underlying stochastic optimization problem. The procedure can be improved by the construction of a distribution with a common lower bound on state transition probabilities; this distribution may also be used in the more general context of structural statistical testing of software. |
| |
Keywords: | Dependability Importance sampling Markov chains Optimization Reliability Risk Safety Software usage models Software testing |
本文献已被 ScienceDirect 等数据库收录! |
|