Speaker : | Marco Scarpa |
University of Messina | |
Date: | 17/06/2014 |
Time: | 10:30 am - 11:45 am |
Location: | LINCS Meeting Room 40 |
Abstract
Quantitative evaluation plays a key role in the design of a broad range of systems, where functional and non-functional properties have to be carefully analysed in order to meet the requirements. In this talk, an analytical framework that enables the study of a class of phenomena where different working conditions alternate by changing the stochastic behaviour of the observed system but still preserving the continuity of the quantities to investigate is presented. The proposed solution technique is based on phase type distributions and on an ad-hoc fitting algorithm.
It allows to derive important metrics able to characterize the transient behaviour of the system and it is shown the impact of such a feature on the evaluated metrics with respect more simplistic (and classical) assumptions. On the other hand, the solution technique allows to extend the semantic of some modelling formalism, such as Petri nets, to represent new timing classes. Different cross-domain examples are provided to show the usefulness and wide applicability of the framework.