Real-life systems are often plagued by unanticipated performance problems caused by subtle bugs and bottlenecks. It is thus essential for system designers and engineers to have an understanding of their fundamental performance characteristics, both before and after implementation. Stochastic modelling and analysis respectively provide the means to abstract systems as mathematical descriptions and to derive quantifiable measures of interest from them.
A major, and so far largely unaddressed, challenge is the specification of complex performance queries on models in an accessible manner that does not sacrifice expressiveness. This thesis attempts to address this challenge by introducing Performance Trees, a new formalism for the graphical specification of complex performance queries on stochastic models. Performance Trees are designed to be accessible by providing a more intuitive approach to query specification, expressive by being able to reason about a far broader range of concepts than current alternatives, extensible by supporting additional user-defined concepts, and versatile through their applicability to multiple modelling formalisms. Performance Trees are presented in the context of a rigorous formal framework that defines the syntax, typing and quantitative semantics of operators.
Prototype tool support is implemented in the form of a module of the PIPE2 Petri net tool, which provides graphical user interfacing and Performance Tree query design capabilities. Query evaluation is supported by a set of integrated parallel and distributed analysis tools, and realised by the distribution of computations onto a dedicated Grid cluster.
The practical application of Performance Trees is demonstrated in the context of case study analysis scenarios of an electronic voting system, an online transaction system and a hospital's Accident & Emergency unit.
Information from pubs.doc.ic.ac.uk/suto-thesis.