Home News People Research Study Search

Institute for Computing Systems Architecture

Methods and algorithms for the performance analysis of large sums

Anne Beniot, IMAG, Grenoble.

Methods and algorithms for the performance analysis of large systems. Abstract: Continuous Time Markov Chains (CTMC) facilitate the performance analysis of dynamic systems in many areas of application. They are often used in a high level formalism in which a software package is employed to generate the state space and the infinitesimal generator of the CTMC, as well as to compute stationary and transient solutions. Several high-level formalisms have been proposed to help model very large and complex CTMCs in a compact and structured manner.

CTMCs which model real systems are usually huge and sophisticated algorithms are needed to handle them. Both the amount of available memory and the time taken to generate them and to compute their solutions need to be carefully analyzed.

In the first part of my talk, I shall survey different modeling techniques and compare different algorithms and techniques for the performance analysis of large systems. I shall motivate the choices made and the research directions taken.

In the second part, I shall focus on our particular research. We use the formalism of Stochastic Automata Networks (SANs) to model systems as a set of interconnected subcomponents, and develop some new algorithms in order to be able to handle large systems in this framework.

I shall present a technique to aggregate replicated components, which leads a substantial reduction in the size of the state space. Finally, I shall present numerical algorithms for SANs, with emphasis on the improvement that we propose for the standard numerical "shuffle algorithm" to decrease the amount of memory and computation time.


Home : Colloquium 

Please contact our webadmin with any comments or changes.
Unless explicitly stated otherwise, all material is copyright © The University of Edinburgh.