Visualisation for stochastic process algebras: the graphic truth
EPEW'11 Proceedings of the 8th European conference on Computer Performance Engineering
Compositional approximate markov chain aggregation for PEPA models
EPEW'12 Proceedings of the 9th European conference on Computer Performance Engineering
Compositional approximate markov chain aggregation for PEPA models
EPEW'12 Proceedings of the 9th European conference on Computer Performance Engineering
| Hi-index | 0.00 |
When analysing the performance of a system, we are often interested in long-run properties, such as the proportion of time it spends in a certain state. Stochastic process algebras help us to answer this sort of question by building a compositional model of the system, and using tools to analyse its underlying Markov chain. However, compositionality in the model leads to a state space explosion in the Markov chain, which severely limits the size of models we can analyse. Because of this, we look for abstraction techniques that allow us to analyse a smaller model that safely bounds the properties of the original. In this paper, we present an approach to bounding long-run properties of models in the stochastic process algebra PEPA. We use a method called stochastic bounds to build upper and lower bounds of the underlying Markov chain that are lump able, and therefore can be reduced in size. Importantly, we do this compositionally, so that we bound each component of the model separately, and compose these to obtain a bound for the entire model. We present an algorithm for this, based on extending the algorithm by Fourneau et al to deal with partially-ordered state spaces. Finally, we present some results from our implementation, which forms part of the PEPA plug-in for Eclipse. We compare the precision and state space reduction with results obtained by computing long-run averages on a CTMDP-based abstraction.