Component aggregation for PEPA models: An approach based on approximate strong equivalence

Approximate aggregation for PEPA components involves the construction of a smaller component that approximates the behaviour of the original one. Such an approximation at the component level can be very efficient and it can also result in a considerable reduction of the state-space for the underlying continuous-time Markov chain. We propose an approximate PEPA component aggregation strategy that relies on an approximate form of strong equivalence. The notion of strong equivalence captures behavioural similarity between components of different size. This quality renders approximate strong equivalence appropriate as a criterion to aggregate the state-space of PEPA components. We compare our newly proposed approach with previous work on component aggregation, where only a part of the component behaviour has been used as a criterion for aggregation. Our method requires fewer assumptions regarding the form of the components, and is therefore readily applicable to a larger family of PEPA models.