Modelling and analysis of distributed simulation protocols with distributed graph transformation

Abstract

This paper presents our approach to model distributed discrete event simulation systems in the framework of distributed graph transformation. We use distributed typed attributed graph transformation to describe a conservative simulation protocol. We use local control flows for rule execution in each process, as the use of a global control would imply a completely synchronized evolution of all processes. These are specified by a Statechart in which transitions are labelled with rule executions. States are encoded as process attributes, in such a way that rules are only applicable if the process is in a particular state. For the analysis, we introduce a flattening construction as a functor from distributed to normal graphs. Global consistency conditions can be defined for normal graphs which specify safety properties for the protocol. Once the flattening construction is applied to each rule, the global conditions can then be translated into pre-conditions for the protocol rules, which ensure that the protocol fulfils the global constraints in any possible execution. Finally, the paper also discusses tool support using the AToM/sup 3/ environment.