Simulating Concurrent Behaviors with Worst-Case Cost Bounds

Abstract

Modern software systems are increasingly being developed for deployment on a range of architectures. For this purpose, it is interesting to capture aspects of low-level deployment concerns in high-level modeling languages. In this paper, an executable object-oriented modeling language is extended with resource-restricted deployment components. To analyze model behavior a formal methodology is proposed to assess resource consumption, which balances the scalability of the method and the reliability of the obtained results. The approach applies to a general notion of resource, including traditional cost measures (e.g., time, memory) as well as concurrency-related measures (e.g., requests to a server, spawned tasks). The main idea of our approach is to combine reliable (but expensive) worst-case cost analysis of statically predictable parts of the model with fast (but inherently incomplete) simulations of the concurrent aspects in order to avoid the state-space explosion. The approach is illustrated by the analysis of memory consumption.