The PN-PEM framework: a Petri Net Based Parallel Execution Model

Abstract

This paper introduces the PN-PEM framework. It is based on the representation of an algorithm with Petri Nets. Frequently, a real algorithm needs a large Petri Net to be represented. We present a way to model an algorithm with Colored Petri Nets that simplify the model. After that, this high level model is transformed into a low level but executable model, preserving its semantics. The execution also needs other components of the framework, as the involved processors, data used and executable kernels. The combination of these elements is described in order to obtain a parallel execution. Some tests are also presented as a testbed of the framework in symmetric multiprocessors. Usability, as well as good performance, confirm the quality of the framework.