Approach to a Simulation Virtual Machine: Object Oriented Implementation of CDEVS and PDEVS

Abstract

In the world today, the size of complex problems and systems has increased, due to the advances in technology and the area of computer design and architecture. Modeling and predicting the behaviour of complex systems (Weather forecast, Fire spreading, River floods, Earthquake, Nanotechnology; design of new materials from molecule scale, decoding the human genome and many others) is becoming complex due to huge amount of data, high statistics need, and the need to serve user communities around the world. Therefore there is a need of exploiting the computing power of nowadays technologies by distributing simulation on multiple processors in order to reduce execution time, perform real time execution, and integrate simulators.

In this work we present an approach to Discrete Event System Specification (DEVS) virtual machine that will take a DEVS model and maps its simulation onto any hardware host like a LAN, a WAN, a Grid, a Cluster, the Internet and so on. The virtual machine is structured in 3 layers; the modeling layer which receives the DEVS model, the simulation layer which simulates the model using either the pessimistic synchronization algorithm or the optimistic synchronization algorithm, and the last layer which is the Middleware layer that allows the mapping of the simulation onto any hardware host. The kernel of the virtual machine contains a CDEVS implementation of the simulator, PDEVS implementation of the simulator, and the distributed versions of them. Starting with an existing CDEVS simulator we got its PDEVS implementation using a meta modeling approach. And we finally provide the multilayer simulation package.