SWZ  >  
Research Areas  >  
Distributed Simulation

Distributed Simulation

Numerical simulation of technical and scientific problems traditionally is one of the disciplines with the highest demand of computing power. Accordingly, such problems are processed on supercomputers with vector and parallel computing architecture. The most powerful parallel computers in the world have been installed in the US for simulation applications. The most powerful parallel computers in Germany together have only a fraction of the power of these machines and are concentrated at three sites in Munich (LRZ and RZG), Stuttgart (HLRS) and Jülich (FZ). In Niedersachsen there are HLRN II (North German Network for High-Performance Computing) and some smaller systems in university computer centers available for large simulation tasks. To get even more computing power, it is necessary to split the numerical problems so they can be calculated on the existing hardware infrastructure at universities and their data centers. Unfortunately generally the effort needed to create the software and middleware that sits between applications and operating systems is very high and there are enormous constraints on the applications. Therefore it has paid only in individual cases, to distribute simulations on multiple PC clusters so far. With the advent of Grid Computing is becoming apparent, that the cost of the distribution of simulation applications comes on existing PCs and workstations in an area where it will be interesting for many users.

In this project area we will examine models and methods with which simulations can be distributed on a grid and the resulting problems of software testing and quality assurance can be solved. "Distributed Simulation" also means the simulation of highly distributed real systems, such as Supply chains, supply chain management, which cannot be detected by a conventional, closed model.