Abstract
For a long time the use of centralized supercomputers has been the standard way to perform CAE-calculations. In the past 6 years the power of decentralized compute ressources grew very fast, due to increasing CPU power of modern RISC architectures. The big challenge has been to use mixed environments consisting of supercomputers and decentralized workstation/server clusters. In CAE this problem has been solved using LSF as a workloadmanagement tool, and appropriate job-flow concepts to manage distributed filesystems and job chains. This led to a transparent distributed computing environment, where computing ressources are used whereever these ressources are available. In contrast to that, CAD departments did not yet use a bigger amount of compute power. The increasing growth of CAD-models and the need to generate smaller computing jobs (e.g. structure machanics) during the design process, makes it necessary to use the computepower of the local workstations as well as the power of some decentralized compute-servers. First steps are done in the integration of CAD specific jobs like plotjob processing and NC-program generation. So far CAE and CAD have been treated as different computing environments. In future the ressources of both areas will be shared and the powerful machines of an enterprise will be available for all engineers, that might want to use it. This has to be done carefully, because the interactive usage of ressources (e.g. CAD) must not be disturbed by any job of other users. In this contribution we want to show up some technical possibilities how to create mixed working environments with the support of LSF, the leading tool in workloadmanagement. LSF knows about all available ressources in the computing environment and manages jobs in the framework of the defined policies, like priorization, preemption, run-windows, using of high-power machines, fair share scheduling and so on.