The Institute of Propulsion Technology uses a large PC cluster to perform complex numerical computations. The cluster, which is for the exclusive use of the institute, is maintained and administrated by T-Systems-SfR. It consists of 190 dual nodes (380 processors) with the latest 64bit AMD-Opteron technology. The nodes are linked with a fast Infiniband interconnect using the largest available Infiniband switch. As such the processors communicate with each other with the same speed. The system has a master node with 4 processors and 32 GB RAM. This is fail-safed by a second server (slave), which undertakes the tasks of the master in the case of breakdown. The hard disk space comprises 12TB in a SAN network. Additionally, an integrated pre- and post-processing server with 4 AMD-Dual-Core processors and 64 GB RAM has access to the data space. The pre- and post-processing server enables the large amounts of data typically generated in time-accurate multi-stage compressor and turbine simulations to be analysed efficiently.
Primarily, the CFD code TRACE, developed at the Institute of Propulsion Technology for the simulation of multi-stage turbomachinery flows, is run on the cluster. Alongside standard applications, TRACE is also often used on the cluster in the automatic optimization of compressor blades. Similarly, using commercial CFD programs, numerical simulations of the combustion process in aero-engine combustion chambers are also performed on the cluster.