The several projects of the working group demand different and partly very high standards of the IT environment. These requirements are satisfied with up-to-date and powerful hardware. Particularly, workstations and servers from Dell® are used for this reason. Here are some examples of applications:
- Three dimensional point clouds generated by the terrestrial laser scanner comprise more than 30 million data points. Fluent visualization and post-processing of such datasets is done by using up-to-date workstations with newest Intel® Xeon® hexacore processors and high-end graphic boards by Nvidia®, as for instance the Quadro 4000.
- In addition, the build-up and processing of geological reservoir models inside the software Petrel® requires this sort of high-end graphic devices in respective workstations to display the huge amount of information.
- The calculation of geomechanical Finite Element (FE) – Models of entire reservoirs exceeds the computing power of a workstation due to the large number of elements (~ 4 million) and faults (>80). Such models are calculated with the FE – software Ansys® on the high-performance compute server of the working group. This server comprises two powerful Intel® Xeon® hexacore processors (X5690) in combination with 192GB main memory and six 15K SAS HDDs in a RAID5 assemblage. By utilizing so called high-performance computing (HPC) licenses and SMP parallelization, twelve cores can be used simultaneously and calculation times of less than 20 hours can be reached for the currently largest geomechanical FE models. The post-processing of these models demands a lot of main memory and is currently done on a workstation with 48GB RAM. In addition to the compute server, the working group has a file server with almost 6TB of accessible storage.
The big advantage of these in-house solutions is that they are completely subject to the working group and are thus always available according to internal agreements. If more computing power is temporarily needed, the Hessian High-Performance Computing Cluster (HHLR) can be accessed. A new version of this cluster is built from January 2013 on by IBM® and should provide in the first stage in June 2013 more than 700 computing nodes.