Grid-based cooperation in aerospace research

The objective of the AeroGrid project is to provide an efficient, grid-based working environment for aerospace research in Germany. To this end, a long-term, practice-oriented grid infrastructure is being developed for cooperation between national research institutions, industry and university research. The focus is on real-world usability and transferability to similar networks.

Unlike other industry sectors, a very large proportion of aerospace research is carried out in public institutions and facilities. Thus, the industry uses direct codes with innovative algorithms in the area of numerical computational fluid dynamics (CFD). These algorithms are jointly developed by research institutions, such as the German Aerospace Center (DLR), and universities. The DLR Institute of Propulsion Technology, the Institute of Jet Propulsion of the University of German Armed Forces, Munich, and the engine manufacturer MTU Aero Engines are involved in AeroGrid as developers and/or users.

The AeroGrid working environment will facilitate flexible cooperation in virtual organisations according to project requirements as well as the use of the latest program versions, data and computers on a cross-location basis. It will also ensure that the origins of calculation results can be traced in detail. The DLR Simulation and Software Technology Facility and T-Systems Solutions for Research (SfR) are responsible for developing the grid infrastructure and integrating the grid interfaces in the working environments of the application partners. T¬-Systems SfR will act as the service provider after the project has reached completion in order to ensure that the infrastructure continues to operate productively.

In the project, the grid environment will be evaluated using typical application scenarios from turbine simulations. In particular, this includes the use of computer resources via the AeroGrid, cooperation in the design of engine components and the cooperative further development of the DLR CFD code TRACE.

Actual evaluation scenarios in the project are example simulations of two engine components: an impeller and a blade cascade. In both of these cases, testing will cover the cooperation between the partners in terms of a distributed development of the components, the functionality of the grid (e.g. resource usage, VOs or security), the user interfaces and the automatic generation of metadata.

The UNICORE 6 grid middleware will be used to develop the grid infrastructure in AeroGrid. Currently, the Grid Application Toolkit (GAT) is being used as the client-side grid programming interface. At a later stage, the GAT’s successor, Simple API for Grid Applications (SAGA), will be used to enable communication with other middleware solutions. A web portal based on the GridSphere server and the DataFinder data management client will be provided as user interfaces. Both interfaces will be adjusted in order to support the various user scenarios optimally.

A key aspect here is the seamless integration of the user interfaces in the respective working environments of the application partners. The aim here is to hide the grid functionalityfrom the engineers as much as possible. In order to achieve this, the DataFinder will be used for the research and university partners. This can start CFD calculations entirely transparently on local computers and clusters or in UNICORE 6 grid resources. At MTU, the interfaces to the AeroGrid will be fully integrated into the company’s existing ePROTAS Engineering Workbench working environment.
The integration of a provenance service (that records detailed information about all executed processing steps) will increase the reliability of the results as well as users’ faith in their quality.

Examples of the information recorded in AeroGrid include the versions of the programs used, information about the hardware used, model parameters of the calculations in question and information about the libraries and compilers used. Even in years to come, this information will make it possible to trace past calculations and, if necessary, to use the original parameters and software environment to repeat the calculations.