Waves to Weather
print


Breadcrumb Navigation


Content

Z2 - Computing services

Project leaders: Prof. Dr. George C. Craig, Dr. Christian Keil, Prof. Dr. André Brinkmann, Dr. Marc Rautenhaus

Other researchers: Dr. Robert Redl (scientific programmer), Maicon Hieronymus (PhD student)

 

This IT infrastructure project is responsible for providing the essential support, expertise and facilities for computing-based research in W2W. In Phase 1, this was achieved by a post-doctoral scientific programmer and shared facilities located at LMU. In Phase 2, the computing requirements have increased hugely, with a need to process and share PetaBytes of data. To meet this need, both in Phase 2 and in the future, Project Z2 now includes both a service and a research component.

The service component addresses requirements for a central code repository and collaboration platform, internal and external web pages, and a data management strategy to enable storage and transfer of huge amounts of data, as well as a strategy to archive data in local and community accessible archives to ensure the reproducibility of results. A particular requirement that was only partially met in Phase 1 is for the support of visualization software. To meet these requirements we will take a distributed approach, with staff and hardware located at each of the three W2W locations, taking advantage of local computing facilities and benefiting from university co-funding of staff and equipment.

The research component has two parts. First a 50% post-doctoral position in Munich will be created to investigate the applicability to W2W diagnostic tools, of newly available Python libraries that support code parallelization and reuse, as well as of lossy compression algorithms to greatly reduce data volumes. If these methods live up to their promise, it will have a direct impact on many W2W research projects by enabling them to exploit substantially larger data sets. The second research project, to be carried out by a PhD student in Mainz, will explore the use of Algorithmic Differentiation (AD) on meteorological codes, with initial emphasis on cloud models. These methods have proven valuable in understanding the sensitivity of numerical models to changes in parameters and initial conditions in many areas of engineering, but have not been widely exploited in atmospheric science.