Data Analytics for HPC
Keyvisual image main

CATALYST researches methods for analyzing large datasets produced by modeling and simulation with the goal of implementing a framework that combines HPC and data analytics.

High Performance Computing (HPC) is a key driving factor for both academic and industrial innovation. HPC technology is well-established and actively applied in various areas of applications including modelling and simulation. Obtained results from simulations are constantly increasing, and thus it has become a major challenge for domain experts to manually analyze the data in time. Data analytics is the perfect method to automate these processes; data analytics can support experts in decision-making. Therefore, CATALYST aims at researching methodologies and implementing a framework to combine HPC with data analytics.

In cooperation with Cray, HLRS have extended the current flagship HPC system, Hazel Hen, with specific data analytics hardware. In the first phase of the project, both systems, Hazel Hen and the new Urika-GX, are still operated independently. However, our vision is to combine both in a seamless manner. This means that we have to tackle various challenges including security aspects, data transfer, and accounting, just to name a few.

Since the majority of today’s data analytics algorithms are tailored for text processing (e.g., news clustering) and graph analysis (e.g., social network studies), we are further in need to evaluate these algorithms with respect to their applicability to the High Performance Computing domain. Thus, we are examining future concepts for both hardware and software. The project will therefore pursue multiple case studies from divergent domains throughout the next three years. The project will investigate, for example, the operation in industrial areas by cooperating with the Daimler AG.


01. October 2016 -
31. December 2022


MWK Baden-Württemberg