Presentation
HPC-DA: An Architecture on the Intersection between Flexibility and Maximum I/O Performance
Speaker
Event Type
Machine Learning Day
AI/Machine Learning/Deep Learning
Big Data Analytics
Heterogeneous Systems
Performance Analysis and Optimization
System Software & Runtime Systems
TimeWednesday, June 27th12pm - 12:30pm
LocationPanorama 2
DescriptionIn 2016 the ZIH of TU Dresden presented an innovative approach to the Deutsche Forschungsgemeinschaft (German Research Foundation) to build an HPC machine that is targeted at the technological intersection of Big Data, Machine Learning and HPC. The system has been funded and is currently under construction. In this talk, we will present some use cases from the Big Data Competence Center Dresden/Leipzig and our HPC environment from the I/O perspective. Furthermore, we will depict how the architecture that we designed supports these use cases. We will present the general design guidelines and describe how current technology is integrated into the system.