The strong methodological focus of EXC 2075 with respect to novel computational techniques, Machine Learning and Data Science requires not only production runs in centers like HLRS, but direct access to hardware during code development. Conventionally, this problem is addressed with powerful individual workstations or small cluster installations.
To counteract the high overhead for hardware and software maintenance under this approach, SimTech is currently installing the Experimental Compute Lab with the help of Malte Brunn. He is the EXC’s new Cluster Computing Systems Expert in the field of simulation science and started his work at the end of 2021.
“With Malte Brunn, we have found an expert who is not only fluent in the diversity of administration and monitoring tools to set up the system, keep it running and help with future extensions, but who also understands the specifics of the simulation codes of our researchers. This unique combination makes him the perfect candidate to administrate the SimTech Experimental Compute Lab", explain the members of the SimTech committee "Experimental Platform & Central Hardware" their decision.
Malte Brunn is a SimTech original. He obtained his Bachelor and Master degree in the SimTech study program Simulation Technology and directly started his PhD after his graduation in 2017 at the Institute for Parallel and Distributed Systems (IPVS) in the group of Miriam Schulte. His research is focused on algorithmic and numerical optimization of simulation software for high-performance computing.
“This job at the heart of a computing cluster with a part-time research focus is very appealing. The opportunity to simultaneously pursue my own research agenda as a SimTech Postdoc and to closely work with users of the experimental compute platform (and to guide the development of this platform) is the perfect opportunity to continue my current research and at the same time efficiently share my experience and methodology about code optimization with colleagues from other fields and collaborate in a variety of projects within SimTech.” states Malte Brunn.
Malte has already contributed to the acquisition process, and will play a major role in deploying the SimTech Experimental Compute Lab. Among his tasks will be the continuous adaptation of the software stack to meet the diverse needs of the SimTech researchers, their direct support, and of course the monitoring of the Lab to keep it up and running. Through this continuously expanding experience, he will be instrumental for the design of future extensions of the Lab.
This Lab is called “wolfgang”, in honor of Wolfgang Ehlers, who has been one of the driving forces of establishing SC SimTech and who has lead EXC 310 until 2017. It will be operated with a local queueing system that accommodates not only interactive sessions for debugging and testing implementations but also provides a cloud-like interface for research geared toward pervasive computing. To stay abreast of the rapidly evolving hardware, especially with respect to accelerators such as GPUs and future architectures, the system is designed for extensibility.
The lab will be highly heterogeneous, including a shared memory CPU partition, a GPU partition and a dedicated ML partition in the initial deployment. All nodes are equipped with powerful fine-granular metadata acquisition equipment and provide for a high bandwidth, low latency link to DaRUS, our data and software hub.
As soon as the lab is fully installed further information will follow.