First in-person (hybrid) HPC Café after more than two years
The next HPC Café takes place as a hybrid event (in person + Zoom) on Tuesday, July 12, at 4 p.m.
The HPC Café is an opportunity to get to know each other and to stay up to speed with current developments at NHR@FAU. After a general Q&A phase, the focus will be put on advanced or application-specific topics. In the HPC Café we also offer short talks about anything HPC related, from running job scripts to performance issues. The Café is also an informal platform to give feedback or talk about general requests.
As always, there will also be plenty of time to get in touch with your favorite HPC group. This time, after more than two years, we invite you again to come to RRZE’s seminar room 2.049 to have coffee and cake, just like old times.
Time: Tuesday, July 12, at 4 p.m.
The event will start with the open coffee chat and Q&A. At 4:30 p.m., the presentation will start to allow remote attendees to join on time.
Location: RRZE (Martensstr. 1), seminar room 2.049 and Zoom online session (https://fau.zoom.us/j/69439931483)
Topic: Job monitoring for end users – ClusterCockpit and PyTorch Profiling at work
We will show how job-specific system monitoring can be used to assess how efficiently a cluster job makes use of the hardware resources. Our new monitoring system, ClusterCockpit, provides a plethora of metrics, such as flop rates, memory bandwidth, I/O, and even energy consumption. This type of data can help you identify “low-hanging fruits” in job and workflow optimization and will eventually lead to more science per day for all HPC customers. For users who employ the PyTorch framework, we will briefly demonstrate how useful trace information can be obtained and visualized that will give more insight into the resource utilization on GPUs.