The next HPC Champions meeting will be held virtually on the morning of the 26th October 2021, and will involve talks and discussion sessions. Topics will include updates from supercomputing services, training, and HPC software. The workshop will run between 9 AM and 1 PM GMT. If you have an idea for a discussion topic then please get in touch with the organiser: Luis Fernandez Menchero firstname.lastname@example.org .
9:00 - 10:40 Presentations and lightning talks
9:00 - 9:10 Luis Fernandez Menchero. Information and Technology Services, Queen's University Belfast. Welcome.
9:10 - 9:40 Andrew Turner, Edinburgh Parallel Computing Centre (EPCC), The University of Edinburgh. "ARCHER2: performance at scale and planned RSE work."
9:40 - 10:10 Ales Zemljak. Institute of Information Science (IZUM), Maribor, Slovenia. "Vega HPC - 1st EuroHPC Supercomputer."
10:10 - 10:40 Christopher Edsall. Research Computing Services. Cambridge University. "Updates from Cambridge and Introducing the Open Exascale Lab."
10:40 - 11:10 Break
11:10 - 11:30 Selection of discussion topics
11:30 - 12:30 Discussion session
12:30 - 12:45 Wrap-up and planning for next meeting
To Register :
- Presenter: Dr Andrew Turner, Edinburgh Parallel Computing Centre (EPCC), The University of Edinburgh.
Title: "ARCHER2: performance at scale and planned RSE work"
Abstract: In this presentation I will give an overview of the performance of the full ARCHER2 system for a range of application benchmarks from materials modelling, CFD and biomolecular simulation at a range of parallelism from a single node (128 cores) up to half a million compute cores; and share the lessons we have learnt about running on the AMD processors and HPE Cray Slingshot interconnect over deployment and testing of the ARCHER2 service. I will also provide details on plans and ideas that the EPCC RSEs that make up the ARCHER2 CSE service have on developing the service throughout its lifetime and technical areas for investigation. We are keen to look for areas where we can work collaboratively with other RSEs going forward and so I would be interested to hear feedback on these ideas and suggestions for other areas where we could work together.
- Presenter: Aleš Zemljak. Institute of Information Science (IZUM), Maribor, Slovenia.
Title: "Vega HPC - 1st EuroHPC Supercomputer"
Vega HPC  is a 6.9 PFLOPS (Rmax) supercomputer and the first operational EuroHPC  system. Its conception is the fruit of the Slovenian national infrastructural project HPC-RIVR  that was expanded by the EuroHPC  initiative. It is hosted at IZUM  in Maribor , Slovenia . It is designed to serve as a "general purpose HPC", its hardware configuration distinctions are the rather evenly distributed computing power between the CPU and GPU partitions, a rather large data storage capacity and a high throughput to the outside world. Adding to these are also the compute node’s direct liquid cooling and low latency internal (Infiniband) network, as well as Ceph  being the storage cluster used for the large-capasity storage equipment. This Slurm  based cluster is operated by a young & newly trained team at IZUM , who are joined by HPC experts from all across Slovenia via SLING - Slovenian National Supercomputing Network , and with the help & support of Atos , the machine’s manufacturer. Vega HPC users can use their own custom software, choose from a plephora of different software already installed on the system or work with our support team to enable an environment that suits their needs. Slovenia based users can apply for allocations via SLING, while EU & partner states based users can apply via EuroHPC issued calls .
[ 1] https://www.izum.si/en/hpc-en
[ 2] https://eurohpc-ju.europa.eu
[ 3] https://www.hpc-rivr.si/home_en
[ 4] https://www.izum.si/en/home
[ 5] https://www.visitmaribor.si/en
[ 6] https://www.slovenia.info/en
[ 7] https://ceph.io/en
[ 8] https://slurm.schedmd.com
[ 9] https://www.sling.si/sling/en
- Presenter: Christopher Edsall. Research Computing Services. Cambridge University.
Title: "Updates from Cambridge and Introducing the Open Exascale Lab."
The Cambridge Centre for Data Driven Discovery (CSD3) has long been a significant part of the UK HPC landscape providing services to not only Cambridge researchers but also DiRAC, IRIS and Tier 2 users. Since the last champions meeting there have been several devlopements including the upgrading of the hardware to the point where the new GPU partition Wilkes-3 is the number three most energy efficient supercomputer on the planet. We have also launched the Cambridge Open Exascale Lab and this talk will introduce the lab's aims and the workstreams.