NHR application rules – DRAFT
Since January 2021, NHR@FAU is one of eight (nine) centers of the NHR Alliance.
The final rules and workflows for applying for an NHR project / account are not yet fully fixed within the alliance. There will be a joint portal of all NHR centers for compute time applications. Nevertheless, you’ll find the preliminary conditions below.
You can reach NHR@FAU by email to email@example.com. Reachability by phone cannot be guaranteed as many of us are still partially working from at home.
Individual consulting using Zoom, MS Teams, BigBlueButton, or DFN-Conf can be arranged on short notice.
Offerings of NHR@FAU
NHR@FAU is offering support especially in the area of performance engineering and single-node performance analysis. The application focus is on atomistic simulations, i.e. molecular dynamics, chemistry, and certain areas of material sciences.
NHR@FAU is offering two different types of compute resources:
- GPGPU cluster “Alex” with Nvidia A40 and Nvidia A100 GPGPUs
- Parallel computer “Fritz” with Intel Xeon Platium 8360Y “IceLake” processors and HDR100 interconnect within islands of 64 nodes (i.e. 4.608 cores)
Both clusters will become generally available in the first half of 2022. Early adopters can get access to Alex in December 2021 and limited access to Fritz in January. All dates are still subject to change owing to the general shortage of IT components and delays in delivering the systems.
To register as early adpoter of Alex, fill in a simple form. You’ll get contacted once the system will be opened for first users.
Types of projects at NHR@FAU
Regular projects by default have a duration of 12 month. The review process typically takes less than 3 months. Test / porting projects will be decided within 3-4 weeks.
Details of the resource limits listed in the table are still subject to change.
|Type of project||possible resources on “Alex”||possible resources on “Fritz”||type of review||remarks|
|test / porting project||up to 3.000 GPU hours (A40 or A100)||up to 500.000 core hours||
A scientific advisor from NHR@FAU (liaison scientist) is recommended.
|normal compute time projects for granted DFG or BMBF projects||
||1 – 10 mio core hours||
A scientific advisor from NHR@FAU (liaison scientist) can be requested
|normal compute time projects built without granted DFG or BMBF projects||
||1 – 10 mio core hours||
A scientific advisor from NHR@FAU (liaison scientist) can be requested.
|large compute time projects||
||10 – 20 mio core hours||
||A scientific advisor from NHR@FAU (liaison scientist) is mandatory.|
|follow-up projects||as above||as above||
||a scientific advisor from NHR@FAU (liaison scientist) can be requested|
Projects falling into the “large compute time projects” category should also consider applying for compute time at one of the three federal Tier1 HPC centers of the Gauss Centre for Supercomputing (GCS).
General structure of NHR compute time application proposals
The outline for proposals is a DRAFT and subject to minor changes. NHR sites may also add additional local requirements.
- General information on the applicants
- project manager / technical contact
- responsible scientist (professor / postdoc)
- federal state
- scientific field (DFG categories)
- affiliated scientists (if any; including their federal state)
- information on other compute time proposals
- HPC experience of the planned users (publications, attended HPC training, etc.)
- Project information
- desired NHR center (probably NHR@FAU when you are reading this page)
- duration of the resource application
- new / follow-up compute time project
- project type
- existing reviews from the NHR Alliance, DFG, BMBF
- will sensitive data be processed?
- has a similar compute time proposal been submitted to an other center? If yes, what are the differences?
- Technical description (mandatory also for test / porting projects) – not more than 2 pages
- requested resources (type and amount)
- required software
- other technical requirements
- description of the software to be used; suitability for HPC and batch processing
- justification of using HPC resources (related to the scientific work program; basis for the estimation of the resources)
- proof for efficient HPC usage (e.g. performance plots)
- Scientific project description – up to 4 pages for normal project, up to 10 pages for large projects
- project abstract (may be published on the NHR web pages)
- state of the art
- own prior work
- scientific project aims and differentiation from other projects
- detailed scientific project description including numerical methods and their implementation
- For follow-up projects
- resources used so far and results obtained so far
- short intermediate report for publication on the NHR web page
- outreach of the project (e.g. publications)
- suggested reviewers
Acknowledgement of NHR@FAU
Please use the following formulation for acknowleding the ressources and the support provided by NHR@FAU:
The authors gratefully acknowledge the compute resources and support provided by NHR@FAU.
It might be necessary to also include our funding number or BMBF as funding agency. Details will be provided in the future.