On this site you will find information about how researchers can get access to HPC ressources on either AU's own facilities or through national or international HPC cooperation.
AU offers access to national HPC resources in collaboration with DeiC (Danish e-infrastructure Coorporation).
This collaboration means that researchers at AU can apply for acces to the national HPC resources under DeiC.
Moreover, AU itself has access to a number of HPC resources, that we make available to our reseachers.
HPC stands for High-Performance Computing. This describes a number og different machines capable of calculations and data handling that surpasses the scope of a normal PC. In other words, HPC is a research tool that makes formerly impossible projects possible.
If your project lives up to the criteria described here, you can apply for HPC resources.
DeiC has four different HPC types at their disposal. The HPC types are offered by a number of consortia, and the universities are parties to these. The HPC types are categorised according to which hardware they consist of, as well as which demands the system makes of the user. Here we will go through the different types of HPC you can apply for.
For further information and details, visit DeiC's overview page thorugh this link.
| HPC type | Description |
|---|---|
| DeiC Interactive HPC | This type is easy to use for new users and for educational purposes. The primary target group for Interactive HPC are users who desire an HPC-system as close as possible to a normal computer. The computer is accessed via a browser and offers a large number of pre-installed programmes, just like on your own PC. Because Interactive HPC is quite easy to use, it is ideal for new users and students. On the other hand, the many available programmes offer a wealth of opportunities to the more experienced users. For further information see https://interactivehpc.dk/#/. |
| DeiC Throughput HPC | Ideal for small- and medium-sized tasks that use big data/files, and for analyses of sensitive personal data. DeiC Throughput HPC offers a traditional HPC-setup accessed via an SSH-connection to a Linux server. From here, you can run jobs through a queueing system (e.g. Slurm), which manages resource allocation and runs. The system can handle large quantities og data with a high level of security, and is ideal for parallelisable programmes. All in all you have a large amount of control over your jobs on Throughput HPC, which is why it can take longer for new user to learn how to use it compared with Interactive. For more information anbout the system and how to gain access, see https://www.deic.dk/en/supercomputing-hpc/tjenester/deic-throughput-hpc. |
| LUMI Capability HPC | Capability HPC is the European pre-exascale supercomputer, LUMI. LUMI offers a setup similar to that of Throughput, with the very best hardware. Specifically, the computer's nodes are connected to the memory, the storage, and each other in a way that minimises latency during communication. This means that LUMI can handle calculations that surpass the capacity of Throughput HPC due to latency or memory. LUMI is financed 50% by EuroHPC Joint Undertaking, and 50% by the consortium countries, Denmark among them. Denmark has access to 3% of the LUMI resources, and AU has access to its own part hereof, which is allocated by the HPC forum via quaterly calls. You can read more about LUMI here: https://www.lumi-supercomputer.eu/ |
You can apply for HPC resources for both small and large project. The two kinds of projects each have their own application process.
A small project (or a ”sandbox project”) is defined as a project that demands less than:
You apply by emailing deicfrontoffice@au.dk, with a brief description of the following:
| 13th January: | Call published by DeiC: Apply for HPC ressources |
| 10th March: | Deadline for application to DeiC via e-grant: www.e-grant.dk |
Your can find help on how to fill out your application here: guide for application.
There is no deadline for a small projects, so you can always send an application.
The local HPC Front-office will handle your application and return with a response within 2-3 weeks.
The applications are assessed in relation to research quality, qualification of the applicant, feasibility, publication and dissemination of the result.
Applications that are judged as not fulfilling the criteriae will be rejected.
Upon rejection, the application is informed of the possiblity of applying for resources via the local resource pool.
| 13. January - 10. March | Researchers can submit an application to DeiC via www.e-grant.dk. |
| Early to mid-June | DeiC has completed assessment of the applications. Applicants, universities, and HPC centres are briefed about the result. Resources are expected to be ready for use 1. July. |
| Mid-June - early July | AU is free to dispose of a part of the HPC-ressourcer itself. These resources are granted to those AU-applicants, who did not receive resources from DeiC, if the researchers in question decide to apply to the HPC Forum. The Forum handles these applications at their first meeting after the DeiC results are announced. |
Out of the national HPC resources DeiC coordinates, 50% are allocated locally at the individual universities, while 45% are allocated following calls. The final 5% are so-called "sandbox", and is used for testing calculation methods by both new and experienced users.
The "sandbox" is allocated by DeiC. You can read more about how to apply for sandbox resources here: link.
Should you decide to apply for access to the 45% resources allocated following calls, you application will be assessed by the E-ressource Committee, a national committee for allocation convened by DeiC, with representatives from all the Danish universities and scientific fields. There are two annual calls, in January (with a deadline in March) and July (with a deadline in September).
The individual university is responsible for allocating its own resources among the researchers.
At AU it is decided that 50% of the resources are reserved for an ”AU sandbox”. This is reserved for projects that need a relatively small calculating capacity.