Service Desk
LUMI
Supercomputer
LUMI is a leading platform for AI and one of the EuroHPC world-class supercomputers
Service description
With LUMI, researchers are able to solve completely new types of research questions with the combination of HPC, AI (especially deep learning) and high-performance data analytics to solve a single research question. LUMI Supercomputer service allows customers to run parallel scientific workflows through a batch queueing system. Users can execute their own codes, install Linux compatible applications, and utilize LUMI’s software collection. For code parallelization MPI and OpenMP can be used. HIP, OpenMP offload, and OpenACC can be used with GPUs. Additionally, mathematical libraries are available. The LUMI computing environment includes tools for debugging and performance analysis. LUMI’s architecture is described in detail on the LUMI website.
Technical information
The LUMI software policy allows users to install software themselves. LUMI has also pre-installed applications available. Users benefit from skills in Linux and programming expertise e.g. Fortran, C/C++, or Python. There are multiple training options available for beginners, intermediate and advanced HPC users. The EasyBuild tool is provided with accompanying build recipes for several applications. For more details, see the LUMI software documentation pages in user guide.
Protection level
Restricted access
You can decide who can access or download the data, and manage data sharing.
Restricted and controlled access
You can decide who can access or download data, and manage data sharing and export from isolated environments.
Read more about protection levels
How to obtain the service
In order to access and use the service the customer must have a LUMI user account and a project, which can be applied for either on an academic or commercial basis. You can apply for the national LUMI resources either as an academic user affiliated with a Finnish higher-education or research institute or as an industry user from an organization established or located in Finland.
Service is free of charge for academic use cases defined in the CSC free of charge policy. Depending on the use requirements there are different packages. Compute is billed in units of CPU-core-hours for CPU nodes and GPU-hours for GPU nodes. Storage space is billed in units of TB-hours. Developer project access starts from 1 million core hours. Access is provided for 1 year.