LUMI Benchmark 2026 Allocations¶

This page provides information about the requirements for submitting a LUMI Benchmark 2026 allocation request.

LUMI Benchmark allocations are intended for benchmarking, porting, scalability testing, and performance evaluation of applications on the LUMI system. The allocation is designed to help applicants assess application suitability, estimate future resource requirements, and prepare for future LUMI allocation requests.

Proposals are reviewed throughout the year. Vacations and holidays may affect processing times.

Who is eligible to apply?¶

The Principal Investigator (PI) is responsible for the proposal and usage of any resources granted.

You must be at least a PhD student and hold an appointment of at least 20 percent at a Swedish university or research institution recognised by the Swedish Research Council, as NAISS is financed by Swedish taxpayers with the objective of strengthening science and research in Sweden.

Researchers intending to apply for future LUMI allocations are particularly encouraged to use Benchmark allocations for application porting, optimisation, scalability studies, and resource estimation.

Project information¶

When registering a proposal in SUPR, you must provide the following information:

Project title
Affiliation - name of university/institute or similar
Abstract - a scientific project abstract in English
Requested duration and start date
Classification - according to Research Council classification codes

Resource usage¶

Explain how you have estimated your need for computational resources and justify the requested amounts for each resource.

The purpose of the allocation should be clearly described, for example:

Application porting to LUMI
Performance benchmarking
GPU scaling studies
CPU scaling studies
Workflow testing
Preparation for future LUMI allocation applications

Provide motivation for the system-specific information requested.

Computational resources¶

The proposal must specify the computational resources needed, including the estimated number of GPU hours and/or CPU core hours required for the benchmarking activities.

Indicate which LUMI resources will be used:

LUMI-G for GPU-based benchmarking
LUMI-C for CPU-based benchmarking

The requested resources should be supported by calculations.

Example 1: GPU benchmarking

gpuhours = nodes x used_gpus_per_node x wall_time x number_of_runs

Example 2: CPU benchmarking

corehours = nodes x used_cores_per_node x wall_time x number_of_runs

Describe:

The software and applications to be benchmarked.
Relevant software dependencies.
Whether software is already available on LUMI or must be installed by the project.
Any licensing requirements and confirmation that the requested resources support the necessary licenses.
The benchmarking methodology and expected outcomes.
The scaling range that will be tested (number of nodes, GPUs, or cores).

Storage resources¶

This section may be omitted if no storage, or only default storage, is requested.

If storage is required for benchmarking activities, specify:

Estimated storage size required.
Storage type needed.
Expected duration of storage usage.

LUMI storage allocations are specified in TB-hours rather than fixed storage quotas. Available storage resources include:

LUMI-P (Lustre storage)
LUMI-F (Flash storage)
LUMI-O (Object storage)

Storage accounting follows the LUMI charging model:

LUMI-P: 1 TB-hour per TB-hour used
LUMI-F: 10 TB-hours per TB-hour used
LUMI-O: 0.5 TB-hours per TB-hour used

You may combine the different storage types within the approved storage allocation, provided that the approved quota is never exceeded for any storage type.

Data management plan (DMP)¶

This section may be omitted if no storage, or only default storage, is requested.

The DMP should describe how benchmark data and metadata are managed during and after the project. Users are encouraged to organise their data systematically and maintain clear documentation of file contents.

Include at least the following:

Please describe the downstream plan for your data.
Where will benchmark input and output data be transferred at the end of the allocation?
Describe the mechanism by which data will be transferred to external sites for further analysis or archival.
Give an estimate of the amount of data that will be transferred.
LUMI strongly recommends keeping a copy of primary research data elsewhere. Briefly describe where primary data will be stored and how backup will be managed.
Storage is intended only for active data used in benchmarking and analysis. Explain how outdated or unused data will be routinely removed during the project.

Activity report¶

If you have previously been granted NAISS resources, you must upload activity reports for those projects to be eligible for new allocations.

Include an explanation for any major deviations from the originally planned resource usage.

Review process¶

Benchmark proposals are reviewed throughout the year. Vacations and holidays may affect processing times.

The review focuses primarily on:

Technical feasibility on LUMI.
Appropriateness of the requested resources.
Suitability of the benchmarking plan.
Expected benefit of the benchmarking activities for future use of LUMI resources.