LUMI User Survey 2024: challenges and solutions ahead

The annual LUMI User Survey is conducted to measure the overall satisfaction with the LUMI services, to hear about improvement needs, and to address any issues users may face. An overview of the 2024 results was given by Pekka Manninen, Director of Science and Technology at CSC, at the LUMI User Support Team’s (LUST) virtual coffee break in December 2024.

This year, fewer than a hundred replies were given in the survey – a slightly smaller number of respondents than last year. Hence, the representability of the survey is somewhat questionable as LUMI has over 2000 active users, but it gives indications from the users, nonetheless. The scientific domains represented in the survey correlated reasonably well with the actual use of the system – 1/3 of the respondents came from physical sciences and another 1/3 from AI-related projects. Over half of LUMI’s resources are used for AI-related projects. Most of the respondents were academic users using the EuroHPC JU resource quota; users from Finland’s national LUMI quota represented the 2nd largest responder group in the survey.

Concerns and commitments

The survey revealed a lot of valuable feedback. LUMI-G was seen as the most important computing service when asked about the importance of LUMI’s different services. Whereas LUMI-C had declined in importance compared to last year’s results – the AI use of LUMI also resonates here. The LUMI User Support Team services were also seen as key services, as was the LUMI documentation, especially the software library documentation page. About half of the respondents had attended a LUMI training event, and those were also seen as necessary.

In general, people are satisfied with LUMI’s computing power and file system. On the challenges side, the survey pointed out issues with LUMI’s software stack, especially with how to install software.

– This is the first thing for us to consider, and the LUMI User Support Team will be providing more comprehensive support for software installations and deployment. Just remember to contact them to get help, Manninen pointed out.

The study also revealed concerns about system stability and reliability. Some technical problems during this autumn, now solved, were affecting the results of the survey and LUMI is now more stable than before.

– Next year we will also put even more effort with the system vendor to track down the root causes of crashed jobs in order to fix them. For 2025, we foresee a couple of service breaks: Lustre updates in early 2025 and a two-week maintenance break in June. We will inform users about the exact dates well in advance, Manninen continued.

Another issue in the survey was the queue times and Slurm configuration – the same issues that were reported in last year’s survey.

– Unfortunately, we can’t do much for Slurm configuration as we are not manually adjusting job priorities; they follow Slurm practices. But we will have fewer projects on the system starting in Q1/2025, and this will shorten the queue times. We thank everyone who took part in the survey, your feedback is very valuable to us. We will do our best to make the use of LUMI even better for the users, Manninen concludes.

See also:

LUMI User Support web pages

Video introduction of the LUMI web interface