News on HPC systems @ UiO, application deadline for CPU time through Sigma2, interesting conferences and external courses.
- The IT department IT i forskning, formidling og utdanning FFU(NO), or Division for Research, Dissemination and Education (RDE)(EN), is responsible for delivering IT support for research at University of Oslo.
- The division's departments operate infrastructure for research, and support researchers in the use of computational resources, data storage, application portals, parallelization and optimizing of code, and advanced user support.
- Announcement of this newsletter is done on the hpc-users mailing list. To join the hpc-users list, send a mail to sympa@usit.uio.no with subject "subscribe hpc-users Mr Fox" (if your name is Mr Fox). The newsletter will be issued at least twice a year.
A new year is sailing up, and since 2024 is a leap year, great leaps around HPC and AI are to be expected. As we saw last year with tools like ChatGPT and Whisper taking us all by surprise, the biggest game changers are not the ones we know about, but rather the ones we don't already know about. But before looking for more unknown unknowns, let's have a look at what we already know is going on and/or will happen around HPC and AI @ UiO in this first HPC Users@UiO newsletter of 2024.
Testing NVIDIA MIG for splitting up GPUs
On Fox, we are currently testing splitting up NVIDIA GPUs into smaller parts, Multi-Instance GPUs (MIGs), that can be handed out individually to jobs. This allows more jobs to use the GPUs at the same when they don't require a whole GPU. This is especially interesting for the Educloud OnDemand service. Users should not need to make any changes to their job scripts and SLURM will see each partition as one GPU. Only thing to consider is that, there will be only 16GB (1/5 of 80GB) of GPU memory available per GPU partition.
BioNT - BIO Network for Training
BioNT - BIO Network for Training - is an international consortium of academic entities and small and medium-sized enterprises (SMEs). BioNT is dedicated to providing training programs and fostering a community for digital skills relevant to the biotechnology industry and biomedical sector, and University of Oslo is part of this consortium through the IT Department.
Norwegian AI Cloud
The Norwegian AI Cloud (NAIC) project is aiming at providing infrastructure and support for Norwegian machine learning (ML)/AI researchers, students and industry.
More details: https://www.naic.no/om/index.html
Fox Supercomputer - get access
The Fox cluster is the 'general use' HPC system within Educloud, open to researchers and students at UiO and their external collaborators. There are 24 regular compute nodes with 3,000 total cores and five GPU accelerated nodes with NVIDIA RTX 3090 and NVIDIA A100 cards available. Access to Fox requires having an Educloud user, see registration instructions. About 250 projects have already joined Educloud!
For instructions and guidelines on how to use Fox, see Foxdocs - the Fox User Manual.
Software request form
If you need additional software or want us to upgrade an existing software package, we are happy to do this for you (or help you to install it yourself if you prefer that). In order for us to get all the relevant information and take care of the installation as quick as possible, we have created a software request form. After filling in the form, a ticket will be created in RT and we will get back to you with the installation progress.
To request software, go to the software request form.
Support request form
We have had great experience with users requesting software installations since we introduced the software request form. We now usually get all the information we need at first contact. We want to further improve our support and get to the root cause of an issue faster. Therefore we now encourage you to fill in a form when you need help with other types of issues as well. When the support form is submitted it will be sent to our hpc-drift queue in RT and will be handled as usual. The difference from emailing us directly is that we will now immediately get needed bits of information and your tickets will be labelled according to what system you are on and what issue you are facing.
The link to the new support form will be shown when you log in our servers and it has also been added to relevant documentation pages. You can have a look at it here:
https://nettskjema.no/a/hpc-support
We encourage users of all our HPC resources to use this form. Whether it concerns Fox, LightHPC, ML-nodes, Educloud OnDemand, Galaxy-Fox or our individual appnodes.
New web pages for Sigma2 NRIS
Sigma2 NRIS is announcing the launch of their new website. While you'll find that much of the content and the main navigation remain the same, there are significant enhancements to meet universal design requirements. The new setup is designed to be more user-friendly and intuitive, ensuring that visitors can easily find the information they need. This is part of the ongoing commitment to provide a seamless and efficient online experience for all users of the national e-infrastructure services.
Sigma2 cost for UiO-users
As was mentioned in the previous newsletter, during the renegotiation of the Sigma2-BOTT-RCN agreement, UiO managed to lower our cost from 2023 onwards to approximately 24MNOK instead of the planned 37,5MNOK. The short story is that UIO, as all other users, will pay the actual operational costs, but we will pay up-front as a guarantee, and not per use.
It is of utmost importance that all researchers and especially those at UiO are aware of the actual operational cost connected to their work, and that you try to apply for external funding to cover your part of the operational cost. How much, and if, you will have to pay for your operational costs directly from your project depends on your Faculty and how they choose to handle the invoice.
Centrally at UIO the administration will shave a significant part of the invoice off, this will function as a “centrally covered discount” before the remainder of the invoice is sent to the Faculties and Museums based on their usage.
TSD resources and Sigma2 application deadline
As a side-effect of the above mentioned cost calculation, we at UiO can spend a lot more of the TSD resources that Sigma2 owns without getting any extra expenses either for individual research project or for UiO as a whole. We therefore strongly encourage TSD users to apply for TSD resources from Sigma2 for the coming allocation period.
New Sigma2 e-Infrastructure allocation period 2024.1, application deadline 1 February 2024
The Sigma2 e-Infrastructure period 2024.1 (01.04.2024 - 30.09.2024) is getting nearer, and the deadline for applications for HPC CPU hours and storage (for both regular and sensitive data), is 1 February. This also includes access to the Sigma2 part of TSD, as well as LUMI-C and LUMI-G.
Please note that although applications for allocations can span multiple allocation periods, they require verification from the applicants prior to each application deadline to be processed by the Resource Allocation Committee for a subsequent period. Hence any existing multi-period application must be verified before the deadline to be evaluated and receive an allocation before the new period starts. This does not apply to LARGE projects.
Kind reminder: If you have many CPU hours remaining in the current period, you should of course try to utilize them asap, but since many users will be doing the same there is likely going to be a resource squeeze and potentially long queue times. The quotas are allocated according to several criteria, of which publications registered to Cristin is an important one (in addition to historical usage). The quotas are based on even use throughout the allocation period. If you think you will be unable to spend all your allocated CPU hours, it is highly appreciated to notify sigma@uninett.no so that the CPU hours may be released for someone else. You may get extra hours if you need more later. For those of you that have run out of hours already, or are about to run out of hours, take a look at the Sigma2 extra allocation page to see how to ask for more. No guarantees of course.
Run
projects
to list project accounts you are able to use.
Run
cost -p nn0815k
to check your allocation (replace 0815 with your project's account name).
Run
cost -p nn0815k --detail
to check your allocation and print consumption for all users of that allocation.
HPC Course week/training
Norwegian Research Infrastructure Services NRIS has an extensive education and training program to assist existing and future users of our services. UiO has joined NRIS training providing training to all Norwegian HPC users, instead of just focusing on UiO users, this makes it possible to provide a more streamlined and consistent training by consolidating the training events. The courses are aimed to give the participants an understanding of our services as well as using the resources effectively.
There is a HPC on-boarding course coming up April 9-11 as well as an intermediate HPC course where dates will be announced very soon.
See the following for list of all events: https://documentation.sigma2.no/training/events.html
Training video archive: https://documentation.sigma2.no/training/videos.html
Please do not hesitate to request new topics or uncovered areas of training to use the services more optimal or make your work with our systems easier.
Other hardware needs
If you are in need of particular types of hardware (fancy accelerators, GPUs, ARM, Kunluns, Dragens, Graphcore, etc.) not provided through our local infrastructure, please contact us (hpc-drift@usit.uio.no), and we'll try to help you as best we can.
Also, if you have a computational challenge where your laptop is too small but a full-blown HPC solution is a bit of an overkill, it might be worth checking out NREC. This service can provide you with your own dedicated server, with a range of operating systems to choose from.
With the ongoing turmoil about computing architectures we are also looking into RISC-V. The European Processor Initiative is aiming for ARM and RISC-V and UiO needs to stay on top of things.
With the advent of integrated accelerators (formerly known as GPUs) with shared cache-coherent among all execution units including accelerators (like AMD MI300 and NVIDIA Grace/Hopper) these might be of interest for early adopters. Call out if this sounds interesting.
Publication tracker
The Division for Research, Dissemination and Education (RDE) is interested in keeping track of publications where computation on RDE services are involved. We greatly appreciate an email to:
hpc-publications@usit.uio.no
about any publications (including in the general media). If you would like to cite use of our services, please follow this information.
Fox trying out Educloud OnDemand
