Hello!
First of all, happy New Year! We hope that you’ve enjoyed the holidays.
In 2021 you all submitted 19 mio. jobs and used 21 mio. billing hours worth of compute. During the year, storage usage grew from 7.5 PB to 9.6 PB, and peaked at 10 PB.
Last, but not least, welcome to the 200 users that joined us during 2021!
Anders, Rasmus, and Dan
The GenomeDK Team
Our office has moved to the newly renovated buildings in the University City. You can now find us in building 1872, room 359.
We have now switched completely to our own disk-based backup solution! This means that we can now much more quickly recover files from backup. In the future we will also provide a more flexible way of specifying which files should be backed up.
We’re now on social media as @GenomeDK_AU -on Twitter. Follow us for casual status updates and other HPC-related content. Get in touch if you have ideas for users, projects or research we should feature.
GenomeDK is a ISO 27001-compliant HPC facility. This means that we have a formal information security management system (ISMS) in place to guide our information security choices and implementation, and that you can safely store your sensitive data on GenomeDK.
During the past six months many of you have experienced unstable access to faststorage due to a range of (mostly) software-related issues in the filesystem software (BeeGFS) that is responsible for faststorage. We have previously described the actions we have taken to resolve these issues in detail. Since then we have introduced additional automation to self-heal faststorage and are working with the company behind BeeGFS to address any remaining issues.
The space command has been updated and now provides a monthly overview of billing hours used by users, projects, and all projects owned by you. To see a report for your user specifically:
[local]$ space user
Or for a specific project:
[local]$ space project --project <name>
Or get an overview for all projects owned by you:
[local]$ space overview
Due to supply issues and significant price jumps in 2021, we did not purchase any new hardware. In the coming year we’ll be acquiring new hardware for both compute and storage, which should result in shorter waiting times when submitting jobs and a more stable faststorage.
We’ll retire the s04 nodes during 2022 as they reach end-of-life.
Short version: Always do conda install to install packages, no matter what language the package is for. Not doing so can cause hard-to-debug issues and broken Conda environments.
Long version: When adding new R packages to Rstudio, you should always install it via Conda. For example, to install the Tidyverse package, do this:
[local]$ conda install r-tidyverse
Instead of doing this in the Rstudio console:
install.packages("tidyverse")
The same applies for Python packages. For example, to install SciPy, do this:
[local]$ conda install scipy
Instead of this:
[local]$ pip install scipy
Thank you for reading!