Newsletter, summer 2019

Hello!

We’re quickly approaching summer vacation here in Denmark, so this is probably a good time to take a look at the past year at GenomeDK.

We’ve collected numbers that demonstrate how we’re growing, how the cluster is used, and who the users are and where they come from. We also have a few tips for you, as well as updates from the last year.

We would like to feature your scientific work with GenomeDK in future newsletters. If you’d like to write a few lines about your research and how it’s being supported by GenomeDK, please get in touch!

Enjoy!

The GenomeDK Team

A few updates

Website and documentation

During the past year we’ve been working on improving our documentation and website. The two sites have now been merged and live at the same domain, genome.au.dk.

Most of the documentation has been rewritten and we tried hard to improve the structure and make it easier for new users to get started.

We welcome all feedback on the new site!

Coming up, more storage!

After manually pushing 840 hard drives into place we can increase our storage to 12 PB! We will also be saying goodbye to some of our old storage servers and will be migrating data from them to the new storage servers.

The new storage will be available as soon as the data migration and testing has been completed.

We’ve got GPU nodes

We recently added two GPU nodes with two GPUs each to GenomeDK. We encourage you to experiment with these nodes. Several users have confirmed that they get large speed-ups when running e.g. Tensorflow on the GPU nodes. You can read more here.

Some interesting facts

Who uses GenomeDK?

In the last year we’ve welcomed approximately 80 new users to the cluster which is now serving more than 400 people from 17 countries.

../../_images/worldmap.svg

Countries with at least one GenomeDK user marked in red.

We’re well on our way to world domination if you consider continents. You just need to find collaborators in Africa.

Although Aarhus University is by far the largest user of GenomeDK, we’ve seen quite a bit of interest in GenomeDK from outside AU too. Actually, 47% of our users are not from AU. Many of these are, of course, your external collaborators.

../../_images/top5_users.svg

Top five organizations based on the number of active user accounts.

How much compute is used?

In the last year you submitted 12 million jobs, averaging 33000 jobs per day. Our most busy day was July 26th, 2018, where 390223 jobs were submitted in one day!

We measure compute usage in units of billing hours which takes into account the time a job was running, how many cores it allocated, and how much memory it used.

In total you have used 18 million billing hours in the last year. We estimate that the cost of one billing hours is 0.12 DKK, so you collectively burned through almost 2.2 million DKK worth of compute time.

../../_images/top5_organizations.svg

Top five organizations based on their compute usage (which we measure in billing hours).

Region Midt is a substantial user due to research and clinical data analysis from the Department of Molecular Medicine. Likewise, Copenhagen University consumes a large chunk of our capacity.

How much storage is used?

Bioinformatics is data-intensive so storage is extremely important at GenomeDK. There’s currently at least 4.8 PB data stored on GenomeDK spread across 370 million files. Most of these files are tiny. In fact, 99% of files are less than 26 MB. In the past year, the biggest file on the cluster was 13.2 TB.

The estimated cost of storing 1 TB of data for one year is 180 DKK. Thus, you collectively spent 864000 DKK on data storage in the past year. If you had to store this data in the cloud, for example on Microsoft Azure, you would pay 18 mio. DKK… per month!

On top of that, we currently have 850 TB of data in backup, which cost you an additional 595000 DKK per year. On average, 87 GB of data is added to the backup every single day. This number is, however, increasing significantly.

../../_images/top5_organizations_storage.svg

Storage usage by organization. Aarhus University is of course at the top here, but has also been used GenomeDK for the longest time. Copenhagen University is a relatively new user and thus being number two in this plot shows that they’ve had explosive storage usage growth.

Let’s look at some of the projects that call GenomeDK their home. First, here’s a list of the top five largest projects by the total size of the files in the project.

Project name

Size

DBS_Exomes

371 TB

DanishPanGenome2

329 TB

MammalianMutation

302 TB

RetroVirus

246 TB

MolGen

191 TB

That’s a lot of data! But who then has the most files?

Project name

Number of files

IGdata

68 mio.

MammalianMutation

21 mio.

MomaRAWfiles

21 mio.

norfab

13 mio.

NChain

13 mio.

Almost 70 million files for the winner, which is exactly the same as the total number of files for number 2, 3, 4, and 5.

When are you active?

The heat map below shows job submissions from the past year grouped by day of the week and hour of the day.

../../_images/activity_week.svg

Activity map of job submissions. A darker color means more jobs were submitted.

GenomeDK users are clearly not morning people, but prefer to work late. There’s usually a lot of activity on Tuesday, Wednesday, and Friday afternoons. We’re also happy to observe that most users actually seem to take time off during the weekend!

A few tips

  • Do you want faster and more stable access to files in your home folder? We can now offer to move your NFS home folder to a new, faster, and more stable file server! To be moved the size of your home folder must be less than 100 GB. We encourage you to clean up your home folder, move big files to faststorage, and then contact us to have your user moved.

    To get the size of your home folder, run space. This gives you a nice report of your disk usage.

  • Our software repository, /com/extra is deprecated. We won’t add new software or update existing software in this repository. Instead, we encourage all users to use a proper package manager, like Conda.

Thank you for reading!