Hélios

From CC Doc
Jump to: navigation, search
This page is a translated version of the page Hélios and the translation is 100% complete.

Other languages:
English • ‎français
Availability: March 2014
Login node: helios3.calculquebec.ca
Globus endpoint: computecanada#helios-dtn
Data transfer node (rsync, scp, sftp,...): helios3.calculquebec.ca

Hélios is a supercomputer with general purpose graphics processor nodes (GPGPU). It was installed in Laval University's computing centre in the spring of 2014. The server was purchased with funds from Calcul Québec and from researchers at Laval University and Université de Montréal. Until the commissioning of Cedar and Graham, Helios had been the largest GPU deployment administered by Compute Canada, both in terms of computing power and number of GPUs.

Transferring your data

As of January 2020, Hélios is being aligned with the new Compute Canada standards.

  • Connection identifiers are migrated from Calcul Québec to Compute Canada;
  • Scheduling is migrated from Moab to Slurm;
  • Filesystems are reorganised to conform to the structure used on the other national clusters;
  • The modules system and the software are updated.

In this context, you must transfer your files from the old to the new filesystem. We suggest you use Globus from endpoint computecanada#colosse to endpoint computecanada#helios-dtn.

Site-specific policies

By policy, Hélios compute nodes cannot access the internet. If you need an exception to this rule, contact technical support justifying your needs.

The crontab tool is not offered.

Each job should have a duration of at least one hour (at least five minutes for test jobs) and a user cannot have more than 1000 jobs, running and queued, at any given moment. The maximum duration for a job is 7 days (168 hours).

Backups

A difference between Hélios and the new national clusters is that Hélios filesystems are not backed up. Please make sure your files are safely backed up elsewhere.

Jupyter

In addition to the traditional SSH access interface, you can also use a JupyterHub interface by connecting through https://jupyterhub.helios.calculquebec.ca/hub/spawn.

Software environment

Hélios uses version AVX of the Compute Canada software environment. See the list of available software under the AVX (Helios) tab.

Available GPUs

Two types of GPUs are available: NVIDIA K20 and NVIDIA K80. To indicate the type you require, use the following options in your submission script:

 #SBATCH --gres=gpu:k20:1

or

 #SBATCH --gres=gpu:k80:1

Storage

Hélios has only one 392TB Lustre filesystem. As is the case with the new clusters, there are three distinct areas which each have different quotas. The filesystem is not backed up.

HOME
Lustre filesystem
  • this is a small space that cannot be enlarged: we suggest you use your project space for large backups
  • 50GB storage space, 500K files per user
SCRATCH
Lustre filesystem
  • large space to store temporary files during compute operations
  • 20TB storage space, 1M files per user
PROJECT
Lustre filesystem
  • space designed for sharing data within a group and storing a high-volume data
  • 1TB storage space, 500K files per group

For transferring data via Globus, you should use the endpoint computecanada#beluga-dtn, while for tools like rsync and scp you can use a login node.

Node characteristics

nœuds cœurs mémoire disponible CPU stockage GPU
15 20 110G 2 x Intel Xeon Ivy Bridge E5-2670 v2 @ 2.5 GHz 1 x HDD of 2T 8 x NVidia K20 (5G memory)
8 24 256G 2 x Intel Xeon Ivy Bridge E5-2697 v2 @ 2.7 GHz 2 x SSD of 180G (330G usable) 8 x NVidia K80 (16 GPU, 12G/GPU memory)

Globus is a file transfer service