High-Performance Computing (HPC) Facility: Speed

The current infrastructure comprises:

• Twenty four (24), 32-core nodes, each with 512 GB of memory and approximately 1 TB
  of volatile-scratch disk space.
• Twelve (12) NVIDIA Tesla P6 GPUs, with 16 GB of memory (compatible with the CUDA,
  OpenGL, OpenCL, and Vulkan APIs).
• One AMD FirePro S7150 GPUs, with 8 GB of memory (compatible with the Direct X,
  OpenGL, OpenCL, and Vulkan APIs).

Job Management is handled by Univa Grid Engine.

The "cluster" mounts multi-TB, NFS-provided storage, which serves both persistent-scratch data and persistent-store data (not backed up; backed up, respectively).

Software (both open-source and commercial):

• Scientific Linux 7, supports containers
• Singularity (supports conversion from Docker containers), various machine- and specifically deep learning frameworks, Conda, Ubuntu Lambda Stack, TensorFlow, OpenCV, OpenMPI, OpenISS, MARF, OpenFOAM
• Commercial tools, subject to licensing, Fluent, MATLAB, Ansys, and many others.

All this infrastructure is continuously maintained by dedicated and professional AITS staff for sysadmin, applications, storage, and networking needs.

In alignment with the University’s digital strategy and open learning, GCS ENCS Network, Security, and HPC (NAG) group began to release some HPC resources for Speed publicly, including job submission script samples. The GCS HPC users / community are encouraged to contribute their own scripts, tricks, and hints via pull requests or report issues with the existing ones on our GitHub page:

● GCS NAG Speed GitHub Page

For more information, please e-mail: rt-ex-hpc@encs.concordia.ca

There are ongoing plans and work in progress for future expansion of Speed compute, GPU, and storage capabilities.