Computing Resources

As a member of the uWaterloo Applied Math Fluids Group you have access to a fairly substantial array of computing resources. These are correct as of March 2018.

Compute Canada

Compute Canada is perhaps the largest supplier of computing services available to us, and it provides significant computing power. To begin, you will need to create a Compute Canada account (follow instructions on the Compute Canada page). Account creation will require approval from your supervisor. The main components of Compute Canada that are relevant to us are: SHARCNet, SciNet, and WestGrid.

SHARCNet

SHARCNet has many computing clusters. However, there are two clusters to which we have a degree of priority access: Orca and Graham

Help with these systems can be accessed through the SHARCNet ticket system.

Orca (listed to be decommissioned)

Orca is now (as of Graham) a legacy system. Fluids contributed 512 processors to the compute cluster and, as a result, has priority on those compute nodes. For information on using Orca, see Orca_Tips. For information on the Orca system and hardware, see the official SHARCNet page here and here.

Graham

Graham is a significantly larger computing cluster than Orca, and provides the opportunity to run much larger simulations. As of writing this, the group has been awarded dedicated computation time. For information on using Graham, see Graham Tips. For information on the Graham system and hardware, see the official documentation.

SciNet

???

MFCF-administered Fluids Systems

The Math Faculty Computing Facility (MFCF) maintains several Linux/Unix servers that belong to the Fluids group. These servers are much smaller in scale than the Compute Canada clusters (typically tens of processors), but they are hosted locally and can be very useful tools for running smaller simulations or getting your code working properly before scaling up to Compute Canada. The MFCF staff are also very helpful regarding software installation and general computing problems / questions.

kazan

This system is fairly old, and unless you specifically need to use it you may be better served by some of the other machines. However, some information can be found on our page about Compiling on winisk/kazan.

hood

more information can be found here.

bow, minnewanka, waterton

These are new (2017) systems with high-speed interconnects and are managed through the SLURM scheduler (see Graham Tips for some useful SLURM-related commands). More information can be found here.

kesagami, kuujjua

These new (2018) systems, not yet in production, are intended primarily to run the IGW model.

faculty-wide machines

MFCF also operates a faculty-wide shared computing environment comprising numerous Linux and Windows servers and a small number of GPU servers. Even if you do not use those machines, the central file server is a smart place to store a copy of your important work, such as your thesis in progress, for safety. The central file service "files.math" is accessible via the Mac mini provided to you by the department, as well as via ssh/scp to the shared Linux machines.

Lab Systems

Lab machines are maintained by lab members (currently Aaron Coutino), not MFCF. Accounts should exist for Fluid Lab members (if you don't have one but would like one, ask Aaron / your supervisor). Some standard software is installed.

Note: these machines do not have a queue system, so please compute responsibly and do not swamp the machines.

Belize

As of 18 March 2018, Belize has been taken off-line. Data can still be accessed, but requires using the machine directly / in-person.

Boogaloo

Boogaloo can be accessed either in-person in the Fluids Lab or via ssh with yourUserID@boogaloo.math.uwaterloo.ca. See our wiki page for information on the available hardware. Boogaloo has both CPU and GPU capabilities.

Onyx

Onyx is a Windows machine that is primarily intended for visualization and is new as of March 2018. Both VisIt and ParaView are installed and should be GPU-aware. This machine should be used in-person to perform high-powered visualization of your datasets. Onyx is not intended for heavy computation, but is capable to running CUDA models.