HPC clusters
There are multiple HPC clusters with thousands of cores and hundreds of CPUs. They can be used for serial, parallel, GPU and large memory workloads. HPC clusters are configured with a variety of pre-installed software you can use for your research projects. Cluster computing is easy to use and simple to access:
- Sign in to a shared HPC cluster.
- Choose the software you want to run.
- Upload your data onto the cluster. (n.b. or your calculations will not compute)
- Submit your jobs into a queue and your computations will run when it's your turn.
Details
- Type
- HPC service
- For
- Principle investigators (PIs) and sponsored users
- Cost
- Free
Use cases
HPC clusters can be used in a wide variety of ways. Many researchers use HPC clusters for batch processing and running computational workloads like:
- artificial intelligence (AI), machine learning (ML) and deep learning
- big data and data mining
- natural language processing (NLP) and text mining
- fluid dynamics simulations
- computational chemistry, particle physics and astrophysics simulations
- weather, land and ocean simulations
- bioinformatics jobs
Similarly, researchers use the clusters for collaboration:
- sharing data within a research group
- data storage and archiving
Examples
- The Owens lab analyzes genomic data of plants and animals. They run software to identify genomic differences between samples and uses alliance resources to assemble genomes for new species.
- The XCITE lab runs large parallel Monte Carlo simulations of medical imaging systems.
Cluster storage
Cluster storage is available for researchers using an Alliance HPC cluster. This type of storage makes your data accessible from the compute resources you're using for your research. It features:
- up to 10TB per research group in long term storage
- additional storage options available for temporary data and archived data
- automatic nightly backups
- storage in secure data centres in Canada
- HPC cluster computing integration
- support for multiple ways to transfer data
- file sharing options
Self-help resources
HPC cluster resources are free for all Canadian researchers. Every Alliance member gets access to the same default computing allocation. This means users have a standard priority when waiting for their submitted computation requests to be processed. For high resource requests, or when resources are requested frequently, users will have to wait longer for their jobs to be processed.
Examples of high resource requests are:
- Using many computing cores simultaneously (e.g. 48+)
- Long computing times (24h+)
- GPUs (demand is high!)
- More than 10TB of storage required.
If your computations require a significant amount of resources, and/or you will be requesting resources very frequently, you can submit an application to the resource allocation competition (RAC). We've put together some resources to help you with your RAC application.
You can login to the HPC clusters by command line using SSH or through a web browser using JupyterHub. Both options use your Alliance account information.
For example, if you needed to access software on the Cedar cluster the login credential would be allianceusername@cedar.computecanada.ca
There are multiple ways to store your data in cluster storage. The Alliance uses Globus, which is a file transfer service used to move research data between cluster storage and other devices. Globus includes a web portal and desktop client.
Both options require some technical configuration. The Alliance support wiki has walkthrough instructions for using the web portal and setting up the desktop client.
Get help now
Technical support for HPC clusters and cluster storage is provided through the Alliance. Your support ticket might be resolved by research computing staff from other Canadian universities.
Cluster storage related tickets can also be sent to the Alliance's Globus support team.
For more tech support topics, see the Alliance tech support wiki page.