HPC resources available
Storage
PIs can request a maximum of 40 TB of project storage and 100 TB of nearline storage in any General Purpose cluster without submitting a RAC application. The storage resources can be requested all in one cluster or split across multiple clusters, but the total amount across all clusters must not exceed 40 TB of project storage or 100 TB of nearline storage. Resources allocated via RAS will be available in the Default Resource Allocation Project (RAP).
|
Cluster |
Project storage |
Nearline storage |
|
Cedar, Graham, Béluga, Narval |
Max 40 TB |
Max 100 TB |
|
Niagara, HPSS |
RAS storage not available |
RAS storage not available |
How to request storage resources
Only PIs can request storage resources via RAS. To do so, please send an email to support@tech.alliancecan.ca with the appropriate justification when required.
Please note: Some storage resources are made available in the Default RAP to PIs immediately after creating an Alliance account on CCDB. These resources are also available to all the users sponsored by that PI. Visit this page for details about the storage resources available by default.
CPU
CPU resources are available for opportunistic use to all research groups with an active Alliance account. The great majority of jobs submitted this way get executed, albeit with sometimes a lower priority than jobs submitted with a RAC allocation.
Starting on April 1, 2024, research groups with an Alliance account needing CPU resources may be able to use, on average, up to 200 core years on each cluster with their Default RAP. Note that 200 core years is a variable target: it is not a reservation nor a cap, which means that research groups could utilize more or less than that amount depending on the shape and size of the jobs and on the overall utilization of the cluster.
Most research groups are able to meet their need for CPU resources by submitting jobs opportunistically with their Default RAP and without having to apply for the RAC. Jobs with large memory requirements may take longer to run: in those cases, applying for RAC may be the best option if the amount of CPU resources needed exceeds the minimum required to apply.
Please read about Allocations and compute scheduling. This information can help you better understand how jobs are scheduled in our clusters and how CPU usage is charged.
GPU
GPU resources are available for opportunistic use to all research groups with an active Alliance account.
The demand for GPUs is increasing quickly with the advances in Artificial Intelligence. The availability of GPU resources varies greatly over the year, becoming more limited in periods preceding major conferences.
We cannot therefore guarantee any amount of resources available to each group for opportunistic use, especially during times of high demand. Users with a RAC award should be able to consistently use their allocated amount.
Cloud resources available
PIs may request access to modest amounts of cloud resources at any time and allow others with an Alliance account to access these resources within the limits detailed in the cloud RAS documentation. You may request RAS cloud resources using this form.
If you have questions or need help, please contact cloud@tech.alliancecan.ca.