We are pleased to announce that the CANFAR Science Platform transition to new bare-metal hardware on a new cluster is complete. You should see more capacity right away, and if you encounter issues, contact us at support@canfar.net.
Posted Aug 05, 2026 - 18:48 UTC
In progress
Scheduled maintenance is currently in progress. We will provide updates as necessary.
Posted Aug 05, 2026 - 17:00 UTC
Scheduled
Dear CANFAR Community,
We are pleased to announce that the CANFAR Science Platform is moving to new bare-metal hardware on a new cluster. You should see more capacity right away, with further resource growth and improved /arc storage I/O speed in the coming months.
What improves for you - Roughly +40% CPU, +60% memory, and +50% local /scratch across the platform. - Much more GPU capacity — on the order of 14× more GPU slices, so GPU-enabled sessions should be far easier to obtain and faster for many workloads. - Sessions run on modern bare-metal CPUs with DDR5 memory, instead of the previous virtualized machines which is better for concurrent use and vectorized science codebases. - Alongside the new machines, much of the other data-centre infrastructure has been replaced, which should improve resiliency and uptime.
The Switchover The move to the new hardware will occur on Wednesday 5 August 2026 at ~10:00 PST / 17:00 UT. The process is expected to take ~2 hours. At that point all existing CANFAR Sessions will be terminated. Please shut down all your sessions by this time, then start new sessions on the new cluster when it is available. There will be no change to how you start and interact with CANFAR sessions.
For questions about this process, or for support with CANFAR issues, contact us at support@canfar.net.