The HPC will be upgraded this summer, however, unlike previous years, this will be done in two distinct phases. In phase 1, the compute nodes from the 2007 acquisition will be moved from the Sliger data center to the server room in the Department of Scientific Computing (DSC) and the general access queue will be disabled. In phase 2, new compute nodes will be installed in the Sliger building.

Because the warranty will expire on the 2007 nodes, it was decided to move those nodes out of the existing HPC system. These nodes will still be available for HPC user through our moab queuing system, but will be part of special queues.

Currently, these nodes are in use by the general access queue and this queue will therefore be temporary disabled during this period. Once these nodes are moved to DSC, we will try to temporary add them back to the HPC queue until the end of phase 2. Owner queues will not be affected during phase 1 and general access users can continue to use the backfill queue.

In short:

Phase 1

• June 30: general access queue will be disabled and jobs on the 2007 nodes will be removed
• July 2: moving services will transfer the 2007 compute nodes
• (tentative) July 12: compute nodes will be back online, however without infiniband.
• Owner queues will not be affected during this phase and the backfill queue will stay open for everyone.

Phase 2.

• Mid-End August: new compute nodes will be installed. Dates will be announced once we have more information.
• The old nodes will be partitioned into "old" queues.