Update on HPC Scavenge Partition 7/9

Dear HPC Users,

This week, the HPC team has been busy installing our new Discovery cluster. During the installation, we have had to decommission a number of nodes that had become too old (10-12 years) for adequate performance. This decrease in nodes means that the pool of available nodes in our scavenge partition has also decreased. As such, we have made changes to this partition’s resource limits.

The maximum resource limits for the scavenge partition have been updated as follows:

· Maximum core count per user: 200
· Maximum number of running jobs per user: 10
· Maximum number of jobs submitted to the queue by a single user: 50

For more information on the scavenge partition, please see this page on our website: https://hpcc.usc.edu/resources/documentation/scavenge/

We will send out another announcement once the new system installation has been completed, and we will notify you of any other changes in the meantime.

If you have any questions, please direct them to hpc@usc.edu.

Thank you,

USC Center for High-Performance Computing

University of Southern California
3434 South Grand Avenue
Los Angeles, CA 90089-0706

Are there any plans to have a Scavenge Partition for the new Discovery system?

@geneyu There are no plans to include a scavenge partition on Discovery. The condo compute nodes are actually not accessible from Discovery, only from Endeavour. However, we will be adding more compute nodes to Discovery later in the year.