Dear NERSC users, Perlmutter has been an extremely valuable system for many NERSC users during its current “early science” phase. The system has delivered 4.2 million GPU Node Hours and 2 million CPU Node Hours to science so far in 2022. Because the system has been very popular during the non-charged “early science” phase, average queue wait times are long (many days) for all but the smallest jobs. In order to make sure that system usage aligns with the allocation award decisions made by Office of Science program managers through the 2022 ERCAP process, we will start charging for Perlmutter usage following the successful completion of the Lustre scratch file system upgrade, which is planned to complete on August 31, 2022. As long as your project’s balance remains above zero, jobs that run under your project’s account will receive priority scheduling. Jobs associated with projects that have exhausted their GPU Node Hour and/or CPU Node Hour allocation may still be submitted and run using the lower-priority “overrun” QOS. Please note that, while Perlmutter is not yet in its final configuration, NERSC has made every effort to make it available and productive for users while we continue to work on updating its high speed network and integrating additional CPU-only nodes. These updates and changes are moving into their final stages, please see https://docs.nersc.gov/systems/perlmutter/ for current status and configuration. Regards, Richard Gerber Richard Gerber, Ph.D. Senior Science Advisor | HPC Department Head National Energy Research Scientific Computing Center <http://www.nersc.gov/> | Berkeley Lab <http://www.lbl.gov/> ragerber@lbl.gov | phone: (510) 486-6820 fax: 510-486-4316 _______________________________________________ Users mailing list Users@nersc.gov