From: Francesca Verdier (FVerdier_at_lbl.gov)
Date: 08/10/2006
Dear NERSC users, DOE and NERSC have set aside 2.5 million MPP (SP POWER3-equivalent) hours for projects to investigate code efficiencies on 64 or more Seaborg nodes (1,024+ processors) and to encourage running codes in these large node configurations. This program is available to projects that have not yet scaled their codes to 1,024+ processors, or that would like to investigate scaling new codes. NERSC intends to migrate this scaling program from Seaborg to the recently announced NERSC-5 Cray system, Franklin, once it is in production. Franklin will have 19,344 compute CPUs. Reimbursements for 1,024+ processor jobs will be made towards the beginning of each month so long as you submit a paragraph to NERSC explaining the science results achieved with these large runs and so long as you gather performance information for your large runs using IPM. For more information on IPM see: http://www.nersc.gov/nusers/resources/SP/ipm.php IPM must be invoked for each run that gets reimbursed. To request participation in the scaling reimbursement program send a message to NERSC Consulting explaining briefly what you hope to accomplish by participating in the program (both the science, briefly, and potential code improvements): Web: http://help.nersc.gov Email: consult_at_nersc_dot_gov Please note that NERSC will continue to discount all Seaborg jobs run on 48 or more nodes. The current discount for such jobs is 50%. Sincerely, -- Francesca Verdier email: fverdier@lbl.gov Associate Manager, NERSC Services phone: 510-486-7193
This archive was generated by hypermail 2.1.6 : 08/21/2008 PDT