On XSEDE, how are compute jobs charged?

Computational services on the Extreme Science and Engineering Discovery Environment (XSEDE) are allocated and charged in service units (SUs). SUs are defined locally on each system, with conversion factors among systems based on High-Performance LINPACK (HPL) benchmark results. To calculate the conversion of SUs from one XSEDE digital service to the equivalent amount of SUs on another, use the XSEDE SU Conversion Calculator.

Current XSEDE supercomputers have complex multi-core and memory hierarchies. Each digital service has a specific configuration that determines how many cores can be dedicated to a job without slowing the code (and other user and system codes). Each Service Provider defines for its system the minimum number of SUs charged for a job running in the default batch queue (calculated by multiplying the job's wall time by the number of cores it used).

For the following XSEDE digital services, the minimum charge in SUs for a job is calculated as follows:

Digital service Minimum charge
Comet (SDSC) wall time x 1 core

Note: Node-exclusive jobs will be charged for using 24 cores (wall time x 1 node x 24 cores).
Darter (NICS) wall time x [number of nodes] x 32
Gordon (SDSC) wall time x 16 cores
Stampede (TACC) wall time x [number of nodes] x 16

Note: Each Stampede node can be assigned to only one user at a time; consequently, each user's job accrues wall time for all 16 cores per node. Use of each node's Phi coprocessors and NVIDIA GPUs are "free" components for which the user is not charged. For jobs running in the largemem queue, the charge is doubled (wall time x [number of nodes] x 32).
SuperMIC (LSU) wall time x [number of cores]
Wrangler (TACC/IU) 1 SU per node-hour

Note: Jobs on Wrangler are charged differently than jobs on other compute services. Wrangler uses the node-hour metric to measures the use of a node and its storage over a period of time. The user is charged for reserved nodes, whether they are used for computation or not. Reservation cancellations are refunded for the unused portions. Jobs submitted without reservations are charged for the duration of the job.

The actual charge will depend on the specific requirements of your job (e.g., the mapping of the cores across the machine, or the priority you wish to obtain). Also, not all XSEDE allocations are requested in SUs (e.g., storage requests are made in terabytes). Consult each system's user guide for details. If you have questions, contact the XSEDE Help Desk.

For more about XSEDE compute, advanced visualization, storage, and special purpose systems, see the Resources Overview, Systems Monitor, and User Guides. For scheduled maintenance windows, outages, and other announcements related to XSEDE digital services, see User News.

This document was developed with support from National Science Foundation (NSF) grants 1053575 and 1548562. Any opinions, findings, conclusions, or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the NSF.