Page History

Running jobs on gadi requires users to have sufficient compute hours available. 

These compute hours are granted to projects rather than directly to the user. 

To view and use an an allocation, users must first be associated with the project. Once they are a member of the project they can run the command

$ nci_account -P <project code>

replacing <project code> with the project they are enquiring about. The image on the right shows information about project ab01. 

It has grant allocation of 471.00 KSU, and it has used 406.30 KSU from that grant. Leaving 64.70 KSU in the allocation, as you can see in the 'Avail' row.

SU is an abbreviation of Service Unit, and KSU is an abbreviation of Kilo-service unit, 1000 SU. Jobs that run on the Gadi normal queue are charged 2 SU to run for an hour on a single core, with a memory allocation of 4 GiB. Different Gadi queues are charged at different rates. 

The grant amount will always be equal to the sum of used, reserved, and avail. Each time that a job is submitted for a project, the estimated walltime will be removed from the avail and placed into the reserve. Upon completion of this job the actual walltime will be deducted and moved to used and any remainder will be returned as available SUs. 

If there is not enough SUs available for a job to run it will wait in the queue indefinitely. in this case, the lead CI should contact their allocation scheme manager and apply for more resources. If you are unsure who to contact you can access more information about the project by running the command

$ nci_account -P <project code> -v

replacing <project code> with the project that you are enquiring into.

As you can see on the right, using '-v', or the the verbose command, gives a more indepth look into the project and allows us to see the stakeholders that are associated. This should help you confirm who to contact in this situation.