To ensure that computing resources are distributed equitably and to discourage irresponsible use users' jobs on the Atos HPCF are charged for the resources they have used against a project account. Each project account is allocated a number of System Billing Units(SBU) at the beginning of each accounting year (1 January to 31 December). You can monitor your usage in the HPC SBU accounting portal.

ECS usage

Resources used on the ECGATE Class Service (ECS) cluster will not be charged to your HPCF resource allocation.

How much is an SBU?

The accounting on the HPCF is currently based on the total elapsed time for the cores allocated to serial or parallel batch jobs. To reflect the current system configuration and performance the formula to compute SBU changed on 1 January 2024 to

SBU = P * N * T

where:

P = 424074479 / (8090 * 128 * 86400)

is a fixed proportionality factor, adjusted so that the entire Atos system will deliver 424,074,479 SBU per day on a "typical"
ECMWF job mix on the HPC system (8090 * 128 is the total number of physical cores used to run batch jobs and 86400
the number of seconds in a day).

N

Is the number of (physical) cores allocated to a job. 

T

is the elapsed time (real time in seconds) recorded for a job.

Examples:

  1. A serial job using 1 hour (3600 seconds) of elapsed time will be charged about 17.06 SBU.
  2. A parallel job requesting 2 nodes, running for 3 hours elapsed time will be charged 13104.90 SBU.

Compared to 2023 the SBU cost of the same job in 2024 is ~10.8% less . 


SBU = P * N * T

where:

P = 470410408 / (8098 * 128 * 86400)

is a fixed proportionality factor, adjusted so that the entire Atos system will deliver 470,410,408 SBU per day on a "typical"
ECMWF job mix on the HPC system (8098 * 128 is the total number of physical cores used to run batch jobs and 86400
the number of seconds in a day).

N

Is the number of (physical) cores allocated to a job. 

T

is the elapsed time (real time in seconds) recorded for a job.

Examples:

  1. A serial job using 1 hour (3600 seconds) of elapsed time will be charged about 18.91 SBU.
  2. A parallel job requesting 2 nodes, running for 3 hours elapsed time will be charged 14522.43 SBU.


End of job information

At the end of your jobs a short report is appended to the corresponding job output file, which amongst other useful information, includes how many SBUs where used and the account they were charged on:

[ECMWF-INFO -ecepilog] ----------------------------------------------------------------------------------------------------
[ECMWF-INFO -ecepilog] This is the ECMWF job Epilogue
[ECMWF-INFO -ecepilog] +++ Please report issues using the Support portal +++ 
[ECMWF-INFO -ecepilog] +++ https://support.ecmwf.int                     +++ 
[ECMWF-INFO -ecepilog] ----------------------------------------------------------------------------------------------------
[ECMWF-INFO -ecepilog] Run at 2022-08-24T09:09:01 on ac
[ECMWF-INFO -ecepilog] JobName                   : myjob
[ECMWF-INFO -ecepilog] JobID                     : 37015044
[ECMWF-INFO -ecepilog] Submit                    : 2022-08-24T09:08:55
[ECMWF-INFO -ecepilog] Start                     : 2022-08-24T09:08:55
[ECMWF-INFO -ecepilog] End                       : 2022-08-24T09:09:01
[ECMWF-INFO -ecepilog] QueuedTime                : 0.0 
[ECMWF-INFO -ecepilog] ElapsedRaw                : 6 
[ECMWF-INFO -ecepilog] ExitCode                  : 0:0 
[ECMWF-INFO -ecepilog] DerivedExitCode           : 0:0 
[ECMWF-INFO -ecepilog] State                     : COMPLETED
[ECMWF-INFO -ecepilog] Account                   : myaccount
[ECMWF-INFO -ecepilog] QOS                       : np
[ECMWF-INFO -ecepilog] User                      : user
[ECMWF-INFO -ecepilog] StdOut                    : /home/user/slurm-37015044.out
[ECMWF-INFO -ecepilog] StdErr                    : /home/user/slurm-37015044.out
[ECMWF-INFO -ecepilog] NNodes                    : 1 
[ECMWF-INFO -ecepilog] NCPUS                     : 256 
[ECMWF-INFO -ecepilog] SBU                       : 4.083
[ECMWF-INFO -ecepilog] ----------------------------------------------------------------------------------------------------

SBU simulator

You can see how much your job would cost in terms of SBUs by copying and pasting here your job headers.

Defining how long your job will run

Use the SLURM time directive for the job duration. For example, for a job that needs to run for 12 hours:

#SBATCH --time=12:00:00


SBUs Nodes Cores Time Notes

What are my project accounts?

All users on the HPCF have at least one project account where their usage will be charged. You can check what accounts you have access to with the command:

account [-l uid]

If more than one account is listed, then by default, job usage will be debited to the first account, which is then considered the primary account.

If you wish to debit the usage of a job to an alternative account (out of your list of assigned accounts) you may do so by using the following Slurm directive in your job header:

#SBATCH --account=your_preferred_account

How do I monitor my usage?

To monitor your SBU usage (Cray HPCF until 31 Oct 2022, Atos HPCF from 1 Nov 2022) charged against allocations please visit the HPC SBU accounting portal and login with your relevant computer user-ID. 

You will find the accumulated SBU usage for your different accounts from the beginning of the accounting year as well as the allocations granted for each of them.

Please note that usage data is only updated once a day and displayed with a delay, so resources used by your latest jobs will not be shown immediately.

This web application is still under development and further enhancements and bug fixes will be implemented. Please send any questions or feedback via our Support Portal.

Accounting and account management

Computing resource allocation rules

Member States are allocated computing resources for their own purposes according to the Council guidelines on distribution of computer resources. Member State Computing Representatives may distribute those resources among projects registered at ECMWF as they feel appropriate. Should a redistribution of resources between the projects of a Member State be required during the course of a year, this may be carried out by the Member State Computing Representative or another designated "privileged user" (see section Administration of project accounts in the Member States  below).

Special Projects

Special Projects are experiments or investigations of a scientific or technical nature undertaken by one or more Member States, likely to be of interest to the general scientific or technical community. In the spring of each year, Directors of national meteorological services are invited to submit requests for Special Project resources to be available for use in the following year. Requests are considered by the Council at its winter session, in conjunction with recommendations from the Scientific Advisory Committee. Since Special Project accounts are separate from Member State resources, their account administration is performed by ECMWF.

For more details see http://www.ecmwf.int/en/research/special-projects.

Daily Accounting Controls

Information is extracted daily from HPCF report files to produce a cumulative total of resources used per user and per account. Once usage reaches 100% no further use of that account should be made. The automatic sending of email alerts when a user specified usage threshold is reached, 90% by default, is not yet active. 

Registering a new account

New accounts are registered by ECMWF only on the authority either of an ECMWF Department Head (for ECMWF staff accounts) or of a Member State Computing Representative (for Member State accounts) or the Principal Investigator (for Special Projects).

For further information get in touch via our Support Portal.

Administration of project accounts in the Member States

Member States may nominate "privileged users" to perform limited administration of Member State accounts using the HPC SBU accounting portal. By default, the Centre will grant this status to all Member State Computing Representatives with a registered user identifier, but Member State Computing Representatives may propose an alternative person and/or nominate a backup if they wish.

A "privileged user" may be authorised to modify all or a reduced set of Member States' accounts. For instance, Member States are free to allocate control to several "privileged" users, each having control of a particular sub-node. It should be noted, however, that the nomination of privileged users should be kept to a minimum.

For security reasons, the establishment of new project accounts and the addition of new users remains under the control of ECMWF.