The BiG Grid distributed facilities are deployed at twelve universities and academic hospitals throughout the Netherlands.
The sites participating in BiG Grid have agreed in the Operational Steering Team to define a consistent set of queues on the computing resources. This plan aims to improve the uniformity of the computing resources at the various sites towards the users. All sites will define a set of queues with identical names and identical properties, that will replace (some of) the current historical queue names and their properties. The historic queue names and properties were inconsistent among the sites.
The table below presents the queues that may be present on every site and the maximum duration of a job (as wall clock time).
Queue Name | Wall time | Access | Remark |
|---|---|---|---|
infra | 0:30 | ops, dteam, pvier | mandatory |
express | 0:30 | all supported VOs | optional |
short | 4:00 | all supported VOs | mandatory |
medium | 36:00 | all supported VOs | default local users |
long | 72:00 | per request | optional |
Note that certain queues are optional, meaning that it is up to the sites to deploy them; however, if they are deployed, they will have the properties listed above. Access to the optional long queue may be granted per request of the VOs. In addition to the queues defined above, sites may offer additional queues to selected VOs with arbitrary properties.
All sites will inform their users of planned changes to the queue setup.
How does this affect users?
- Users who do not explicitly submit jobs to a specific queue, there should not be any impact;
- Users who put a statement in the .jdl file to select a specific queue may have to change the queue name in the .jdl file;
- Users who directly submit jobs to a CE / queue via the commandglite-wms-job-submit using the option "--resource <CE>:2119/jobmanager-pbs-<QUEUE>"may have to change the queue name.
The Life Science Grid is managed by SARA, and documentation is maintained at https://hpcv.projects.sara.nl/wiki/index.php/Life_Science_Grid.