|This is a detailed information page about queues and job submission on Cy-Tera.|
The queues implemented on Cy-Tera are the following:
- cpuq - for allocation of cpu jobs only
- gpuq - for allocation of gpu jobs only
- cpui - for interactive cpu jobs
- gpui - for interactive gpu jobs
- cpub - for allocation of cpu jobs only, for projects whose CPU quota is exhausted
- gpub - for allocation of gpu jobs only, for projects whose GPU quota is exhausted
- specialq - for jobs that exceed the cpuq limits
The following tables indicate for each queue how many nodes can be used and the respective maximum walltime per node range.
node range max walltime 1-16 24 hrs 17-32 14 hrs 33-64 8 hrs
node range max walltime 1-4 24 hrs 5-8 14 hrs 9-16 8 hrs
node range max walltime 1-4 1 hr
node range max walltime 1-2 1 hr
node range max walltime 1-64 24 hrs
node range max walltime 1-16 24 hrs
By prior arrangement only, users may submit jobs that exceed limits to the specq. Users must contact the system's administrators through email@example.com to require to submit jobs in that special queue, justifying their request. Jobs submitted to the specq will be started running manually from the system's administrator.
To submit a job on Cy-Tera users have to specify in their script file to which queue the job must be submitted.
- For access to cpu nodes the following line must be added in the script file:
- For access to gpu nodes the following line must be added in the script file:
|The above lines in the script file change according to the queue the job is submitted|
In addition, to submit a job on Cy-Tera msub is used instead of qsub. Hence, users have to type the command:
To submit interactive jobs on Cy-Tera users have to type the command:
The name of the queue must be specified, otherwise the job won't run.