25.06.2015 Views

Administering Platform LSF - SAS

Administering Platform LSF - SAS

Administering Platform LSF - SAS

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Supported Resource Usage Limits and Syntax<br />

Format<br />

Normalized run<br />

time<br />

<strong>Platform</strong><br />

MultiCluster<br />

Thread limit<br />

run_limit is in the form [hour:]minute, where minute can be greater than<br />

59. 3.5 hours can either be specified as 3:30 or 210.<br />

The run time limit is normalized according to the CPU factor of the<br />

submission host and execution host. The run limit is scaled so that the job has<br />

approximately the same run time for a given run limit, even if it is sent to a<br />

host with a faster or slower CPU.<br />

For example, if a job is submitted from a host with a CPU factor of 2 and<br />

executed on a host with a CPU factor of 3, the run limit is multiplied by 2/3<br />

because the execution host can do the same amount of work as the submission<br />

host in 2/3 of the time.<br />

If the optional host name or host model is not given, the run limit is scaled<br />

based on the DEFAULT_HOST_SPEC specified in the lsb.params file. (If<br />

DEFAULT_HOST_SPEC is not defined, the fastest batch host in the cluster is<br />

used as the default.) If host or host model is given, its CPU scaling factor is<br />

used to adjust the actual run limit at the execution host.<br />

The following example specifies that myjob can run for 10 minutes on a<br />

DEC3000 host, or the corresponding time on any other host:<br />

% bsub -W 10/DEC3000 myjob<br />

If ABS_RUNLIMIT=Y is defined in lsb.params, the run time limit is not<br />

normalized by the host CPU factor. Absolute wall-clock run time is used for all<br />

jobs submitted with a run limit.<br />

See “CPU Time and Run Time Normalization” on page 357 for more<br />

information.<br />

For MultiCluster jobs, if no other CPU time normalization host is defined and<br />

information about the submission host is not available, <strong>LSF</strong> uses the host with<br />

the largest CPU factor (the fastest host in the cluster). The ABS_RUNLIMIT<br />

parameter in lsb.params is is not supported in either MultiCluster model; run<br />

time limit is normalized by the CPU factor of the execution host.<br />

Job syntax (bsub) Queue syntax (lsb.queues) Fomat/Units<br />

-T thread_limit THREADLIMIT=[default] maximum integer<br />

Sets the limit of the number of concurrent threads to thread_limit for the<br />

whole job. The default is no limit.<br />

Exceeding the limit causes the job to terminate. The system sends the following<br />

signals in sequence to all processes belongs to the job: SIGINT, SIGTERM, and<br />

SIGKILL.<br />

If a default thread limit is specified, jobs submitted to the queue without a joblevel<br />

thread limit are killed when the default thread limit is reached.<br />

If you specify only one limit, it is the maximum, or hard, thread limit. If you<br />

specify two limits, the first one is the default, or soft, thread limit, and the<br />

second one is the maximum thread limit.<br />

354<br />

<strong>Administering</strong> <strong>Platform</strong> <strong>LSF</strong>

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!