hpc:hpc_tubit:clust_conf:usage

Unterschiede

Hier werden die Unterschiede zwischen zwei Versionen angezeigt.

Link zu dieser Vergleichsansicht

Beide Seiten der vorigen Revision Vorhergehende Überarbeitung
hpc:hpc_tubit:clust_conf:usage [2019/06/24 12:20] – kraushpc:hpc_tubit:clust_conf:usage [2019/06/24 12:46] (aktuell) – kraus
Zeile 1: Zeile 1:
 +====== Usage guidelines ======
  
 +
 +  * <html> <span style="font-family:monospace;font-size:140%;font-weight:bold"><em>Fast <span style="color:red">local SSD disk scratch available with gpu nodes (gpu[001-021])</span> (mounted at /tmp).</em></span> </html>
 +
 +  * <html> <span style="font-family:monospace;font-size:140%;font-weight:bold"><em>With <span style="color:red">exclusive jobs on large SMP nodes, temporary files could just be placed into /dev/shm/&lt;path/to/your/folder&gt;</span> (in-memory tmpfs).</em></span> </html>
 +
 +  * <html> <span style="font-family:monospace;font-size:140%;font-weight:bold"><em>Please especially <span style="color:red">avoid copying around TB large rwf and chk files from Gaussian runs</span>. Please remove temporary files from scratch at the end of job runs.</em></span> </html>
 +
 +  * <html> <span style="font-family:monospace;font-size:140%;font-weight:bold"><em>Please put <span style="color:red">volatile job data at /scratch/&lt;account-name&gt;</span> or in the working directory of your job and <span style="color:red">delete temporary data at the end of job runs.</span></em></span> </html>
 +
 +  * <html> <span style="font-family:monospace;font-size:140%;font-weight:bold"><em>Please carefully calculate <span style="color:red">the memory limit for your jobs, so that optimal scheduling of your jobs can happen.</em></span> </html>
 +
 +  * <html> <span style="font-family:monospace;font-size:140%;font-weight:bold"><em>There is <span style="color:red">no virtual memory limit</span> applied to your jobs, only <span style="color:red">real memory usage is accounted</span>. This is subject to eventual change in future.</em></span> </html>
 +
 +  * <html> <span style="font-family:monospace;font-size:140%;font-weight:bold"><em>Defining virtual memory limits: The standard memory allocation parameter <span style="font-family:courier">--mem</span> <span style="color:red">defines the total amount of memory available to all processes on one node</span>. For advanced usage, it's also possible to define the amount of memory allocated per process reps. cpu (c. f. <span style="font-family:courier">man 1 sbatch</span> or <span style="font-family:courier">man 1 srun</span>).</em></span> </html>
 +
 +  * <html> <span style="font-family:monospace;font-size:140%;font-weight:bold"><em> In general, hyper-threading should not be used. <span style="color:red">Please do not set the parameter <span style="font-family:courier">--threads-per-core</span> or <span style="font-family:courier">--cpus-per-task</span>, unless you really know what you do</span>.</em></span> </html>
 +
 +
 +\\
 +<WRAP rightalign>
 +date of revision: 06-24-2019 © kraus
 +</WRAP>
  • hpc/hpc_tubit/clust_conf/usage.txt
  • Zuletzt geändert: 2019/06/24 12:46
  • von kraus