next up previous contents
Next: Requesting Multiple Nodes and Up: Requesting Resources - Reference Previous: Requesting Walltime   Contents

Requesting Multiple Nodes and cpus


Directive:


#PBS -l nodes=<Number of Slices>:ppn=<Size of Slices>

Example:

#!/bin/bash
#PBS -l nodes=2:ppn=4

echo "Script begins here"


Command Line:


qsub -l nodes=<Number of Slices>:ppn=<Size of Slices> script



Example:

qsub -l nodes=2:ppn=4 blast.sh


Praticum:

	[jdpoisso@umms-amino blast_test]$ qsub blast.sh 
	2500607.umms-amino.ccmb.med.umich.edu 
	[jdpoisso@umms-amino blast_test]$ qsub -l nodes=2:ppn=4 blast.sh 
	2500608.umms-amino.ccmb.med.umich.edu 
	[jdpoisso@umms-amino blast_test]$ qstat -f 2500607 
	Job Id: 2500607.umms-amino.ccmb.med.umich.edu 
	    Job_Name = blast.sh 
	<- output omitted ->
	    \textbf{exec_host = compute-6-18/5+compute-6-18/4+compute-6-18/3+compute-6-18/2+co mpute-6-17/5+compute-6-17/4+compute-6-17/3+compute-6-17/2} 
        <- output omitted ->
	    \textbf{Resource_List.nodect = 2} 
	    \textbf{Resource_List.nodes = 2:ppn=4}
	<- output omitted ->
	[jdpoisso@umms-amino blast_test]$ qstat -f 2500608 
	Job Id: 2500608.umms-amino.ccmb.med.umich.edu 
	    Job_Name = blast.sh 
	<- output omitted ->
	    \textbf{exec_host = compute-6-15/5+compute-6-15/4+compute-6-15/3+compute-6-15/0+co mpute-6-13/5+compute-6-13/4+compute-6-13/3+compute-6-13/0} 
	<- output omitted ->
	    \textbf{Resource_List.nodect = 2} 
	    \textbf{Resource_List.nodes = 2:ppn=4} 
	<- output omitted ->
	[jdpoisso@umms-amino blast_test]$


Note - exec_host : The exec host line includes the location of each task assigned to the job. It takes the form of node_name / cpu_number, adding each node name and cpu number together until you reach the total number of tasks selected. Tasks is equal to the number of nodes (nodes) multiplied by the processors per node (ppn).


Note: The preferred method of requesting multiple cores and nodes is to use the nodes : processors per node (nodes=X:ppn=X) method. There are other methods and features that can be used when requesting multiple tasks. Some of these methods are discussed in the advanced section, while others are deprecated or unique to certain cluster configurations. If the method of selection is relevant to your cluster, you should reference any cluster specific documentation available.



next up previous contents
Next: Requesting Multiple Nodes and Up: Requesting Resources - Reference Previous: Requesting Walltime   Contents
2010-08-27